- developers - lists.mariadb.org

Re: [Maria-developers] df203d1daa1: MDEV-15990: REPLACE on a precise-versioned table returns duplicate key error (ER_DUP_ENTRY)
by Sergei Golubchik 28 Feb '19

28 Feb '19

Hi, Nikita! Sorry, I don't understand what are you fixing and why :( Could you please elaborate? See few minor comments below and the main question at the very end. On Dec 26, Nikita Malyavin wrote: > revision-id: df203d1daa1 (versioning-1.0.5-42-gdf203d1daa1) > parent(s): 5e89a0dc25a > author: Nikita Malyavin <nikitamalyavin(a)gmail.com> > committer: Nikita Malyavin <nikitamalyavin(a)gmail.com> > timestamp: 2018-07-21 00:14:57 +1000 > message: > > MDEV-15990: REPLACE on a precise-versioned table returns duplicate key error (ER_DUP_ENTRY) > > * handle zero value in `row_start` `trx_id` > * refuse to do more than one versioned insert on the same transaction and same row > * generalize versioned cases with and without triggers > > [closes tempesta-tech/mariadb#517] > [closes tempesta-tech/mariadb#520] > > diff --git a/mysql-test/suite/versioning/r/replace.result b/mysql-test/suite/versioning/r/replace.result > --- a/mysql-test/suite/versioning/r/replace.result > +++ b/mysql-test/suite/versioning/r/replace.result > @@ -7,10 +7,31 @@ period for system_time (row_start, row_end) > ) with system versioning; > insert t values (1, 2); > replace t values (1, 3); > -select *, current_row(row_end) as current from t for system_time all order by x; > -id x current > -1 2 0 > -1 3 1 > +replace t values (1, 3); > +# MDEV-15990: REPLACE on a precise-versioned table > +# returns duplicate key error (ER_DUP_ENTRY) > +replace t values (1, 3), (1, 30), (1, 40); > +create table log(s varchar(3)); > +create trigger tr1del before delete on t > +for each row insert into log values('DEL'); > +replace t values (1, 4), (1, 40), (1, 400); > +select *, check_row(row_start, row_end), row_start != 0 as start_nonzero > +from t for system_time all order by x, row_end; > +id x check_row(row_start, row_end) start_nonzero > +1 2 HISTORICAL ROW 1 > +1 3 HISTORICAL ROW 1 > +1 3 HISTORICAL ROW 1 > +1 40 HISTORICAL ROW 1 > +1 400 CURRENT ROW 1 > +select * from log; > +s > +DEL > +DEL > +DEL better to insert OLD.x too, to be able to see where the DEL comes from > +# ensure that row_start is not duplicated > +select count(row_start) as empty from t for system_time all > +group by row_start having count(row_start) > 1; > +empty > drop table t; > create table t ( > id int unique, > diff --git a/sql/sql_insert.cc b/sql/sql_insert.cc > --- a/sql/sql_insert.cc > +++ b/sql/sql_insert.cc > @@ -1969,23 +1952,22 @@ int write_record(THD *thd, TABLE *table,COPY_INFO *info) > TRG_ACTION_BEFORE, TRUE)) > goto before_trg_err; > > - if (!table->versioned(VERS_TIMESTAMP)) > + if (table->versioned(VERS_TRX_ID) && table->vers_write) > + { can table->vers_write be false here? I suspect it can never be false for REPLACE, so I'd write it as if (table->versioned(VERS_TRX_ID)) { DBUG_ASSERT(table->vers_write); > + bitmap_set_bit(table->write_set, table->vers_start_field()->field_index); > + table->vers_start_field()->store(0, false); > + } > + > + if (!table->versioned()) > error= table->file->ha_delete_row(table->record[1]); > else > - { > - store_record(table, record[2]); > - restore_record(table, record[1]); > - table->vers_update_end(); > error= table->file->ha_update_row(table->record[1], > table->record[0]); > - restore_record(table, record[2]); > - } > - if (unlikely(error)) > + if (unlikely(error && error != HA_ERR_RECORD_IS_THE_SAME)) > goto err; > - if (!table->versioned(VERS_TIMESTAMP)) > + if (likely(!error)) > info->deleted++; > - else > - info->updated++; > + error= 0; > if (!table->file->has_transactions()) > thd->transaction.stmt.modified_non_trans_table= TRUE; > if (table->triggers && > @@ -1995,6 +1977,17 @@ int write_record(THD *thd, TABLE *table,COPY_INFO *info) > trg_error= 1; > goto ok_or_after_trg_err; > } > + if (table->versioned(VERS_TIMESTAMP) && table->vers_write) same > + { > + store_record(table, record[2]); > + error = vers_insert_history_row(table); why are you doing that (insert + goto after_trg_n_copied_inc) instead of going the normal REPLACE code path? > + restore_record(table, record[2]); > + if (unlikely(error)) > + goto err; > + } > + if (table->versioned()) > + goto after_trg_n_copied_inc; > + > /* Let us attempt do write_row() once more */ > } > } > diff --git a/storage/innobase/handler/ha_innodb.cc b/storage/innobase/handler/ha_innodb.cc > index db755077e0b..139c43ced38 100644 > --- a/storage/innobase/handler/ha_innodb.cc > +++ b/storage/innobase/handler/ha_innodb.cc > @@ -8916,6 +8916,19 @@ ha_innobase::update_row( > upd_t* uvect = row_get_prebuilt_update_vector(m_prebuilt); > ib_uint64_t autoinc; > > + bool force_insert= table->versioned_write(VERS_TRX_ID) > + && table->vers_start_id() == 0; hmm, so you're using vers_start_id() == 0 as a special hack to signal REPLACE to the storage engine... > + > + if (force_insert) { > + table->vers_start_field()->store(trx->id); > + > + Field *start= table->vers_start_field(); > + trx_id_t old_trx_id= uint8korr(start->ptr_in_record(old_row)); > + > + /* avoid double versioned insert on the same transaction and same row */ > + force_insert= (old_trx_id != trx->id); okay, I'm confused. What was wrong in the old code? Normally REPLACE is DELETE + INSERT, why was it not enough for InnoDB and you needed some special logic here to handle the REPLACE case? > + } > + > /* Build an update vector from the modified fields in the rows > (uses m_upd_buf of the handle) */ > Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

2 1

Re: [Maria-developers] 773d24d457a: MDEV-15951 system versioning by trx id doesn't work with partitioning
by Sergei Golubchik 28 Feb '19

28 Feb '19

Hi, Nikita! On Dec 25, Nikita Malyavin wrote: > revision-id: 773d24d457a (versioning-1.0.6-100-g773d24d457a) > parent(s): b26736cdb11 > author: Nikita Malyavin <nikitamalyavin(a)gmail.com> > committer: Nikita Malyavin <nikitamalyavin(a)gmail.com> > timestamp: 2018-12-21 03:13:47 +1000 > message: > > MDEV-15951 system versioning by trx id doesn't work with partitioning > > Fix partitioning for trx_id-versioned tables. > `partition by hash`, `range` and others now work. > `partition by system_time` is forbidden. > Currently we cannot use row_start and row_end in `partition by`, because > insertion of versioned field is done by engine's handler, as well as > row_start/row_end's value set up, which is a transaction id -- so it's > also forbidden. > > The drawback is that it's now impossible to use `partition by key()` > without parameters for such tables, because it references row_start and > row_end implicitly. > > * add handler::vers_can_native() > * drop Table_scope_and_contents_source_st::vers_native() > * drop partition_element::find_engine_flag as unused > * forbid versioning partitioning for trx_id as not supported > * adopt vers tests for trx_id partitioning > * forbid any row_end referencing in `partition by` clauses, > including implicit `by key()` > > diff --git a/sql/sql_partition.cc b/sql/sql_partition.cc > index 9e6c333d3c9..48bcb828d56 100644 > --- a/sql/sql_partition.cc > +++ b/sql/sql_partition.cc > @@ -341,7 +341,16 @@ static bool set_up_field_array(THD *thd, TABLE *table, > while ((field= *(ptr++))) > { > if (field->flags & GET_FIXED_FIELDS_FLAG) > + { > + if (table->versioned(VERS_TRX_ID) > + && unlikely(field->flags & VERS_SYSTEM_FIELD)) > + { > + my_error(ER_VERS_NOT_SUPPORTED, MYF(0), > + "Using ROW START/END for patitioning", "transactional"); This is incorrect usage of error message parameters. The error message can be translated, parameters can be reordered - you should not use English as parameter values. Consider: transactional cистемно-версионированные таблицы не поддерживают Using ROW START/END for patitioning But looks like ER_VERS_NOT_SUPPORTED is not used anywhere at the moment. So you can change it to be anything you want, even remove all parameters: Transactional system versioned tables do not support partitioning by ROW START or ROW END > + DBUG_RETURN(TRUE); > + } > num_fields++; > + } > } > if (unlikely(num_fields > MAX_REF_PARTS)) > { > diff --git a/sql/table.cc b/sql/table.cc > index ce7a34a8fe2..0c2ff1eea5a 100644 > --- a/sql/table.cc > +++ b/sql/table.cc > @@ -1776,7 +1776,7 @@ int TABLE_SHARE::init_from_binary_frm_image(THD *thd, bool write, > goto err; > DBUG_PRINT("info", ("Columns with system versioning: [%d, %d]", row_start, row_end)); > versioned= VERS_TIMESTAMP; > - vers_can_native= plugin_hton(se_plugin)->flags & HTON_NATIVE_SYS_VERSIONING; > + vers_can_native= handler_file->vers_can_native(thd); Interesting. How does that work? Your ha_partition::vers_can_native() is bool can= !thd->lex->part_info || thd->lex->part_info->part_type != VERSIONING_PARTITION; for (uint i= 0; i < m_tot_parts && can; i++) can= can && m_file[i]->vers_can_native(thd); return can; But it doesn't look like thd->lex->part_info->part_type is initialized at this point in time. Try to add to your test case flush tables; select * from t1; Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

2 2

Re: [Maria-developers] bc7d600a27f: MDEV-15951 system versioning by trx id doesn't work with partitioning
by Sergei Golubchik 28 Feb '19

28 Feb '19

Hi, Nikita! On Dec 21, Nikita Malyavin wrote: > > > @@ -4938,6 +4938,7 @@ int ha_create_table(THD *thd, const char *path, > > > !create_info->tmp_table(); > > > > > > share.frm_image= frm; > > > + share.orig_table_name= table_name; > > > > why? if orig_table_name is NULL, error_table_name() will use table_name, > > so there's never need to make orig_table_name==table_name. > > > > and I don't understand all your other manipulations with > > orig_table_name. Could you please explain that? > > It affects ALTER TABLE error messages. table_name will contain temporary name, > and error_table_name() will output "(temporary)", which is rather inconvenient. > AFAICS, orig_table_name purpose is exactly for fixing that, but some > corner cases were not covered to this moment. I don't understand that. You've made a lot of changes in many places passing orig table name around, it's difficult to see what corner cases you were trying to fix. Could you please fix it in a separate commit? With tests and an explanation in a commit comment, describing what exactly corner cases were not covered. > > > // open an frm image > > > if (share.init_from_binary_frm_image(thd, write_frm_now, Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

2 1

[Maria-developers] Exposing Item::store_native and val_native to the SQL layer
by Eric Herman 25 Feb '19

25 Feb '19

Hello Bar, As you have done the work on store_native and val_native, I would especially like to know your thoughts with regards to how to expose these to the SQL layer. In my particular case, I'd like to re-work the "Add FLOAT and DOUBLE data interchange functions" patch ( https://github.com/ericherman/mariadb-server/commit/dcc7ccf65689bd7d84bef37… ) to use these new API functions. I've only just started on this task ( https://github.com/ericherman/mariadb-server/tree/10.4-eherman-float-native ) and first thing that I need to figure out is how to write the tests. In commit 34eb98387f8f46a80fb053081dbe20d415f23b39 ( MDEV-13995 MAX(timestamp) returns a wrong result near DST change ), I see that in tests the SQL calls various time functions which ultimately call into these functions, yet I do not see any way to access these two new functions directly. Are you thinking of adding new SQL functions to expose the underlying bytes to the user? If not, I have a sense that casting to and from BINARY could expose the native byte arrays, but I'm not sure if that's the direction you imagine. If you tell me which direction to go, I will try to make that happen. Cheers, -Eric Other foo_native related commits: * commit a8a757c6bb32bbf291afdf33df861127489889ab wl#173 - temporal types with sub-second resolution * commit c353b2a8fc10e16107ee6c7e26877f4243d4eaef MDEV-17979 Assertion `0' failed in Item::val_native upon SELECT with timestamp, NULLIF, GROUP BY --

2 1

Re: [Maria-developers] [Commits] dadf5f78411: Fix few compiler warnings.
by Marko Mäkelä 20 Feb '19

20 Feb '19

On February 20, 2019, Jan Lindström wrote: > revision-id: dadf5f78411b8f46ac53b8ea12c08ec317d9aab4 (mariadb-10.1.38-7-gdadf5f78411) > parent(s): 346e46089621e6951e076c82ed5690aa23dcb5fe > author: Jan Lindström > committer: Jan Lindström > timestamp: 2019-02-20 08:10:26 +0200 > message: > > Fix few compiler warnings. The commit message should include the ticket number. I would also suggest fixing the grammar: MDEV-18659: Fix a few compiler warnings Could you make the title more specific? > dict_table_rename_in_cache > dict_create_add_foreign_id > Use of strncpy caused string truncation warnings and is > replaced with memcpy. Specifically, which warnings were issued? I am concerned that this could be replacing dubious code with worse. With memcpy(), you should be sure that the source strings is really safe to access up to the specified length. With strncpy(), the copying will stop at the first NUL character in the output, or at the specified length, whichever comes first. > fts_fetch_index_words > dfield_get_len() can return UNIV_SQL_NULL (=ULINT32_UNDEFINED) > that is too big for used short datatype. Replaced with correct > ulint. Is NULL really an allowed value there? I do not think that the "words" or "tokens" (the keys of the inverted index) should ever be NULL. Likewise, if the entire fulltext-indexed column is NULL, we should not be inserting anything into the inverted index. Please ensure that these cases are covered by test cases in mysql-test-run, and share the detailed results: especially, which pieces of code are handling the NULL values. The data that is being fed to fts_fetch_index_words() comes from this query to the InnoDB internal SQL parser: " SELECT word\n" " FROM $table_name\n" " WHERE word > :word\n" " ORDER BY word;\n" These tables appear to be created in fts_create_one_index_table(), and it seems to me that it is PRIMARY KEY(word,first_doc_id). Strangely, I do not see DATA_NOT_NULL in the column creation, so it is possible that we are reserving NULL flags for the "word" column. But, InnoDB is not prepared to deal with NULL values in the primary key. Lots of things would break if "word" were NULL. Please add assertions to the places where "word" is being written or read in these tables. Here is the call that forgets to declare "word" as NOT NULL: dict_mem_table_add_col(new_table, heap, "word", charset == &my_charset_latin1 ? DATA_VARCHAR : DATA_VARMYSQL, field->col->prtype, FTS_MAX_WORD_LEN_IN_CHAR * unsigned(field->col->mbmaxlen)); > row_rename_table_for_mysql > Target string was exactly same length as maximum length > of copied string. Fixing this does make sense. > diff --git a/storage/innobase/dict/dict0dict.cc b/storage/innobase/dict/dict0dict.cc > index 06c6c3effab..04b14e3ae06 100644 > --- a/storage/innobase/dict/dict0dict.cc > +++ b/storage/innobase/dict/dict0dict.cc > @@ -1845,7 +1845,7 @@ dict_table_rename_in_cache( > > ulint db_len; > char* old_id; > - char old_name_cs_filename[MAX_TABLE_NAME_LEN+20]; > + char old_name_cs_filename[MAX_TABLE_NAME_LEN+20]=""; > uint errors = 0; > > /* All table names are internally stored in charset What is the initialization needed for? It was not mentioned in the commit comment. Note that this would likely not only initialize the first byte of the string, but all bytes. What kind of machine code is this generating? memcpy()? Or is the compiler smart enough to recognize that the stack variable is going to be zero-filled, and will emit something like memset() instead? Either way, it is not good to write MAX_TABLE_NAME_LEN+20 bytes if only 1 byte (a terminating NUL character at the start of the array) would suffice. And you would have to convince me that this initialization is necessary in the first place. Best regards, Marko Mäkelä Lead Developer InnoDB MariaDB Corporation

1 0

Re: [Maria-developers] 46734eb1632: MDEV-371 long unique
by Sergei Golubchik 19 Feb '19

19 Feb '19

Hi, Sachin! On Feb 14, Sachin Setiya wrote: > revision-id: 46734eb1632 (mariadb-10.4.1-97-g46734eb1632) > parent(s): acf4269a5b1 > author: Sachin <sachin.setiya(a)mariadb.com> > committer: Sachin <sachin.setiya(a)mariadb.com> > timestamp: 2019-02-12 19:01:17 +0530 > message: > diff --git a/mysql-test/main/long_unique.result b/mysql-test/main/long_unique.result > --- /dev/null > +++ b/mysql-test/main/long_unique.result > @@ -0,0 +1,1391 @@ please add a test with 64 indexes. Say, one long unique and 63 normal indexes on ints or whatever. The point is to show that long unique creates internally only _one_ index, not two, so it does not limit user's ability to create all MAX_INDEXES indexes > +#Structure of tests > +#First we will check all option for > +#table containing single unique column > +#table containing keys like unique(a,b,c,d) etc > +#then table containing 2 blob unique etc > +set @allowed_packet= @@max_allowed_packet; > +#table with single long blob column; > +create table t1(a blob unique ); > +insert into t1 values(1),(2),(3),(56),('sachin'),('maria'),(123456789034567891),(null),(null),(123456789034567890); > +#blob with primary key not allowed > +create table t2(a blob,primary key(a(10000))); > +ERROR 42000: Specified key was too long; max key length is 1000 bytes > +create table t3(a varchar(10000) primary key); > +ERROR 42000: Specified key was too long; max key length is 1000 bytes > +insert into t1 values(2); > +ERROR 23000: Duplicate entry '2' for key 'a' > +#table structure; > +desc t1; > +Field Type Null Key Default Extra > +a blob YES UNI NULL > +show create table t1; > +Table Create Table > +t1 CREATE TABLE `t1` ( > + `a` blob DEFAULT NULL, > + UNIQUE KEY `a` (`a`) > +) ENGINE=MyISAM DEFAULT CHARSET=latin1 > +show keys from t1; > +Table t1 > +Non_unique 0 > +Key_name a > +Seq_in_index 1 > +Column_name a > +Collation A > +Cardinality NULL > +Sub_part NULL > +Packed NULL > +Null YES > +Index_type HASH > +Comment > +Index_comment > + > +MyISAM file: DATADIR/test/t1 > +Record format: Packed > +Character set: latin1_swedish_ci (8) > +Data records: 10 Deleted blocks: 0 > +Recordlength: 12 > + > +table description: > +Key Start Len Index Type > +1 12 8 multip. ulonglong NULL > +select * from information_schema.columns where table_schema = 'test' and table_name = 't1'; > +TABLE_CATALOG TABLE_SCHEMA TABLE_NAME COLUMN_NAME ORDINAL_POSITION COLUMN_DEFAULT IS_NULLABLE DATA_TYPE CHARACTER_MAXIMUM_LENGTH CHARACTER_OCTET_LENGTH NUMERIC_PRECISION NUMERIC_SCALE DATETIME_PRECISION CHARACTER_SET_NAME COLLATION_NAME COLUMN_TYPE COLUMN_KEY EXTRA PRIVILEGES COLUMN_COMMENT IS_GENERATED GENERATION_EXPRESSION > +def test t1 a 1 NULL YES blob 65535 65535 NULL NULL NULL NULL NULL blob UNI select,insert,update,references NEVER NULL > +select * from information_schema.statistics where table_schema = 'test' and table_name = 't1'; > +TABLE_CATALOG TABLE_SCHEMA TABLE_NAME NON_UNIQUE INDEX_SCHEMA INDEX_NAME SEQ_IN_INDEX COLUMN_NAME COLLATION CARDINALITY SUB_PART PACKED NULLABLE INDEX_TYPE COMMENT INDEX_COMMENT > +def test t1 0 test a 1 a A NULL NULL NULL YES HASH > +select * from information_schema.key_column_usage where table_schema= 'test' and table_name= 't1'; > +CONSTRAINT_CATALOG def > +CONSTRAINT_SCHEMA test > +CONSTRAINT_NAME a > +TABLE_CATALOG def > +TABLE_SCHEMA test > +TABLE_NAME t1 > +COLUMN_NAME a > +ORDINAL_POSITION 1 > +POSITION_IN_UNIQUE_CONSTRAINT NULL > +REFERENCED_TABLE_SCHEMA NULL > +REFERENCED_TABLE_NAME NULL > +REFERENCED_COLUMN_NAME NULL > +# table select we should not be able to see db_row_hash_column; > +select * from t1 order by a; > +a > +NULL > +NULL > +1 > +123456789034567890 > +123456789034567891 > +2 > +3 > +56 > +maria > +sachin > +select db_row_hash_1 from t1; > +ERROR 42S22: Unknown column 'db_row_hash_1' in 'field list' > +#duplicate entry test; > +insert into t1 values(2); > +ERROR 23000: Duplicate entry '2' for key 'a' > +insert into t1 values('sachin'); > +ERROR 23000: Duplicate entry 'sachin' for key 'a' > +insert into t1 values(123456789034567891); > +ERROR 23000: Duplicate entry '123456789034567891' for key 'a' > +select * from t1 order by a; > +a > +NULL > +NULL > +1 > +123456789034567890 > +123456789034567891 > +2 > +3 > +56 > +maria > +sachin > +insert into t1 values(11),(22),(33); > +insert into t1 values(12),(22); > +ERROR 23000: Duplicate entry '22' for key 'a' > +select * from t1 order by a; > +a > +NULL > +NULL > +1 > +11 > +12 > +123456789034567890 > +123456789034567891 > +2 > +22 > +3 > +33 > +56 > +maria > +sachin > +insert into t1 values(repeat('s',4000*10)),(repeat('s',4001*10)); > +insert into t1 values(repeat('m',4000*10)),(repeat('m',4000*10)); > +ERROR 23000: Duplicate entry 'mmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmmm' for key 'a' > +insert into t1 values(repeat('m',4001)),(repeat('m',4002)); > +truncate table t1; > +insert into t1 values(1),(2),(3),(4),(5),(8),(7); > + > +MyISAM file: DATADIR/test/t1 > +Record format: Packed > +Character set: latin1_swedish_ci (8) > +Data records: 7 Deleted blocks: 0 > +Recordlength: 12 > + > +table description: > +Key Start Len Index Type > +1 12 8 multip. ulonglong NULL > +#now some alter commands; > +alter table t1 add column b int; > +desc t1; > +Field Type Null Key Default Extra > +a blob YES UNI NULL > +b int(11) YES NULL > +show create table t1; > +Table Create Table > +t1 CREATE TABLE `t1` ( > + `a` blob DEFAULT NULL, > + `b` int(11) DEFAULT NULL, > + UNIQUE KEY `a` (`a`(65535)) hmm, again, (65535) appeared from nowhere > +) ENGINE=MyISAM DEFAULT CHARSET=latin1 > +insert into t1 values(1,2); > +ERROR 23000: Duplicate entry '1' for key 'a' > +insert into t1 values(2,2); > +ERROR 23000: Duplicate entry '2' for key 'a' > +select db_row_hash_1 from t1; > +ERROR 42S22: Unknown column 'db_row_hash_1' in 'field list' > +#now try to change db_row_hash_1 column; > +alter table t1 drop column db_row_hash_1; > +ERROR 42000: Can't DROP COLUMN `db_row_hash_1`; check that it exists > +alter table t1 add column d int , add column e int , drop column db_row_hash_1; ... > +insert into t2 values(1,1); > +insert into t2 values(2,1); > +drop table t2; > +#not null test //todo solve warnings is that solved? I don't see any warnings > +create table t1(a blob unique not null); > +desc t1; > +Field Type Null Key Default Extra > +a blob NO UNI NULL > +show create table t1; > +Table Create Table > +t1 CREATE TABLE `t1` ( > + `a` blob NOT NULL, > + UNIQUE KEY `a` (`a`) > +) ENGINE=MyISAM DEFAULT CHARSET=latin1 > +insert into t1 values(1); > +insert into t1 values(3); > +insert into t1 values(1); > +ERROR 23000: Duplicate entry '1' for key 'a' > +drop table t1; > +create table t1(a int primary key, b blob unique , c blob unique not null); > +insert into t1 values(1,1,1); > +insert into t1 values(2,1,2); > +ERROR 23000: Duplicate entry '1' for key 'b' > +insert into t1 values(3,3,1); > +ERROR 23000: Duplicate entry '1' for key 'c' > +drop table t1; > +create table t1 (a blob unique not null, b blob not null, c blob not null, unique(b,c)); > +desc t1; > +Field Type Null Key Default Extra > +a blob NO UNI NULL > +b blob NO MUL NULL > +c blob NO NULL > +show create table t1; > +Table Create Table > +t1 CREATE TABLE `t1` ( > + `a` blob NOT NULL, > + `b` blob NOT NULL, > + `c` blob NOT NULL, > + UNIQUE KEY `a` (`a`), > + UNIQUE KEY `b` (`b`,`c`) > +) ENGINE=MyISAM DEFAULT CHARSET=latin1 > +insert into t1 values (1, 2, 3); > +insert into t1 values (2, 1, 3); > +insert into t1 values (2, 1, 3); > +ERROR 23000: Duplicate entry '2' for key 'a' > +drop table t1; > +#partition > +create table t1(a blob unique) partition by hash(a); > +ERROR HY000: A BLOB field is not allowed in partition function > +#key length > 2^15 -1 > +create table t1(a blob, unique(a(100001))); > +ERROR 42000: Specified key was too long; max key length is 1000 bytes better to say "max length is 32767 bytes" or, may be, a new error message "key segment length is 32767 bytes"? > +set @@GLOBAL.max_allowed_packet= @allowed_packet; > diff --git a/mysql-test/main/long_unique_debug.result b/mysql-test/main/long_unique_debug.result > --- /dev/null > +++ b/mysql-test/main/long_unique_debug.result > @@ -0,0 +1,574 @@ > +#In this test case we will check what will happen in the case of hash collision > +SET debug_dbug="d,same_long_unique_hash"; > +create table t1(a blob unique); > +insert into t1 values('xyz'); > +insert into t1 values('abc'); > +insert into t1 values('sachin'); > +insert into t1 values('sachin'); > +ERROR 23000: Duplicate entry 'sachin' for key 'a' > +insert into t1 values('maria'); > +insert into t1 values('maria'); > +ERROR 23000: Duplicate entry 'maria' for key 'a' I think you can use SHOW STATUS LIKE 'Handler_read_key' (or some other Handler_% variable) to verify here that there was indeed a hash collision - that is, this variable counts row reads and you have an extra row read here, so it should be visible. Try: add FLUSH STATUS at the beginning of the test, SHOW STATUS LIKE 'Handler_read_key' after errors and run the test with and without d,same_long_unique_hash if there'll be a difference, let's keep SHOW STATUS in the test. > +drop table t1; > +create table t1(a blob unique, b blob unique); > +insert into t1 values('xyz', 11); > +insert into t1 values('abc', 22); > +insert into t1 values('sachin', 1); > +insert into t1 values('sachin', 4); > +ERROR 23000: Duplicate entry 'sachin' for key 'a' > +insert into t1 values('maria', 2); > +insert into t1 values('maria', 3); > +ERROR 23000: Duplicate entry 'maria' for key 'a' > +drop table t1; > +create table t1(a blob , b blob , unique(a,b)); > +insert into t1 values('xyz', 11); > +insert into t1 values('abc', 22); > +insert into t1 values('sachin', 1); > +insert into t1 values('sachin', 1); > +ERROR 23000: Duplicate entry 'sachin-1' for key 'a' > +insert into t1 values('maria', 2); > +insert into t1 values('maria', 2); > +ERROR 23000: Duplicate entry 'maria-2' for key 'a' > +drop table t1; > +##Internal State of long unique tables > +SET debug_dbug="d,print_long_unique_internal_state"; > +create table t1 ( a blob unique); > +Warnings: > +Note 1 Printing Table state, It will print table fields, fields->offset,field->null_bit, field->null_pos and key_info ... > + > +Printing Table keyinfo > + > +table->s->reclength 19 > +table->s->fields 2 Rather unorthodox, but fine, as you like :) > diff --git a/mysql-test/main/long_unique_innodb.result b/mysql-test/main/long_unique_innodb.result > --- /dev/null > +++ b/mysql-test/main/long_unique_innodb.result > @@ -0,0 +1,121 @@ > +create table t1(a blob unique) engine= InnoDB; > +insert into t1 values('RUC'); > +insert into t1 values ('RUC'); > +ERROR 23000: Duplicate entry 'RUC' for key 'a' > +drop table t1; > +#test for concurrent insert of long unique in innodb > +create table t1(a blob unique) engine= InnoDB; > +show create table t1; > +Table Create Table > +t1 CREATE TABLE `t1` ( > + `a` blob DEFAULT NULL, > + UNIQUE KEY `a` (`a`) > +) ENGINE=InnoDB DEFAULT CHARSET=latin1 > +connect 'con1', localhost, root,,; > +connect 'con2', localhost, root,,; > +connection con1; > +set transaction isolation level READ UNCOMMITTED; > +start transaction; > +insert into t1 values('RUC'); > +connection con2; > +set transaction isolation level READ UNCOMMITTED; > +start transaction; > +insert into t1 values ('RUC'); > +ERROR 23000: Duplicate entry 'RUC' for key 'a' > +connection con1; > +commit; > +set transaction isolation level READ COMMITTED; > +start transaction; > +insert into t1 values('RC'); > +connection con2; > +commit; > +set transaction isolation level READ COMMITTED; > +start transaction; > +insert into t1 values ('RC'); > +ERROR 23000: Duplicate entry 'RC' for key 'a' Is this your HA_ERR_LOCK_WAIT_TIMEOUT that you replace with HA_ERR_FOUND_DUPP_KEY? I think HA_ERR_LOCK_WAIT_TIMEOUT is correct here. Compare with normal indexes, there should be HA_ERR_LOCK_WAIT_TIMEOUT with normal indexes too. Because the second INSERT should wait for the first to commit or rollback. And if the first doesn't neither, the second times out. Looks reasonable. > +commit; > +connection con1; > +commit; > +set transaction isolation level REPEATABLE READ; > +start transaction; > +insert into t1 values('RR'); > +connection con2; > +commit; > +set transaction isolation level REPEATABLE READ; > +start transaction; > +insert into t1 values ('RR'); > +ERROR 23000: Duplicate entry 'RR' for key 'a' Same here, I guess > +connection con1; > +commit; > +set transaction isolation level SERIALIZABLE; > +start transaction; > +insert into t1 values('S'); > +connection con2; > +commit; > +set transaction isolation level SERIALIZABLE; > +start transaction; > +insert into t1 values ('S'); > +ERROR 23000: Duplicate entry 'S' for key 'a' and here > +commit; > +connection con1; > +commit; > +select * from t1; > +a > +RUC > +RC > +RR > +S > +drop table t1; > +create table t1(a blob unique) engine=Innodb; > +connection con1; > +set transaction isolation level READ UNCOMMITTED; > +start transaction; > +insert into t1 values('RUC'); > +connection con2; > +set transaction isolation level READ UNCOMMITTED; > +start transaction; > +insert into t1 values ('RUC');; > +connection con1; > +rollback; > +connection con2; > +commit; > +connection con1; > +set transaction isolation level READ COMMITTED; > +start transaction; > +insert into t1 values('RC'); > +connection con2; > +set transaction isolation level READ COMMITTED; > +start transaction; > +insert into t1 values ('RC');; > +connection con1; > +rollback; > +connection con2; > +commit; > +connection con1; > +set transaction isolation level REPEATABLE READ; > +start transaction; > +insert into t1 values('RR'); > +connection con2; > +set transaction isolation level REPEATABLE READ; > +start transaction; > +insert into t1 values ('RR');; > +connection con1; > +rollback; > +connection con2; > +commit; > +connection con1; > +set transaction isolation level SERIALIZABLE; > +start transaction; > +insert into t1 values('S'); > +connection con2; > +set transaction isolation level SERIALIZABLE; > +start transaction; > +insert into t1 values ('S');; > +connection con1; > +rollback; > +connection con2; > +commit; > +connection default; > +drop table t1; > +disconnect con1; > +disconnect con2; > diff --git a/mysql-test/main/long_unique_innodb.test b/mysql-test/main/long_unique_innodb.test > --- /dev/null > +++ b/mysql-test/main/long_unique_innodb.test > @@ -0,0 +1,131 @@ > +--source include/have_innodb.inc > +--source include/big_test.inc instead of big_test, better set the innodb lock timeout to something small for the first half of your test. and please make sure all tests pass. This one doesn't even start for me now in your branch. > + > +create table t1(a blob unique) engine= InnoDB; > +insert into t1 values('RUC'); > +--error ER_DUP_ENTRY > +insert into t1 values ('RUC'); > +drop table t1; > + > +--echo #test for concurrent insert of long unique in innodb > +create table t1(a blob unique) engine= InnoDB; > +show create table t1; > +connect ('con1', localhost, root,,); > diff --git a/mysql-test/main/mdev-504.test b/mysql-test/main/mdev-504.test > --- a/mysql-test/main/mdev-504.test > +++ b/mysql-test/main/mdev-504.test > @@ -1,4 +1,3 @@ > ---source include/not_valgrind.inc and restore this line too, please. Or, if it make sense to remove it, make sure you'll push it in a separate commit, not in MDEV-371 > --source include/no_protocol.inc > > set @save_use_stat_tables=@@global.use_stat_tables; > diff --git a/sql/field.h b/sql/field.h > --- a/sql/field.h > +++ b/sql/field.h > @@ -553,7 +553,7 @@ class Virtual_column_info: public Sql_alloc > name.str= NULL; > name.length= 0; > }; > - ~Virtual_column_info() {} > + ~Virtual_column_info(); Hmm, I didn't find ~Virtual_column_info() destructor anywhere. Where is it now? > enum_vcol_info_type get_vcol_type() const > { > return vcol_type; > diff --git a/sql/item_func.h b/sql/item_func.h > --- a/sql/item_func.h > +++ b/sql/item_func.h > @@ -1019,6 +1019,19 @@ class Item_long_func: public Item_int_func > }; > > > +class Item_func_hash: public Item_int_func > +{ > +public: > + Item_func_hash(THD *thd, List<Item> &item): Item_int_func(thd, item) > + {} > + longlong val_int(); > + bool fix_length_and_dec(); > + const Type_handler *type_handler() const { return &type_handler_long; } > + Item *get_copy(THD *thd) > + { return get_item_copy<Item_func_hash>(thd, this); } > + const char *func_name() const { return "HASH"; } perhaps "<hash>" ? Like other internal items, e.g. <in_optimizer> or <rowid> > +}; > + > class Item_longlong_func: public Item_int_func > { > public: > diff --git a/sql/sql_show.cc b/sql/sql_show.cc > --- a/sql/sql_show.cc > +++ b/sql/sql_show.cc > @@ -2341,7 +2341,7 @@ int show_create_table(THD *thd, TABLE_LIST *table_list, String *packet, > } > } > packet->append(')'); > - store_key_options(thd, packet, table, key_info); > + store_key_options(thd, packet, table, &table->key_info[i]); why? table->s->key_info[i] doesn't show correct options? I've tried to revert this change and long_unique test passed. > if (key_info->parser) > { > LEX_CSTRING *parser_name= plugin_name(key_info->parser); > @@ -6586,15 +6586,20 @@ static int get_schema_stat_record(THD *thd, TABLE_LIST *tables, > table->field[8]->set_notnull(); > } > KEY *key=show_table->key_info+i; > - if (key->rec_per_key[j]) > + if (key->rec_per_key[j] && key->algorithm != HA_KEY_ALG_LONG_HASH) > { > ha_rows records= (ha_rows) ((double) show_table->stat_records() / > key->actual_rec_per_key(j)); > table->field[9]->store((longlong) records, TRUE); > table->field[9]->set_notnull(); > } When you'll do the optimizer support, I think you can print rec_per_key[] here for your virtual HASH key. But it's not important now in this patch, so let's not delay the push because of it. > - const char *tmp= show_table->file->index_type(i); > - table->field[13]->store(tmp, strlen(tmp), cs); > + if (key->algorithm == HA_KEY_ALG_LONG_HASH) > + table->field[13]->store(STRING_WITH_LEN("HASH"), cs); > + else > + { > + const char *tmp= show_table->file->index_type(i); > + table->field[13]->store(tmp, strlen(tmp), cs); > + } > } > if (!(key_info->flags & HA_FULLTEXT) && > (key_part->field && > diff --git a/sql/sql_table.cc b/sql/sql_table.cc > --- a/sql/sql_table.cc > +++ b/sql/sql_table.cc > @@ -4106,12 +4127,27 @@ mysql_prepare_create_table(THD *thd, HA_CREATE_INFO *create_info, > if (key->type == Key::UNIQUE && !(key_info->flags & HA_NULL_PART_KEY)) > unique_key=1; > key_info->key_length=(uint16) key_length; > - if (key_length > max_key_length && key->type != Key::FULLTEXT) > + if (key_length > max_key_length && key->type != Key::FULLTEXT && > + !is_hash_field_needed) > { > my_error(ER_TOO_LONG_KEY,MYF(0),max_key_length); > DBUG_RETURN(TRUE); > } > > + if (is_hash_field_needed) > + { > + if (key_info->algorithm != HA_KEY_ALG_UNDEF && > + key_info->algorithm != HA_KEY_ALG_HASH ) > + { > + my_error(ER_TOO_LONG_KEY, MYF(0), file->max_key_length()); > + DBUG_RETURN(TRUE); > + } > + key_info->algorithm= HA_KEY_ALG_LONG_HASH; > + } > + // If user forces hash index for storage engine other then memory > + else if (key_info->algorithm == HA_KEY_ALG_HASH && > + create_info->db_type->db_type != DB_TYPE_HEAP) Please, no. No engine type checks. Better add a new engine capability flag HA_CAN_HASH_KEYS (see HA_CAN_RTREEKEYS for example). And only MEMORY engine will have it for now. > + key_info->algorithm= HA_KEY_ALG_LONG_HASH; > if (validate_comment_length(thd, &key->key_create_info.comment, > INDEX_COMMENT_MAXLEN, > ER_TOO_LONG_INDEX_COMMENT, > @@ -8319,6 +8355,11 @@ mysql_prepare_alter_table(THD *thd, TABLE *table, > enum Key::Keytype key_type; > LEX_CSTRING tmp_name; > bzero((char*) &key_create_info, sizeof(key_create_info)); > + if (key_info->algorithm == HA_KEY_ALG_LONG_HASH) > + { > + key_info->flags|= HA_NOSAME; > + key_info->algorithm= HA_KEY_ALG_UNDEF; Why HA_KEY_ALG_UNDEF? > + } > key_create_info.algorithm= key_info->algorithm; > /* > We copy block size directly as some engines, like Area, sets this > @@ -10262,6 +10304,8 @@ copy_data_between_tables(THD *thd, TABLE *from, TABLE *to, > to->s->default_fields= 0; > for (Field **ptr=to->field ; *ptr ; ptr++) > { > + if ((*ptr)->flags & LONG_UNIQUE_HASH_FIELD) > + continue; Really? I'd think it would skip all virtual fields anyway, so you should not need to have any special checks for your HASH field. > def=it++; > if (def->field) > { > @@ -10485,14 +10529,23 @@ copy_data_between_tables(THD *thd, TABLE *from, TABLE *to, > if ((int) key_nr >= 0) > { > const char *err_msg= ER_THD(thd, ER_DUP_ENTRY_WITH_KEY_NAME); > + KEY *long_key= NULL; > if (key_nr == 0 && to->s->keys > 0 && > (to->key_info[0].key_part[0].field->flags & > AUTO_INCREMENT_FLAG)) > err_msg= ER_THD(thd, ER_DUP_ENTRY_AUTOINCREMENT_CASE); > + if (key_nr <= to->s->keys && to->key_info[key_nr].algorithm > + == HA_KEY_ALG_LONG_HASH) > + { > + long_key= to->key_info + key_nr; > + setup_keyinfo_hash(long_key); > + } > print_keydup_error(to, > key_nr >= to->s->keys ? NULL : > &to->key_info[key_nr], > err_msg, MYF(0)); > + if (long_key) > + re_setup_keyinfo_hash(long_key); You've already had this exactly code in handler::print_error. May be it'd be better to move it inside print_keydup_error() to have it only in one place? > } > else > to->file->print_error(error, MYF(0)); > diff --git a/sql/table.cc b/sql/table.cc > --- a/sql/table.cc > +++ b/sql/table.cc > @@ -1143,10 +1150,59 @@ bool parse_vcol_defs(THD *thd, MEM_ROOT *mem_root, TABLE *table, > pos+= expr_length; > } > > - /* Now, initialize CURRENT_TIMESTAMP fields */ > + /* Now, initialize CURRENT_TIMESTAMP and UNIQUE_INDEX_HASH_FIELD fields */ > for (field_ptr= table->field; *field_ptr; field_ptr++) > { > Field *field= *field_ptr; > + if (field->flags & LONG_UNIQUE_HASH_FIELD) > + { > + List<Item > *field_list= new (mem_root) List<Item >(); Why do you add a space before a closing angle bracket? I don't think I've see this style anywhere in our codebase. > + Item *list_item; > + KEY *key; > + uint key_index; > + for (key_index= 0; key_index < table->s->keys; key_index++) > + { > + key=table->key_info + key_index; > + if (key->algorithm == HA_KEY_ALG_LONG_HASH && > + key->key_part[key->user_defined_key_parts].fieldnr == field->field_index+ 1) > + break; > + } > + KEY_PART_INFO *keypart; > + for (uint i=0; i < key->user_defined_key_parts; i++) > + { > + keypart= key->key_part + i; > + if (!keypart->length) > + { > + list_item= new(mem_root)Item_field(thd, keypart->field); > + } > + else > + { > + list_item= new(mem_root)Item_func_left(thd, > + new (mem_root)Item_field(thd, keypart->field), > + new (mem_root) Item_int(thd, keypart->length)); > + list_item->fix_fields(thd, NULL); > + } > + field_list->push_back(list_item, mem_root); > + } > + Item_func_hash *hash_item= new(mem_root)Item_func_hash(thd, *field_list); > + Virtual_column_info *v= new (mem_root) Virtual_column_info(); > + field->vcol_info= v; > + field->vcol_info->expr= hash_item; > + uint parts= key->user_defined_key_parts; > + key->user_defined_key_parts= key->ext_key_parts= key->usable_key_parts= 1; > + key->key_part+= parts; > + > + if (key->flags & HA_NULL_PART_KEY) > + key->key_length= HA_HASH_KEY_LENGTH_WITH_NULL; > + else > + key->key_length= HA_HASH_KEY_LENGTH_WITHOUT_NULL; > + > + if (key->flags & HA_NULL_PART_KEY) > + key->key_length= HA_HASH_KEY_LENGTH_WITH_NULL; > + else > + key->key_length= HA_HASH_KEY_LENGTH_WITHOUT_NULL; copy-paste error? same if() repeated twice > + *(vfield_ptr++)= *field_ptr; > + } > if (field->has_default_now_unireg_check()) > { > expr_str.length(parse_vcol_keyword.length); > @@ -1287,6 +1366,7 @@ int TABLE_SHARE::init_from_binary_frm_image(THD *thd, bool write, > > keyinfo= &first_keyinfo; > share->ext_key_parts= 0; > + share->long_unique_table= 0; I don't think that's needed, TABLE_SHARE is probably bzero-ed by this time (otherwise there would've been a lot more initializations here) > thd->mem_root= &share->mem_root; > > if (write && write_frm_image(frm_image, frm_length)) > @@ -2162,6 +2255,51 @@ int TABLE_SHARE::init_from_binary_frm_image(THD *thd, bool write, > share->default_fields++; > } > } > + > + keyinfo= share->key_info; > + for (uint i=0; i < hash_fields; i++, field_ptr++) > + { > + LEX_CSTRING comment, name; > + uint32 flags= 0; > + Column_definition_attributes attr; > + comment.str= (char*) ""; > + comment.length=0; > + //There are n no of hash fields for n long unique key > + while(keyinfo->algorithm != HA_KEY_ALG_LONG_HASH) > + keyinfo++; > + attr.length= HA_HASH_FIELD_LENGTH; > + if (keyinfo->flags & HA_NULL_PART_KEY) > + attr.pack_flag= FIELDFLAG_MAYBE_NULL; > + attr.unireg_check= Field::NONE; > + attr.charset= share->table_charset;//STODO > + name.str= (const char *)alloc_root(&share->mem_root, LONG_HASH_FIELD_NAME_LENGTH); > + make_long_hash_field_name(&name, &fieldnames); > + fieldnames.count++; > + fieldnames.type_names[share->fields + i] = name.str; > + Record_addr addr(share->default_values + share->reclength, null_pos, null_bit_pos); > + *field_ptr= reg_field= > + attr.make_field(share, &share->mem_root, &addr, &type_handler_longlong, &name, flags); > + if (!reg_field) // Not supported field type > + goto err; > + reg_field->field_index= i + share->fields; > + reg_field->comment=comment; > + reg_field->flags|= LONG_UNIQUE_HASH_FIELD; > + reg_field->invisible= INVISIBLE_FULL; > + if (!(reg_field->flags & NOT_NULL_FLAG)) > + { > + if (!(null_bit_pos= (null_bit_pos + 1) & 7)) > + null_pos++; > + } > + share->reclength+= HA_HASH_FIELD_LENGTH; > + keyinfo++; > + } I'm sorry :( After I saw your chat with Marko I've realized a big problem with this approach. If we do it this way, it'll be very diffucult to change anything in the future. Say, you'd want to do persistent generated columns (I'm not saying you should, it's just an example) - if LONG_UNIQUE_HASH_FIELD fields are stored in the frm, you simply make generated columns persistent, and everything works. If LONG_UNIQUE_HASH_FIELD fields are generated on the fly, you'd need here to generate persistent or virtual fields to match what the actual table in the storage engine has, so here you'd still need to be able to write in the frm that some LONG_UNIQUE_HASH_FIELD are persistent. It looks like it'd be more future-proof just to write LONG_UNIQUE_HASH_FIELD fields into frm as you did in a previous patch. Sorry, for this. > + keyinfo= share->key_info; > + for (uint i =0; hash_fields && i < share->keys; i++) > + { > + if(keyinfo[i].algorithm != HA_KEY_ALG_LONG_HASH) > + for (uint j= 0; j < keyinfo->user_defined_key_parts; j++) > + keyinfo[i].key_part[j].offset+= extra_null_field_bytes; > + } > *field_ptr=0; // End marker > /* Sanity checks: */ > DBUG_ASSERT(share->fields>=share->stored_fields); > diff --git a/sql/table_cache.cc b/sql/table_cache.cc > --- a/sql/table_cache.cc > +++ b/sql/table_cache.cc > @@ -824,11 +824,9 @@ TABLE_SHARE *tdc_acquire_share(THD *thd, TABLE_LIST *tl, uint flags, > DBUG_RETURN(0); > else if (res == 1) > continue; > - > element= (TDC_element*) lf_hash_search_using_hash_value(&tdc_hash, > thd->tdc_hash_pins, hash_value, (uchar*) key, key_length); > lf_hash_search_unpin(thd->tdc_hash_pins); > - DBUG_ASSERT(element); Why? > > if (!(share= alloc_table_share(tl->db.str, tl->table_name.str, key, key_length))) > { Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

2 1

Re: [Maria-developers] 5e265de00c5: triggers fix: delete
by Sergei Golubchik 19 Feb '19

19 Feb '19

Hi, Nikita! On Feb 18, Nikita Malyavin wrote: > revision-id: 5e265de00c5 (mariadb-10.4.1-185-g5e265de00c5) > parent(s): 8ec8242e962 > author: Nikita Malyavin <nikitamalyavin(a)gmail.com> > committer: Nikita Malyavin <nikitamalyavin(a)gmail.com> > timestamp: 2019-02-18 23:48:33 +1000 > message: > > triggers fix: delete Commit comment, please. Like MDEV-16973 Application-time periods: DELETE fire triggers in the order: BEFORE DELETE, BEFORE INSERT, AFTER INSERT, AFTER DELETE. > --- > mysql-test/suite/period/r/delete.result | 74 ++++++++++++++++----------------- > sql/sql_delete.cc | 45 ++++++++++---------- > 2 files changed, 58 insertions(+), 61 deletions(-) > > diff --git a/mysql-test/suite/period/r/delete.result b/mysql-test/suite/period/r/delete.result > index f0949f5b0d9..0ac11f518f8 100644 > --- a/mysql-test/suite/period/r/delete.result > +++ b/mysql-test/suite/period/r/delete.result > @@ -35,33 +35,33 @@ id s e > select * from log_tbl order by id; > id log > 1 >DEL: 1, 1999-01-01, 2018-12-12 > -2 <DEL: 1, 1999-01-01, 2018-12-12 > -3 >INS: 1, 1999-01-01, 2000-01-01 > -4 <INS: 1, 1999-01-01, 2000-01-01 > -5 >INS: 1, 2018-01-01, 2018-12-12 > -6 <INS: 1, 2018-01-01, 2018-12-12 > +2 >INS: 1, 1999-01-01, 2000-01-01 > +3 <INS: 1, 1999-01-01, 2000-01-01 > +4 >INS: 1, 2018-01-01, 2018-12-12 > +5 <INS: 1, 2018-01-01, 2018-12-12 > +6 <DEL: 1, 1999-01-01, 2018-12-12 > 7 >DEL: 1, 1999-01-01, 2017-01-01 > -8 <DEL: 1, 1999-01-01, 2017-01-01 > -9 >INS: 1, 1999-01-01, 2000-01-01 > -10 <INS: 1, 1999-01-01, 2000-01-01 > +8 >INS: 1, 1999-01-01, 2000-01-01 > +9 <INS: 1, 1999-01-01, 2000-01-01 > +10 <DEL: 1, 1999-01-01, 2017-01-01 > 11 >DEL: 1, 2017-01-01, 2019-01-01 > -12 <DEL: 1, 2017-01-01, 2019-01-01 > -13 >INS: 1, 2018-01-01, 2019-01-01 > -14 <INS: 1, 2018-01-01, 2019-01-01 > +12 >INS: 1, 2018-01-01, 2019-01-01 > +13 <INS: 1, 2018-01-01, 2019-01-01 > +14 <DEL: 1, 2017-01-01, 2019-01-01 > 15 >DEL: 2, 1998-01-01, 2018-12-12 > -16 <DEL: 2, 1998-01-01, 2018-12-12 > -17 >INS: 2, 1998-01-01, 2000-01-01 > -18 <INS: 2, 1998-01-01, 2000-01-01 > -19 >INS: 2, 2018-01-01, 2018-12-12 > -20 <INS: 2, 2018-01-01, 2018-12-12 > +16 >INS: 2, 1998-01-01, 2000-01-01 > +17 <INS: 2, 1998-01-01, 2000-01-01 > +18 >INS: 2, 2018-01-01, 2018-12-12 > +19 <INS: 2, 2018-01-01, 2018-12-12 > +20 <DEL: 2, 1998-01-01, 2018-12-12 > 21 >DEL: 3, 1997-01-01, 2015-01-01 > -22 <DEL: 3, 1997-01-01, 2015-01-01 > -23 >INS: 3, 1997-01-01, 2000-01-01 > -24 <INS: 3, 1997-01-01, 2000-01-01 > +22 >INS: 3, 1997-01-01, 2000-01-01 > +23 <INS: 3, 1997-01-01, 2000-01-01 > +24 <DEL: 3, 1997-01-01, 2015-01-01 > 25 >DEL: 4, 2016-01-01, 2020-01-01 > -26 <DEL: 4, 2016-01-01, 2020-01-01 > -27 >INS: 4, 2018-01-01, 2020-01-01 > -28 <INS: 4, 2018-01-01, 2020-01-01 > +26 >INS: 4, 2018-01-01, 2020-01-01 > +27 <INS: 4, 2018-01-01, 2020-01-01 > +28 <DEL: 4, 2016-01-01, 2020-01-01 > 29 >DEL: 5, 2010-01-01, 2015-01-01 > 30 <DEL: 5, 2010-01-01, 2015-01-01 add a test case with no BEFORE INSERT trigger. For example, here drop the BEFORE INSERT trigger (but keep the other three) and delete something again. Please, make sure that "delete something" covers all cases, like above (left end, right end, middle, etc). > # INSERT trigger only also works > diff --git a/sql/sql_delete.cc b/sql/sql_delete.cc > index 9607718d755..d79a9deaa4d 100644 > --- a/sql/sql_delete.cc > +++ b/sql/sql_delete.cc > @@ -802,9 +796,11 @@ bool mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds, > only if there are no triggers set. > It is also meaningless for system-versioned table Fix the comment above to explain why you only look at BEFORE INSERT trigger here. > */ > - portion_of_time_through_update= !has_triggers > - && !table->versioned() > - && table->file->has_transactions(); > + portion_of_time_through_update= > + !(table->triggers && table->triggers->has_triggers(TRG_EVENT_INSERT, > + TRG_ACTION_BEFORE)) > + && !table->versioned() > + && table->file->has_transactions(); > > THD_STAGE_INFO(thd, stage_updating); > while (likely(!(error=info.read_record())) && likely(!thd->killed) && > @@ -840,6 +836,12 @@ bool mysql_delete(THD *thd, TABLE_LIST *table_list, COND *conds, > else > { > error= table->delete_row(); > + > + ha_rows rows_inserted; > + if (likely(!error) && table_list->has_period() > + && !portion_of_time_through_update) > + error= table->insert_portion_of_time(thd, table_list->period_conditions, > + &rows_inserted); > } Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

2 2

[Maria-developers] Input on Optimizer trace code, part #2
by Sergey Petrunia 14 Feb '19

14 Feb '19

Hi Varun, Please find more input below. > --- 10.4-optimizer-trace-orig/sql/sql_derived.cc 2019-02-10 14:25:59.217015365 +0300 > +++ 10.4-optimizer-trace-cl/sql/sql_derived.cc 2019-02-12 14:04:05.446571305 +0300 > @@ -486,6 +486,24 @@ exit_merge: > DBUG_RETURN(res); > > unconditional_materialization: > + > + if (unlikely(thd->trace_started())) > + { > + /* > + Add to the optimizer trace the change in choice for merged > + derived tables/views to materialised ones. > + */ > + Json_writer_object trace_wrapper(thd); > + Json_writer_object trace_derived(thd, derived->is_derived() ? > + "derived" : "view"); > + trace_derived.add("table", derived->alias.str ? derived->alias.str : "<NULL>") > + .add_select_number(derived->get_unit()-> > + first_select()->select_number) > + .add("initial_choice", "merged") > + .add("final_choice", "materialized") > + .add("cause", cause); Please add test coverage for this (I cannot find it in the testsuite). > +select * from v1 { > + "steps": [ > + { > + "join_preparation": { > + "select_id": 1, > + "steps": [ > + { > + "view": { > + "table": "v1", > + "select_id": 2, > + "merged": true The above looks a bit odd, "view", inside it "table", and then either "materilized" or "merged". I would prefer "algorithm": "materialize" or "algorithm": "merge" > diff -urpN '--exclude=.*' 10.4-optimizer-trace-orig/sql/sql_select.cc 10.4-optimizer-trace-cl/sql/sql_select.cc > --- 10.4-optimizer-trace-orig/sql/sql_select.cc 2019-02-10 14:25:59.225015425 +0300 > +++ 10.4-optimizer-trace-cl/sql/sql_select.cc 2019-02-12 14:04:05.450571322 +0300 > @@ -345,6 +349,40 @@ bool dbug_user_var_equals_int(THD *thd, > } > #endif > > +static void trace_table_dependencies(THD *thd, > + JOIN_TAB *join_tabs, uint table_count) > +{ > + Json_writer_object trace_wrapper(thd); > + Json_writer_array trace_dep(thd, "table_dependencies"); > + for (uint i = 0; i < table_count; i++) > + { > + TABLE_LIST *table_ref = join_tabs[i].tab_list; > + Json_writer_object trace_one_table(thd); > + trace_one_table.add_table_name(&join_tabs[i]); > + trace_one_table.add("row_may_be_null", > + (bool)table_ref->table->maybe_null); > + const table_map map = table_ref->get_map(); > + DBUG_ASSERT(map < (1ULL << table_count)); > + for (uint j = 0; j < table_count; j++) > + { > + if (map & (1ULL << j)) > + { > + trace_one_table.add("map_bit", static_cast<longlong>(j)); > + break; > + } > + } > + Json_writer_array depends_on(thd, "depends_on_map_bits"); > + static_assert(sizeof(table_ref->get_map()) <= 64, > + "RAND_TABLE_BIT may be in join_tabs[i].dependent, so we test " > + "all 64 bits."); sizeof() is in bytes, not bits, so the above assert actually checks for whether table_map is 64 bytes large (which is incorrect). I would suggest to switch to using Table_map_iterator and forget about the 64-byte limitation (and the value of RAND_TABLE_BIT, too) > + for (uint j = 0; j < 64; j++) > + { > + if (join_tabs[i].dependent & (1ULL << j)) > + depends_on.add(static_cast<longlong>(j)); > + } > --- 10.4-optimizer-trace-orig/sql/opt_table_elimination.cc 2019-02-10 14:25:59.209015304 +0300 > +++ 10.4-optimizer-trace-cl/sql/opt_table_elimination.cc 2019-02-12 14:04:05.442571289 +0300 > @@ -522,7 +524,8 @@ eliminate_tables_for_list(JOIN *join, > List<TABLE_LIST> *join_list, > table_map tables_in_list, > Item *on_expr, > - table_map tables_used_elsewhere); > + table_map tables_used_elsewhere, > + Json_writer_array* eliminate_tables); I think I wrote this already - I think 'eliminate_tables' is a very poor name for the argument- it doesn't show that this is just a trace output object. If something is named 'eliminate_tables' I would expect it to play some active role (e.g. being a list of tables we are trying to eliminate or something like that). Please change to something like trace, trace_eliminated, trace_for_eliminated, ... > @@ -2340,8 +2377,13 @@ int pull_out_semijoin_tables(JOIN *join) > bool optimize_semijoin_nests(JOIN *join, table_map all_table_map) > { > DBUG_ENTER("optimize_semijoin_nests"); > + THD *thd= join->thd; > List_iterator<TABLE_LIST> sj_list_it(join->select_lex->sj_nests); > TABLE_LIST *sj_nest; > + Json_writer_object wrapper(thd); > + Json_writer_object trace_semijoin_nest(thd, > + "execution_plan_for_potential_materialization"); > + Json_writer_array trace_steps_array(thd, "steps"); > while ((sj_nest= sj_list_it++)) > { > /* semi-join nests with only constant tables are not valid */ The most common output this produces is a useless "execution_plan_for_potential_materialization": { "steps": [] } Please add "if (!join->select_lex->sj_nests.elements) \n DBUG_RETURN(FALSE);" to avoid it. Please also add a testcase where a non-empty output is produced (it's nice to have a semi-representative set of outputs in main/opt_trace*result). > --- 10.4-optimizer-trace-orig/mysql-test/main/opt_trace.result 1970-01-01 03:00:00.000000000 +0300 > +++ 10.4-optimizer-trace-cl/mysql-test/main/opt_trace.result 2019-02-12 14:04:05.438571272 +0300 > +QUERY TRACE MISSING_BYTES_BEYOND_MAX_MEM_SIZE INSUFFICIENT_PRIVILEGES > +EXPLAIN SELECT DISTINCT a FROM t1 { ... > + "best_group_range_summary": { > + "type": "index_group", > + "index": "a", > + "group_attribute": null, > + "min_aggregate": false, > + "max_aggregate": false, > + "distinct_aggregate": false, > + "rows": 5, > + "cost": 7.5, > + "key_parts_used_for_access": ["a"], > + "ranges": [], > + "chosen": true > + }, How come group_attribute is NULL? Loose Index Scan requires groping on a column. Are you sure you are printing the right value? --> - 10.4-optimizer-trace-orig/sql/opt_trace.cc 1970-01-01 03:00:00.000000000 +0300 > +++ 10.4-optimizer-trace-cl/sql/opt_trace.cc 2019-02-12 14:04:05.442571289 +0300 > ... > +void opt_trace_print_expanded_query(THD *thd, SELECT_LEX *select_lex, > + Json_writer_object *writer) > + > +{ > + if (!thd->trace_started()) > + return; > + char buff[1024]; > + String str(buff, sizeof(buff), system_charset_info); > + str.length(0); Please use StringBuffer<1024> instead. > + select_lex->print(thd, &str, > + enum_query_type(QT_TO_SYSTEM_CHARSET | > + QT_SHOW_SELECT_NUMBER | > + QT_ITEM_IDENT_SKIP_DB_NAMES | > + QT_VIEW_INTERNAL > + )); > + /* > + The output is not very pretty lots of back-ticks, the output > + is as the one in explain extended , lets try to improved it here. > + */ > + writer->add("expanded_query", str.c_ptr_safe(), str.length()); > +} Did you try using QT_EXPLAIN? (if it doesn't work, I'd like to know why) ... > + trace->missing_privilege(); > + return; > +} A 'return;' at the end of void function looks very confusing. Please remove it > +void opt_trace_disable_if_no_stored_proc_func_access(THD *thd, sp_head *sp) > +{ > + if (likely(!(thd->variables.optimizer_trace & > + Opt_trace_context::FLAG_ENABLED)) || > + thd->system_thread) > + return; > + > + Opt_trace_context *const trace = &thd->opt_trace; > + if (!thd->trace_started()) > + return; > + bool full_access; > + Security_context *const backup_thd_sctx = thd->security_context(); > + thd->set_security_context(&thd->main_security_ctx); > + const bool rc = check_show_routine_access(thd, sp, &full_access) || !full_access; > + thd->set_security_context(backup_thd_sctx); > + if (rc) > + trace->missing_privilege(); > + return; > +} A 'return;' at the end of void function looks very confusing. Please remove it. Also, check the assignments in the function: they should follow coding style and be "x= y" and not "x = y". > +int fill_optimizer_trace_info(THD *thd, TABLE_LIST *tables, Item *) > +{ > + TABLE *table = tables->table; > + Opt_trace_info info; > + > + /* get_values of trace, query , missing bytes and missing_priv > + > + @todo: Need an iterator here to walk over all the traces > + */ > + Opt_trace_context* ctx= &thd->opt_trace; > + > + if (thd->opt_trace.empty()) > + { So thd->opt_trace.empty() returns FALSE when there are NO traces. And it returns TRUE when there are traces. THIS IS EXTREMELY CONFUSING! Please fix it. > + Opt_trace_stmt *stmt= ctx->get_top_trace(); > + stmt->fill_info(&info); > + > + table->field[0]->store(info.query_ptr, static_cast<uint>(info.query_length), > + info.query_charset); ... > +void Json_writer::add_str(Item *item) > +{ > + if (item) > + { > + THD *thd= current_thd; > + char buff[256]; > + String str(buff, sizeof(buff), system_charset_info); > + str.length(0); Please use StringBuffer. > + ~Opt_trace_stmt() > + { > + delete current_json; > + missing_priv= false; > + ctx= NULL; > + I_S_disabled= 0; > + } What is this for? Why set member variables in destructor? This either does nothing, or is a unsuccessful attempt to prevent "reading freed memory"-type crashes. Please remove this. > +private: > + Opt_trace_context *ctx; > + String query; // store the query sent by the user > + Json_writer *current_json; // stores the trace > + bool missing_priv; ///< whether user lacks privilege to see this trace > + uint I_S_disabled; > +}; What is I_S_disabled and why is it called like that? Please document as it is totally confusing. > +void opt_trace_disable_if_no_security_context_access(THD *thd) > ... > + if (!thd->trace_started()) > + { > + /* > + @@optimizer_trace has "enabled=on" but trace is not started. > + Either Opt_trace_start ctor was not called for our statement (3), or it > + was called but at that time, the variable had "enabled=off" (4). > + > + There are no known cases of (3). > + > + (4) suggests that the user managed to change the variable during > + execution of the statement, and this statement is using > + view/routine (note that we have not been able to provoke this, maybe > + this is impossible). If it happens it is suspicious. > + > + We disable I_S output. And we cannot do otherwise: we have no place to > + store a possible "missing privilege" information (no Opt_trace_stmt, as > + is_started() is false), so cannot do security checks, so cannot safely > + do tracing, so have to disable I_S output. And even then, we don't know > + when to re-enable I_S output, as we have no place to store the > + information "re-enable tracing at the end of this statement", and we > + don't even have a notion of statement here (statements in the optimizer > + trace world mean an Opt_trace_stmt object, and there is none here). So > + we must disable for the session's life. > + > + COM_FIELD_LIST opens views, thus used to be a case of (3). To avoid > + disabling I_S output for the session's life when this command is issued > + (like in: "SET OPTIMIZER_TRACE='ENABLED=ON';USE somedb;" in the 'mysql' > + command-line client), we have decided to create a Opt_trace_start for > + this command. The command itself is not traced though > + (SQLCOM_SHOW_FIELDS does not have CF_OPTIMIZER_TRACE). > + */ > + return; > + } As far as I understand, this was copied from MySQL? Is this still true in MariaDB codebase? It sounds scary, let's discuss it. > diff -urpN '--exclude=.*' 10.4-optimizer-trace-orig/sql/my_json_writer.cc 10.4-optimizer-trace-cl/sql/my_json_writer.cc > --- 10.4-optimizer-trace-orig/sql/my_json_writer.cc 2019-02-10 14:25:59.201015244 +0300 > +++ 10.4-optimizer-trace-cl/sql/my_json_writer.cc 2019-02-12 14:04:05.438571272 +0300 > ... > +Json_writer_array::Json_writer_array(THD *thd) : > + Json_writer_struct(thd) Json_writer_object and Json_writer_array are widely used. I guess performance-wise, it is be beneficial if their calls are inlined. A while ago, compiler would only inline a function if its body is in the .h file (and so is visible in the compilation unit where inlining should occur). I'm wondering if this is still true? (if yes, we better move these methods to the .h file). Can you look at release-build disassembly and check if there are calls to Json_writer_array member functions there or they were inlined? (the same question for THD::trace_started) > +{ > + if (my_writer) > + my_writer->start_array(); > +} > + > +Json_writer_array::Json_writer_array(THD *thd, const char *str) > + :Json_writer_struct(thd) Please change the indentation to be like in the ctor above. > +{ > + if (my_writer) ... > @@ -123,9 +209,18 @@ public: > + /* > + One can set a limit of how large a JSON document should be. > + Writes beyond that size will be counted, but will not be collected. > + */ > + void set_size_limit(size_t mem_size) { output.set_size_limit(mem_size); } > + > + // psergey: return how many bytes would be required to store everything Remove the comment. > + size_t get_truncated_bytes() { return output.get_truncated_bytes(); } > + > Json_writer() : > indent_level(0), document_start(true), element_started(false), > - first_child(true) > + first_child(true), allowed_mem_size(0) > { > fmt_helper.init(this); > } > @@ -140,15 +235,335 @@ private: > bool element_started; > bool first_child; > > + /* > + True when we are using the optimizer trace > + FALSE otherwise > + */ > + size_t allowed_mem_size; The comment above is wrong, remove it. Also remove the 'allowed_mem_size' as it is not used anymore. > +void Opt_trace_context::flush_optimizer_trace() > +{ > + inited= false; > + if (traces) > + { > + while (traces->elements()) > + { > + Opt_trace_stmt *prev= traces->at(0); > + delete prev; > + traces->del(0); > + } > + delete traces; > + traces= NULL; > + } > +} The name of this function is confusing. It is not clear what is flushed and where. How about renaming it to "delete_traces()" ? > +Opt_trace_context::~Opt_trace_context() > +{ > + inited= FALSE; > + /* > + would be nice to move this to a function > + */ And this function is already there - it's currently called flush_optimizer_trace(). I am also confused with setting member variables (inited, max_mem_size) in the destructor. This object is going away, why do it? Please remove that. > + if (traces) > + { > + while (traces->elements()) > + { > + Opt_trace_stmt *prev= traces->at(0); > + delete prev; > + traces->del(0); > + } > + delete traces; > + traces= NULL; > + } > + max_mem_size= 0; > +} > + > --- 10.4-optimizer-trace-orig/sql/opt_range.cc 2019-02-10 14:25:59.209015304 +0300 > +++ 10.4-optimizer-trace-cl/sql/opt_range.cc 2019-02-12 14:04:05.438571272 +0300 > @@ -3214,17 +3431,25 @@ bool calculate_cond_selectivity_for_tabl > SEL_ARG *key= tree->keys[idx]; > if (key) > { ... > rows= records_in_column_ranges(&param, idx, key); > if (rows != DBL_MAX) > + { > key->field->cond_selectivity= rows/table_records; > + selectivity_for_column.add("selectivity_from_histograms", > + key->field->cond_selectivity); > + } That's not from multiple histogramS, that's from one histogram. Please change the name. And I think I have requested this already: at the end of the function, please print the table->cond_selectivity as the total selectivity of condition for this table. ... Coding style: can you search through your patch for instances of if (x) { } and change them to if (x) { } ? The same for "x = y;" -> "x= y;" > --- 10.4-optimizer-trace-orig/sql/opt_trace_context.h 1970-01-01 03:00:00.000000000 +0300 > +++ 10.4-optimizer-trace-cl/sql/opt_trace_context.h 2019-02-10 14:25:38.576860203 +0300 > + Opt_trace_stmt *current_trace; > + /* > + TRUE: if we allocate memory for list of traces > + FALSE: otherwise > + */ > + bool inited; 'inited' is not necessary, it would be sufficient to check if 'traces' is NULL. (On the other hand, why traces is a pointer?) BR Sergei -- Sergei Petrunia, Software Developer MariaDB Corporation | Skype: sergefp | Blog: http://s.petrunia.net/blog

1 0

Re: [Maria-developers] [Commits] b428e82: MDEV-7974 backport fix for mysql bug#12161 (XA and binlog).
by andrei.elkin＠pp.inet.fi 14 Feb '19

14 Feb '19

Привет, Алексей! Я просматриваю патч. Где-нить через 2-3 часа должны быть комментарии от меня. Можно будет созвониться. И сразу - полагаю, ты сейчас прямо усердно трудишься над тестами (в патче не присутствующими). Andrei. holyfoot(a)askmonty.org (Alexey Botchkov) writes: > revision-id: b428e822da09b2bc82a2447332bb980f93c80262 (mariadb-10.4.1-103-gb428e82) > parent(s): 8aae31cf494678b6253031c627566e50bc666920 > committer: Alexey Botchkov > timestamp: 2019-02-14 02:46:57 +0400 > message: > > MDEV-7974 backport fix for mysql bug#12161 (XA and binlog). > > XA transactions now are kept persistent after prepare. > XA_prepare_log_event implamented, and XA tranasctions are logged > as XA transactions. > > --- > sql/handler.cc | 9 ++ > sql/handler.h | 10 ++ > sql/log.cc | 115 +++++++++++++--- > sql/log.h | 10 ++ > sql/log_event.cc | 397 +++++++++++++++++++++++++++++++++++++++++++++++++++-- > sql/log_event.h | 81 +++++++++++ > sql/sql_class.cc | 18 +-- > sql/sql_class.h | 20 ++- > sql/sql_connect.cc | 1 + > sql/transaction.cc | 69 +++++++++- > sql/transaction.h | 1 + > 11 files changed, 691 insertions(+), 40 deletions(-) > > diff --git a/sql/handler.cc b/sql/handler.cc > index 001055c..3b2a3e0 100644 > --- a/sql/handler.cc > +++ b/sql/handler.cc > @@ -1214,6 +1214,9 @@ static int prepare_or_error(handlerton *ht, THD *thd, bool all) > } > > > +/*static inline */int > +binlog_commit_flush_xid_caches(THD *thd, binlog_cache_mngr *cache_mngr, > + bool all, my_xid xid); > /** > @retval > 0 ok > @@ -1225,6 +1228,7 @@ int ha_prepare(THD *thd) > int error=0, all=1; > THD_TRANS *trans=all ? &thd->transaction.all : &thd->transaction.stmt; > Ha_trx_info *ha_info= trans->ha_list; > + > DBUG_ENTER("ha_prepare"); > > if (ha_info) > @@ -1250,6 +1254,11 @@ int ha_prepare(THD *thd) > > } > } > + if (unlikely(tc_log->log_prepare(thd, all))) > + { > + ha_rollback_trans(thd, all); > + error=1; > + } > } > > DBUG_RETURN(error); > diff --git a/sql/handler.h b/sql/handler.h > index fc6246c..613c1c3 100644 > --- a/sql/handler.h > +++ b/sql/handler.h > @@ -810,6 +810,16 @@ struct xid_t { > long gtrid_length; > long bqual_length; > char data[XIDDATASIZE]; // not \0-terminated ! > + /* > + The size of the string containing serialized Xid representation > + is computed as a sum of > + eight as the number of formatting symbols (X'',X'',) > + plus 2 x XIDDATASIZE (2 due to hex format), > + plus space for decimal digits of XID::formatID, > + plus one for 0x0. > + */ > + static const uint ser_buf_size= > + 8 + 2 * XIDDATASIZE + 4 * sizeof(long) + 1; > > xid_t() {} /* Remove gcc warning */ > bool eq(struct xid_t *xid) > diff --git a/sql/log.cc b/sql/log.cc > index a56117a..316b871 100644 > --- a/sql/log.cc > +++ b/sql/log.cc > @@ -87,6 +87,9 @@ static bool binlog_savepoint_rollback_can_release_mdl(handlerton *hton, > static int binlog_commit(handlerton *hton, THD *thd, bool all); > static int binlog_rollback(handlerton *hton, THD *thd, bool all); > static int binlog_prepare(handlerton *hton, THD *thd, bool all); > +static int binlog_xa_recover(handlerton *hton, XID *xid_list, uint len); > +static int binlog_commit_by_xid(handlerton *hton, XID *xid); > +static int binlog_rollback_by_xid(handlerton *hton, XID *xid); > static int binlog_start_consistent_snapshot(handlerton *hton, THD *thd); > > static const LEX_CSTRING write_error_msg= > @@ -1688,6 +1691,9 @@ int binlog_init(void *p) > binlog_hton->commit= binlog_commit; > binlog_hton->rollback= binlog_rollback; > binlog_hton->prepare= binlog_prepare; > + binlog_hton->recover= binlog_xa_recover; > + binlog_hton->commit_by_xid = binlog_commit_by_xid; > + binlog_hton->rollback_by_xid = binlog_rollback_by_xid; > binlog_hton->start_consistent_snapshot= binlog_start_consistent_snapshot; > binlog_hton->flags= HTON_NOT_USER_SELECTABLE | HTON_HIDDEN; > return 0; > @@ -1883,23 +1889,16 @@ static inline int > binlog_commit_flush_xid_caches(THD *thd, binlog_cache_mngr *cache_mngr, > bool all, my_xid xid) > { > - if (xid) > - { > - Xid_log_event end_evt(thd, xid, TRUE); > - return (binlog_flush_cache(thd, cache_mngr, &end_evt, all, TRUE, TRUE)); > - } > - else > + /* Mask XA COMMIT ... ONE PHASE as plain BEGIN ... COMMIT */ > + if (!xid) > { > - /* > - Empty xid occurs in XA COMMIT ... ONE PHASE. > - In this case, we do not have a MySQL xid for the transaction, and the > - external XA transaction coordinator will have to handle recovery if > - needed. So we end the transaction with a plain COMMIT query event. > - */ > - Query_log_event end_evt(thd, STRING_WITH_LEN("COMMIT"), > - TRUE, TRUE, TRUE, 0); > - return (binlog_flush_cache(thd, cache_mngr, &end_evt, all, TRUE, TRUE)); > + DBUG_ASSERT(thd->transaction.xid_state.xa_state == XA_IDLE && > + thd->lex->xa_opt == XA_ONE_PHASE); > + xid= thd->query_id; > } > + > + Xid_log_event end_evt(thd, xid, TRUE); > + return (binlog_flush_cache(thd, cache_mngr, &end_evt, all, TRUE, TRUE)); > } > > /** > @@ -1961,11 +1960,77 @@ static int binlog_prepare(handlerton *hton, THD *thd, bool all) > do nothing. > just pretend we can do 2pc, so that MySQL won't > switch to 1pc. > - real work will be done in MYSQL_BIN_LOG::log_and_order() > + real work is done in MYSQL_BIN_LOG::log_and_order() > */ > return 0; > } > > + > +static int serialize_xid(XID *xid, char *buf) > +{ > + size_t size; > + buf[0]= '\''; > + memcpy(buf+1, xid->data, xid->gtrid_length); > + size= xid->gtrid_length + 2; > + buf[size-1]= '\''; > + if (xid->bqual_length == 0 && xid->formatID == 1) > + return size; > + > + memcpy(buf+size, ", '", 3); > + memcpy(buf+size+3, xid->data+xid->gtrid_length, xid->bqual_length); > + size+= 3 + xid->bqual_length; > + buf[size]= '\''; > + size++; > + if (xid->formatID != 1) > + size+= sprintf(buf+size, ", %ld", xid->formatID); > + return size; > +} > + > + > +static int binlog_xa_recover(handlerton *hton __attribute__((unused)), > + XID *xid_list __attribute__((unused)), > + uint len __attribute__((unused))) > +{ > + /* Does nothing. */ > + return 0; > +} > + > + > +static int binlog_commit_by_xid(handlerton *hton, XID *xid) > +{ > + THD *thd= current_thd; > + const size_t xc_len= sizeof("XA COMMIT ") - 1; // do not count trailing 0 > + char buf[xc_len + xid_t::ser_buf_size]; > + size_t buflen; > + binlog_cache_mngr *const cache_mngr= thd->binlog_setup_trx_data(); > + > + DBUG_ASSERT(thd->lex->sql_command == SQLCOM_XA_COMMIT); > + > + if (!cache_mngr) > + return 1; > + > + memcpy(buf, "XA COMMIT ", xc_len); > + buflen= xc_len + serialize_xid(xid, buf+xc_len); > + Query_log_event qe(thd, buf, buflen, TRUE, FALSE, FALSE, 0); > + return binlog_flush_cache(thd, cache_mngr, &qe, TRUE, TRUE, TRUE); > +} > + > + > +static int binlog_rollback_by_xid(handlerton *hton, XID *xid) > +{ > + THD *thd= current_thd; > + const size_t xr_len= sizeof("XA ROLLBACK ") - 1; // do not count trailing 0 > + char buf[xr_len + xid_t::ser_buf_size]; > + size_t buflen; > + > + DBUG_ASSERT(thd->lex->sql_command == SQLCOM_XA_ROLLBACK); > + memcpy(buf, "XA ROLLBACK ", xr_len); > + buflen= xr_len + serialize_xid(xid, buf+xr_len); > + Query_log_event qe(thd, buf, buflen, FALSE, TRUE, TRUE, 0); > + return mysql_bin_log.write_event(&qe); > +} > + > + > /* > We flush the cache wrapped in a beging/rollback if: > . aborting a single or multi-statement transaction and; > @@ -9809,6 +9874,24 @@ int TC_LOG_BINLOG::unlog(ulong cookie, my_xid xid) > DBUG_RETURN(BINLOG_COOKIE_GET_ERROR_FLAG(cookie)); > } > > + > +int TC_LOG_BINLOG::log_prepare(THD *thd, bool all) > +{ > + XID *xid= &thd->transaction.xid_state.xid; > + XA_prepare_log_event end_evt(thd, xid, FALSE); > + binlog_cache_mngr *cache_mngr= thd->binlog_setup_trx_data(); > + > + if (!cache_mngr) > + { > + WSREP_DEBUG("Skipping empty log_xid: %s", thd->query()); > + return 0; > + } > + > + cache_mngr->using_xa= FALSE; > + return (binlog_flush_cache(thd, cache_mngr, &end_evt, all, TRUE, TRUE)); > +} > + > + > void > TC_LOG_BINLOG::commit_checkpoint_notify(void *cookie) > { > diff --git a/sql/log.h b/sql/log.h > index 7dfdb36..92fdf95 100644 > --- a/sql/log.h > +++ b/sql/log.h > @@ -61,6 +61,7 @@ class TC_LOG > bool need_prepare_ordered, > bool need_commit_ordered) = 0; > virtual int unlog(ulong cookie, my_xid xid)=0; > + virtual int log_prepare(THD *thd, bool all)= 0; > virtual void commit_checkpoint_notify(void *cookie)= 0; > > protected: > @@ -115,6 +116,10 @@ class TC_LOG_DUMMY: public TC_LOG // use it to disable the logging > return 1; > } > int unlog(ulong cookie, my_xid xid) { return 0; } > + int log_prepare(THD *thd, bool all) > + { > + return 0; > + } > void commit_checkpoint_notify(void *cookie) { DBUG_ASSERT(0); }; > }; > > @@ -198,6 +203,10 @@ class TC_LOG_MMAP: public TC_LOG > int log_and_order(THD *thd, my_xid xid, bool all, > bool need_prepare_ordered, bool need_commit_ordered); > int unlog(ulong cookie, my_xid xid); > + int log_prepare(THD *thd, bool all) > + { > + return 0; > + } > void commit_checkpoint_notify(void *cookie); > int recover(); > > @@ -698,6 +707,7 @@ class MYSQL_BIN_LOG: public TC_LOG, private MYSQL_LOG > int log_and_order(THD *thd, my_xid xid, bool all, > bool need_prepare_ordered, bool need_commit_ordered); > int unlog(ulong cookie, my_xid xid); > + int log_prepare(THD *thd, bool all); > void commit_checkpoint_notify(void *cookie); > int recover(LOG_INFO *linfo, const char *last_log_name, IO_CACHE *first_log, > Format_description_log_event *fdle, bool do_xa); > diff --git a/sql/log_event.cc b/sql/log_event.cc > index 7a0d0be..354c5f3 100644 > --- a/sql/log_event.cc > +++ b/sql/log_event.cc > @@ -2139,6 +2139,9 @@ Log_event* Log_event::read_log_event(const char* buf, uint event_len, > case XID_EVENT: > ev = new Xid_log_event(buf, fdle); > break; > + case XA_PREPARE_LOG_EVENT: > + ev = new XA_prepare_log_event(buf, fdle); > + break; > case RAND_EVENT: > ev = new Rand_log_event(buf, fdle); > break; > @@ -2190,7 +2193,6 @@ Log_event* Log_event::read_log_event(const char* buf, uint event_len, > case PREVIOUS_GTIDS_LOG_EVENT: > case TRANSACTION_CONTEXT_EVENT: > case VIEW_CHANGE_EVENT: > - case XA_PREPARE_LOG_EVENT: > ev= new Ignorable_log_event(buf, fdle, > get_type_str((Log_event_type) event_type)); > break; > @@ -6222,6 +6224,7 @@ Format_description_log_event(uint8 binlog_ver, const char* server_ver) > post_header_len[USER_VAR_EVENT-1]= USER_VAR_HEADER_LEN; > post_header_len[FORMAT_DESCRIPTION_EVENT-1]= FORMAT_DESCRIPTION_HEADER_LEN; > post_header_len[XID_EVENT-1]= XID_HEADER_LEN; > + post_header_len[XA_PREPARE_LOG_EVENT-1]= XA_PREPARE_HEADER_LEN; > post_header_len[BEGIN_LOAD_QUERY_EVENT-1]= BEGIN_LOAD_QUERY_HEADER_LEN; > post_header_len[EXECUTE_LOAD_QUERY_EVENT-1]= EXECUTE_LOAD_QUERY_HEADER_LEN; > /* > @@ -7874,7 +7877,7 @@ Gtid_log_event::Gtid_log_event(const char *buf, uint event_len, > buf+= 8; > domain_id= uint4korr(buf); > buf+= 4; > - flags2= *buf; > + flags2= *(buf++); > if (flags2 & FL_GROUP_COMMIT_ID) > { > if (event_len < (uint)header_size + GTID_HEADER_LEN + 2) > @@ -7882,8 +7885,22 @@ Gtid_log_event::Gtid_log_event(const char *buf, uint event_len, > seq_no= 0; // So is_valid() returns false > return; > } > - ++buf; > commit_id= uint8korr(buf); > + buf+= 8; > + } > + if (flags2 & FL_XA_TRANSACTION) > + { > + xid.formatID= (long) buf[0]; > + xid.gtrid_length= (long) buf[1]; > + xid.bqual_length= (long) buf[2]; > + > + buf+= 3; > + if (xid.formatID >= 0) > + { > + long data_length= xid.bqual_length + xid.gtrid_length; > + memcpy(xid.data, buf, data_length); > + buf+= data_length; > + } > } > } > > @@ -7914,6 +7931,12 @@ Gtid_log_event::Gtid_log_event(THD *thd_arg, uint64 seq_no_arg, > /* Preserve any DDL or WAITED flag in the slave's binlog. */ > if (thd_arg->rgi_slave) > flags2|= (thd_arg->rgi_slave->gtid_ev_flags2 & (FL_DDL|FL_WAITED)); > + if (thd->transaction.xid_state.xa_state == XA_IDLE && > + thd->lex->xa_opt != XA_ONE_PHASE) > + { > + flags2|= FL_XA_TRANSACTION; > + xid= thd->transaction.xid_state.xid; > + } > } > > > @@ -7956,7 +7979,7 @@ Gtid_log_event::peek(const char *event_start, size_t event_len, > bool > Gtid_log_event::write() > { > - uchar buf[GTID_HEADER_LEN+2]; > + uchar buf[GTID_HEADER_LEN+2+sizeof(XID)]; > size_t write_len; > > int8store(buf, seq_no); > @@ -7968,8 +7991,25 @@ Gtid_log_event::write() > write_len= GTID_HEADER_LEN + 2; > } > else > + write_len= 13; > + > + if (flags2 & FL_XA_TRANSACTION) > + { > + buf[write_len]= (uchar) ((char) xid.formatID); > + buf[write_len+1]= (uchar) xid.gtrid_length; > + buf[write_len+2]= (uchar) xid.bqual_length; > + write_len+= 3; > + if (xid.formatID >= 0) > + { > + long data_length= xid.bqual_length + xid.gtrid_length; > + memcpy(buf+write_len, xid.data, data_length); > + write_len+= data_length; > + } > + } > + > + if (write_len < GTID_HEADER_LEN) > { > - bzero(buf+13, GTID_HEADER_LEN-13); > + bzero(buf+write_len, GTID_HEADER_LEN-write_len); > write_len= GTID_HEADER_LEN; > } > return write_header(write_len) || > @@ -8012,7 +8052,7 @@ Gtid_log_event::make_compatible_event(String *packet, bool *need_dummy_event, > void > Gtid_log_event::pack_info(Protocol *protocol) > { > - char buf[6+5+10+1+10+1+20+1+4+20+1]; > + char buf[6+5+10+1+10+1+20+1+4+20+1+5+128]; > char *p; > p = strmov(buf, (flags2 & FL_STANDALONE ? "GTID " : "BEGIN GTID ")); > p= longlong10_to_str(domain_id, p, 10); > @@ -8026,6 +8066,12 @@ Gtid_log_event::pack_info(Protocol *protocol) > p= longlong10_to_str(commit_id, p, 10); > } > > + if (flags2 & FL_XA_TRANSACTION) > + { > + p= strmov(p, " XID :"); > + p= strnmov(p, xid.data, xid.bqual_length + xid.gtrid_length); > + } > + > protocol->store(buf, p-buf, &my_charset_bin); > } > > @@ -8079,11 +8125,25 @@ Gtid_log_event::do_apply_event(rpl_group_info *rgi) > thd->lex->sql_command= SQLCOM_BEGIN; > thd->is_slave_error= 0; > status_var_increment(thd->status_var.com_stat[thd->lex->sql_command]); > - if (trans_begin(thd, 0)) > + if (flags2 & FL_XA_TRANSACTION) > { > - DBUG_PRINT("error", ("trans_begin() failed")); > - thd->is_slave_error= 1; > + thd->lex->xid= &xid; > + thd->lex->xa_opt= XA_NONE; > + if (trans_xa_start(thd)) > + { > + DBUG_PRINT("error", ("trans_xa_start() failed")); > + thd->is_slave_error= 1; > + } > + } > + else > + { > + if (trans_begin(thd, 0)) > + { > + DBUG_PRINT("error", ("trans_begin() failed")); > + thd->is_slave_error= 1; > + } > } > + > thd->update_stats(); > > if (likely(!thd->is_slave_error)) > @@ -8202,9 +8262,29 @@ Gtid_log_event::print(FILE *file, PRINT_EVENT_INFO *print_event_info) > buf, print_event_info->delimiter)) > goto err; > } > - if (!(flags2 & FL_STANDALONE)) > - if (my_b_printf(&cache, is_flashback ? "COMMIT\n%s\n" : "BEGIN\n%s\n", print_event_info->delimiter)) > + if ((flags2 & FL_XA_TRANSACTION) && !is_flashback) > + { > + my_b_write_string(&cache, "XA START '"); > + my_b_write(&cache, (uchar *) xid.data, xid.gtrid_length); > + my_b_write_string(&cache, "'"); > + if (xid.bqual_length > 0 || xid.formatID != 1) > + { > + my_b_write_string(&cache, ", '"); > + my_b_write(&cache, (uchar *) xid.data+xid.gtrid_length, xid.bqual_length); > + my_b_write_string(&cache, "'"); > + if (xid.formatID != 1) > + if (my_b_printf(&cache, ", %d", xid.formatID)) > + goto err; > + } > + if (my_b_printf(&cache, "%s\n", print_event_info->delimiter)) > + goto err; > + } > + else if (!(flags2 & FL_STANDALONE)) > + { > + if (my_b_printf(&cache, is_flashback ? "COMMIT\n%s\n" : "BEGIN\n%s\n", > + print_event_info->delimiter)) > goto err; > + } > > return cache.flush_data(); > err: > @@ -9003,6 +9083,300 @@ Xid_log_event::do_shall_skip(rpl_group_info *rgi) > #endif /* !MYSQL_CLIENT */ > > > +/** > + Function serializes XID which is characterized by by four last arguments > + of the function. > + Serialized XID is presented in valid hex format and is returned to > + the caller in a buffer pointed by the first argument. > + The buffer size provived by the caller must be not less than > + 8 + 2 * XIDDATASIZE + 4 * sizeof(XID::formatID) + 1, see > + XID::serialize_xid() that is a caller and plugin.h for XID declaration. > + > + @param buf pointer to a buffer allocated for storing serialized data > + > + @return the value of the buffer pointer > +*/ > + > +char *XA_prepare_log_event::event_xid_t::serialize(char *buf) const > +{ > + int i; > + char *c= buf; > + /* > + Build a string like following pattern: > + X'hex11hex12...hex1m',X'hex21hex22...hex2n',11 > + and store it into buf. > + Here hex1i and hex2k are hexadecimals representing XID's internal > + raw bytes (1 <= i <= m, 1 <= k <= n), and `m' and `n' even numbers > + half of which corresponding to the lengths of XID's components. > + */ > + c[0]= 'X'; > + c[1]= '\''; > + c+= 2; > + for (i= 0; i < gtrid_length; i++) > + { > + c[0]=_dig_vec_lower[((uchar*) data)[i] >> 4]; > + c[1]=_dig_vec_lower[((uchar*) data)[i] & 0x0f]; > + c+= 2; > + } > + c[0]= '\''; > + c[1]= ','; > + c[2]= 'X'; > + c[3]= '\''; > + c+= 4; > + > + for (; i < gtrid_length + bqual_length; i++) > + { > + c[0]=_dig_vec_lower[((uchar*) data)[i] >> 4]; > + c[1]=_dig_vec_lower[((uchar*) data)[i] & 0x0f]; > + c+= 2; > + } > + c[0]= '\''; > + sprintf(c+1, ",%lu", formatID); > + > + return buf; > +} > + > + > +/************************************************************************** > + XA_prepare_log_event methods > +**************************************************************************/ > +/** > + @note > + It's ok not to use int8store here, > + as long as xid_t::set(ulonglong) and > + xid_t::get_n_xid doesn't do it either. > + We don't care about actual values of xids as long as > + identical numbers compare identically > +*/ > + > +XA_prepare_log_event:: > +XA_prepare_log_event(const char* buf, > + const Format_description_log_event *description_event) > + :Log_event(buf, description_event) > +{ > + uint32 temp= 0; > + uint8 temp_byte; > + > + buf+= description_event->common_header_len + > + description_event->post_header_len[XA_PREPARE_LOG_EVENT-1]; > + memcpy(&temp_byte, buf, 1); > + one_phase= (bool) temp_byte; > + buf += sizeof(temp_byte); > + memcpy(&temp, buf, sizeof(temp)); > + m_xid.formatID= le32toh(temp); > + buf += sizeof(temp); > + memcpy(&temp, buf, sizeof(temp)); > + m_xid.gtrid_length= le32toh(temp); > + buf += sizeof(temp); > + memcpy(&temp, buf, sizeof(temp)); > + m_xid.bqual_length= le32toh(temp); > + buf += sizeof(temp); > + memcpy(m_xid.data, buf, m_xid.gtrid_length + m_xid.bqual_length); > + > + xid= NULL; > +} > + > + > +#if defined(HAVE_REPLICATION) && !defined(MYSQL_CLIENT) > +void XA_prepare_log_event::pack_info(Protocol *protocol) > +{ > + char buf[ser_buf_size]; > + char query[sizeof("XA COMMIT ONE PHASE") + 1 + sizeof(buf)]; > + > + /* RHS of the following assert is unknown to client sources */ > + compile_time_assert(ser_buf_size == XID::ser_buf_size); > + m_xid.serialize(buf); > + sprintf(query, > + (one_phase ? "XA COMMIT %s ONE PHASE" : "XA PREPARE %s"), > + buf); > + > + protocol->store(query, strlen(query), &my_charset_bin); > +} > +#endif > + > + > +#ifndef MYSQL_CLIENT > +bool XA_prepare_log_event::write() > +{ > + uchar data[1 + 4 + 4 + 4]; > + uint8 one_phase_byte= one_phase; > + > + data[0]= one_phase; > + int4store(data+1, static_cast<XID*>(xid)->formatID); > + int4store(data+(1+4), static_cast<XID*>(xid)->gtrid_length); > + int4store(data+(1+4+4), static_cast<XID*>(xid)->bqual_length); > + > + DBUG_ASSERT(xid_bufs_size == sizeof(data) - 1); > + > + return write_header(sizeof(one_phase_byte) + xid_bufs_size + > + static_cast<XID*>(xid)->gtrid_length + > + static_cast<XID*>(xid)->bqual_length) || > + write_data(data, sizeof(data)) || > + write_data((uchar*) static_cast<XID*>(xid)->data, > + static_cast<XID*>(xid)->gtrid_length + > + static_cast<XID*>(xid)->bqual_length) || > + write_footer(); > +} > +#endif > + > + > +#ifdef MYSQL_CLIENT > +bool XA_prepare_log_event::print(FILE* file, PRINT_EVENT_INFO* print_event_info) > +{ > + Write_on_release_cache cache(&print_event_info->head_cache, file, > + Write_on_release_cache::FLUSH_F, this); > + char buf[ser_buf_size]; > + > + m_xid.serialize(buf); > + > + if (!print_event_info->short_form) > + { > + print_header(&cache, print_event_info, FALSE); > + if (my_b_printf(&cache, "\tXID = %s\n", buf)) > + goto error; > + } > + > + if (my_b_printf(&cache, "XA END %s\n%s\n", > + buf, print_event_info->delimiter) || > + my_b_printf(&cache, "XA PREPARE %s\n%s\n", > + buf, print_event_info->delimiter)) > + goto error; > + > + return cache.flush_data(); > +error: > + return TRUE; > +} > +#endif /* MYSQL_CLIENT */ > + > + > +#if defined(HAVE_REPLICATION) && !defined(MYSQL_CLIENT) > +int XA_prepare_log_event::do_apply_event(rpl_group_info *rgi) > +{ > + bool res; > + int err; > + rpl_gtid gtid; > + uint64 sub_id= 0; > + Relay_log_info const *rli= rgi->rli; > + xid_t xid; > + void *hton= NULL; > + > + /* > + XID_EVENT works like a COMMIT statement. And it also updates the > + mysql.gtid_slave_pos table with the GTID of the current transaction. > + > + Therefore, it acts much like a normal SQL statement, so we need to do > + THD::reset_for_next_command() as if starting a new statement. > + */ > + thd->reset_for_next_command(); > + /* > + Record any GTID in the same transaction, so slave state is transactionally > + consistent. > + */ > +#ifdef WITH_WSREP > + thd->wsrep_affected_rows= 0; > +#endif > + > + if (rgi->gtid_pending) > + { > + xa_states c_state= thd->transaction.xid_state.xa_state; > + sub_id= rgi->gtid_sub_id; > + rgi->gtid_pending= false; > + > + gtid= rgi->current_gtid; > + > + thd->transaction.xid_state.xa_state= XA_ACTIVE; > + err= rpl_global_gtid_slave_state->record_gtid(thd, &gtid, sub_id, true, > + false, &hton); > + thd->transaction.xid_state.xa_state= c_state; > + if (err) > + { > + int ec= thd->get_stmt_da()->sql_errno(); > + /* > + Do not report an error if this is really a kill due to a deadlock. > + In this case, the transaction will be re-tried instead. > + */ > + if (!is_parallel_retry_error(rgi, ec)) > + rli->report(ERROR_LEVEL, ER_CANNOT_UPDATE_GTID_STATE, rgi->gtid_info(), > + "Error during XID COMMIT: failed to update GTID state in " > + "%s.%s: %d: %s", > + "mysql", rpl_gtid_slave_state_table_name.str, ec, > + thd->get_stmt_da()->message()); > + thd->is_slave_error= 1; > + return err; > + } > + > + DBUG_EXECUTE_IF("gtid_fail_after_record_gtid", > + { my_error(ER_ERROR_DURING_COMMIT, MYF(0), HA_ERR_WRONG_COMMAND); > + thd->is_slave_error= 1; > + return 1; > + }); > + } > + /* For a slave XA_prepare_log_event is COMMIT */ > + general_log_print(thd, COM_QUERY, > + "COMMIT /* implicit, from Xid_log_event */"); > + thd->variables.option_bits&= ~OPTION_GTID_BEGIN; > + > + xid.set(m_xid.formatID, > + m_xid.data, m_xid.gtrid_length, > + m_xid.data + m_xid.gtrid_length, m_xid.bqual_length); > + > + thd->lex->xid= &xid; > + if (trans_xa_end(thd)) > + return 1; > + > + if (!one_phase) > + { > + res= trans_xa_prepare(thd); > + } > + else > + { > + res= trans_xa_commit(thd); > + thd->mdl_context.release_transactional_locks(); > + } > + > + > + if (!res && sub_id) > + rpl_global_gtid_slave_state->update_state_hash(sub_id, &gtid, hton, rgi); > + > + /* > + Increment the global status commit count variable > + */ > + status_var_increment(thd->status_var.com_stat[SQLCOM_COMMIT]); > + > + return res; > +} > + > + > +Log_event::enum_skip_reason > +XA_prepare_log_event::do_shall_skip(rpl_group_info *rgi) > +{ > + DBUG_ENTER("Xid_log_event::do_shall_skip"); > + if (rgi->rli->slave_skip_counter > 0) > + { > + DBUG_ASSERT(!rgi->rli->get_flag(Relay_log_info::IN_TRANSACTION)); > + thd->variables.option_bits&= ~(OPTION_BEGIN | OPTION_GTID_BEGIN); > + DBUG_RETURN(Log_event::EVENT_SKIP_COUNT); > + } > +#ifdef WITH_WSREP > + else if (wsrep_mysql_replication_bundle && WSREP_ON && > + opt_slave_domain_parallel_threads == 0) > + { > + if (++thd->wsrep_mysql_replicated < (int)wsrep_mysql_replication_bundle) > + { > + WSREP_DEBUG("skipping wsrep commit %d", thd->wsrep_mysql_replicated); > + DBUG_RETURN(Log_event::EVENT_SKIP_IGNORE); > + } > + else > + { > + thd->wsrep_mysql_replicated = 0; > + } > + } > +#endif > + DBUG_RETURN(Log_event::do_shall_skip(rgi)); > +} > +#endif /* !MYSQL_CLIENT */ > + > + > /************************************************************************** > User_var_log_event methods > **************************************************************************/ > @@ -14789,7 +15163,6 @@ bool event_that_should_be_ignored(const char *buf) > event_type == PREVIOUS_GTIDS_LOG_EVENT || > event_type == TRANSACTION_CONTEXT_EVENT || > event_type == VIEW_CHANGE_EVENT || > - event_type == XA_PREPARE_LOG_EVENT || > (uint2korr(buf + FLAGS_OFFSET) & LOG_EVENT_IGNORABLE_F)) > return 1; > return 0; > diff --git a/sql/log_event.h b/sql/log_event.h > index 38a40c9..b5c48c9 100644 > --- a/sql/log_event.h > +++ b/sql/log_event.h > @@ -217,6 +217,7 @@ class String; > #define GTID_HEADER_LEN 19 > #define GTID_LIST_HEADER_LEN 4 > #define START_ENCRYPTION_HEADER_LEN 0 > +#define XA_PREPARE_HEADER_LEN 0 > > /* > Max number of possible extra bytes in a replication event compared to a > @@ -3064,6 +3065,79 @@ class Xid_log_event: public Log_event > #endif > }; > > + > +/** > + @class XA_prepare_log_event > + > + Similar to Xid_log_event except that > + - it is specific to XA transaction > + - it carries out the prepare logics rather than the final committing > + when @c one_phase member is off. > + From the groupping perspective the event finalizes the current "prepare" group > + started with XA START Query-log-event. > + When @c one_phase is false Commit of Rollback for XA transaction are > + logged separately to the prepare-group events so being a groups of > + their own. > +*/ > + > +class XA_prepare_log_event: public Log_event > +{ > +protected: > + /* The event_xid_t members were copied from handler.h */ > + struct event_xid_t > + { > + long formatID; > + long gtrid_length; > + long bqual_length; > + char data[MYSQL_XIDDATASIZE]; // not \0-terminated ! > + char *serialize(char *buf) const; > + }; > + > + /* size of serialization buffer is explained in $MYSQL/sql/xa.h. */ > + static const uint ser_buf_size= > + 8 + 2 * MYSQL_XIDDATASIZE + 4 * sizeof(long) + 1; > + > + /* Total size of buffers to hold serialized members of XID struct */ > + static const int xid_bufs_size= 12; > + event_xid_t m_xid; > + void *xid; > + bool one_phase; > + > +public: > +#ifdef MYSQL_SERVER > + XA_prepare_log_event(THD* thd_arg, XID *xid_arg, bool one_phase_arg): > + Log_event(thd_arg, 0, TRUE), xid(xid_arg), one_phase(one_phase_arg) > + { > + cache_type= Log_event::EVENT_NO_CACHE; > + } > +#ifdef HAVE_REPLICATION > + void pack_info(Protocol* protocol); > +#endif /* HAVE_REPLICATION */ > +#else > + bool print(FILE* file, PRINT_EVENT_INFO* print_event_info); > +#endif > + XA_prepare_log_event(const char* buf, > + const Format_description_log_event *description_event); > + ~XA_prepare_log_event() {} > + Log_event_type get_type_code() { return XA_PREPARE_LOG_EVENT; } > + int get_data_size() > + { > + return xid_bufs_size + m_xid.gtrid_length + m_xid.bqual_length; > + } > + > +#ifdef MYSQL_SERVER > + bool write(); > +#endif > + bool is_valid() const { return 1; } > + > +private: > +#if defined(MYSQL_SERVER) && defined(HAVE_REPLICATION) > + virtual int do_apply_event(rpl_group_info *rgi); > + enum_skip_reason do_shall_skip(rpl_group_info *rgi); > +#endif > +}; > + > + > /** > @class User_var_log_event > > @@ -3376,6 +3450,11 @@ class Gtid_log_event: public Log_event > uint64 seq_no; > uint64 commit_id; > uint32 domain_id; > +#ifdef MYSQL_SERVER > + XID xid; > +#else > + struct st_mysql_xid xid; > +#endif > uchar flags2; > > /* Flags2. */ > @@ -3404,6 +3483,8 @@ class Gtid_log_event: public Log_event > static const uchar FL_WAITED= 16; > /* FL_DDL is set for event group containing DDL. */ > static const uchar FL_DDL= 32; > + /* FL_XA_TRANSACTION is set for XA transaction. */ > + static const uchar FL_XA_TRANSACTION= 64; > > #ifdef MYSQL_SERVER > Gtid_log_event(THD *thd_arg, uint64 seq_no, uint32 domain_id, bool standalone, > diff --git a/sql/sql_class.cc b/sql/sql_class.cc > index fa2f866..cc75da9 100644 > --- a/sql/sql_class.cc > +++ b/sql/sql_class.cc > @@ -1461,12 +1461,19 @@ void THD::cleanup(void) > DBUG_ASSERT(cleanup_done == 0); > > set_killed(KILL_CONNECTION); > -#ifdef ENABLE_WHEN_BINLOG_WILL_BE_ABLE_TO_PREPARE > if (transaction.xid_state.xa_state == XA_PREPARED) > { > -#error xid_state in the cache should be replaced by the allocated value > + trans_detach(this); > + transaction.xid_state.xa_state= XA_NOTR; > + transaction.xid_state.rm_error= 0; > + } > + else > + { > + transaction.xid_state.xa_state= XA_NOTR; > + transaction.xid_state.rm_error= 0; > + trans_rollback(this); > + xid_cache_delete(this, &transaction.xid_state); > } > -#endif > > mysql_ha_cleanup(this); > locked_tables_list.unlock_locked_tables(this); > @@ -1474,11 +1481,6 @@ void THD::cleanup(void) > delete_dynamic(&user_var_events); > close_temporary_tables(); > > - transaction.xid_state.xa_state= XA_NOTR; > - transaction.xid_state.rm_error= 0; > - trans_rollback(this); > - xid_cache_delete(this, &transaction.xid_state); > - > DBUG_ASSERT(open_tables == NULL); > /* > If the thread was in the middle of an ongoing transaction (rolled > diff --git a/sql/sql_class.h b/sql/sql_class.h > index 69fabee..76befcb 100644 > --- a/sql/sql_class.h > +++ b/sql/sql_class.h > @@ -1255,6 +1255,18 @@ typedef struct st_xid_state { > /* Error reported by the Resource Manager (RM) to the Transaction Manager. */ > uint rm_error; > XID_cache_element *xid_cache_element; > + /* > + Binary logging status. > + It is set to TRUE at XA PREPARE if the transaction was written > + to the binlog. > + Naturally FALSE means the transaction was not written to > + the binlog. Happens if the trnasaction did not modify anything > + or binlogging was turned off. In that case we shouldn't binlog > + the consequent XA COMMIT/ROLLBACK. > + The recovered transaction after server restart sets it to TRUE always. > + That can cause inconsistencies (shoud be fixed?). > + */ > + bool is_binlogged; > > /** > Check that XA transaction has an uncommitted work. Report an error > @@ -1278,6 +1290,12 @@ typedef struct st_xid_state { > } > return false; > } > + > + void reset() > + { > + xid.null(); > + is_binlogged= FALSE; > + } > } XID_STATE; > > void xid_cache_init(void); > @@ -2603,7 +2621,7 @@ class THD :public Statement, > then. > */ > if (!xid_state.rm_error) > - xid_state.xid.null(); > + xid_state.reset(); > free_root(&mem_root,MYF(MY_KEEP_PREALLOC)); > DBUG_VOID_RETURN; > } > diff --git a/sql/sql_connect.cc b/sql/sql_connect.cc > index b48070b..3e4a067 100644 > --- a/sql/sql_connect.cc > +++ b/sql/sql_connect.cc > @@ -1414,6 +1414,7 @@ void do_handle_one_connection(CONNECT *connect) > #endif > end_thread: > close_connection(thd); > + thd->get_stmt_da()->reset_diagnostics_area(); > > if (thd->userstat_running) > update_global_user_stats(thd, create_user, time(NULL)); > diff --git a/sql/transaction.cc b/sql/transaction.cc > index 13614d3..64533d7 100644 > --- a/sql/transaction.cc > +++ b/sql/transaction.cc > @@ -790,6 +790,44 @@ bool trans_release_savepoint(THD *thd, LEX_CSTRING name) > > > /** > + Detach the current XA transaction; > + > + @param thd Current thread > + > + @retval FALSE Success > + @retval TRUE Failure > +*/ > + > +bool trans_detach(THD *thd) > +{ > + XID_STATE *xid_s= &thd->transaction.xid_state; > + Ha_trx_info *ha_info, *ha_info_next; > + > + DBUG_ENTER("trans_detach"); > + > +// DBUG_ASSERT(xid_s->xa_state == XA_PREPARED && > +// xid_cache_search(thd, &xid_s->xid)); > + > + xid_cache_delete(thd, xid_s); > + if (xid_cache_insert(&xid_s->xid, XA_PREPARED)) > + DBUG_RETURN(TRUE); > + > + for (ha_info= thd->transaction.all.ha_list; > + ha_info; > + ha_info= ha_info_next) > + { > + ha_info_next= ha_info->next(); > + ha_info->reset(); /* keep it conveniently zero-filled */ > + } > + > + thd->transaction.all.ha_list= 0; > + thd->transaction.all.no_2pc= 0; > + > + DBUG_RETURN(FALSE); > +} > + > + > +/** > Starts an XA transaction with the given xid value. > > @param thd Current thread > @@ -928,6 +966,12 @@ bool trans_xa_commit(THD *thd) > res= !xs; > if (res) > my_error(ER_XAER_NOTA, MYF(0)); > + else if (thd->in_multi_stmt_transaction_mode()) > + { > + my_error(ER_XAER_RMFAIL, MYF(0), > + xa_state_names[thd->transaction.xid_state.xa_state]); > + res= TRUE; > + } > else > { > res= xa_trans_rolled_back(xs); > @@ -978,8 +1022,16 @@ bool trans_xa_commit(THD *thd) > { > DEBUG_SYNC(thd, "trans_xa_commit_after_acquire_commit_lock"); > > - res= MY_TEST(ha_commit_one_phase(thd, 1)); > - if (res) > + if(WSREP_EMULATE_BINLOG(thd) || mysql_bin_log.is_open()) > + { > + res= thd->binlog_query(THD::THD::STMT_QUERY_TYPE, > + thd->query(), thd->query_length(), > + FALSE, FALSE, FALSE, 0); > + } > + else > + res= 0; > + > + if (res || (res= MY_TEST(ha_commit_one_phase(thd, 1)))) > my_error(ER_XAER_RMERR, MYF(0)); > } > } > @@ -1044,7 +1096,18 @@ bool trans_xa_rollback(THD *thd) > DBUG_RETURN(TRUE); > } > > - res= xa_trans_force_rollback(thd); > + if(WSREP_EMULATE_BINLOG(thd) || mysql_bin_log.is_open()) > + { > + res= thd->binlog_query(THD::THD::STMT_QUERY_TYPE, > + thd->query(), thd->query_length(), > + FALSE, FALSE, FALSE, 0); > + } > + else > + res= 0; > + > + res= res || xa_trans_force_rollback(thd); > + if (res || (res= MY_TEST(xa_trans_force_rollback(thd)))) > + my_error(ER_XAER_RMERR, MYF(0)); > > thd->variables.option_bits&= ~(OPTION_BEGIN | OPTION_KEEP_LOG); > thd->transaction.all.reset(); > diff --git a/sql/transaction.h b/sql/transaction.h > index 7e34693..f228cc6 100644 > --- a/sql/transaction.h > +++ b/sql/transaction.h > @@ -29,6 +29,7 @@ bool trans_commit(THD *thd); > bool trans_commit_implicit(THD *thd); > bool trans_rollback(THD *thd); > bool trans_rollback_implicit(THD *thd); > +bool trans_detach(THD *thd); > > bool trans_commit_stmt(THD *thd); > bool trans_rollback_stmt(THD *thd); > _______________________________________________ > commits mailing list > commits(a)mariadb.org > https://lists.askmonty.org/cgi-bin/mailman/listinfo/commits

1 0

Re: [Maria-developers] 297ac6aa5b0: MDEV-7597 Expiration of user passwords
by Sergei Golubchik 13 Feb '19

13 Feb '19

Hi, Robert! On Feb 13, Robert Bindar wrote: > revision-id: 297ac6aa5b0 (mariadb-10.4.1-101-g297ac6aa5b0) > parent(s): ce02e7c9acc > author: Robert Bindar <robert(a)mariadb.org> > committer: Robert Bindar <robert(a)mariadb.org> > timestamp: 2019-02-07 17:47:13 +0200 > message: > > MDEV-7597 Expiration of user passwords > > This patch adds support for expiring user passwords. > The following statements are extended: > CREATE USER user@localhost PASSWORD EXPIRE [option] > ALTER USER user@localhost PASSWORD EXPIRE [option] > If no option is specified, the password is expired with immediate > effect. If option is DEFAULT, global policy applies according to > the default_password_lifetime system var (if 0, password never > expires, if N, password expires every N days). If option is NEVER, > the password never expires and if option is INTERVAL N DAY, the > password expires every N days. > The feature also supports the disconnect_on_expired_password system > var and the --connect-expired-password client option. > > diff --git a/libmariadb b/libmariadb > index 34f8887af03..70f2964dc4d 160000 > --- a/libmariadb > +++ b/libmariadb > @@ -1 +1 @@ > -Subproject commit 34f8887af03d022416dd6593de91d0706e57f46b > +Subproject commit 70f2964dc4de116f4b50732cfec7cb566e082b4c This shouldn't be in your commit > diff --git a/mysql-test/main/grant5.result b/mysql-test/main/grant5.result > index 3d0f331e58a..3bb81266b4d 100644 > --- a/mysql-test/main/grant5.result > +++ b/mysql-test/main/grant5.result > @@ -67,21 +67,21 @@ show create user u1@h; > ERROR 28000: Can't find any matching row in the user table > show create user u2@h; > CREATE USER for u2@h > -CREATE USER 'u2'@'h' IDENTIFIED BY PASSWORD '*975B2CD4FF9AE554FE8AD33168FBFC326D2021DD' ACCOUNT UNLOCK > +CREATE USER 'u2'@'h' IDENTIFIED BY PASSWORD '*975B2CD4FF9AE554FE8AD33168FBFC326D2021DD' PASSWORD EXPIRE DEFAULT ACCOUNT UNLOCK Just like with account locking, I wouldn't print default behavior clauses here For example, we don't print REQUIRE NONE WITH MAX_QUERIES_PER_HOUR 0 MAX_UPDATES_PER_HOUR 0 MAX_CONNECTIONS_PER_HOUR 0 MAX_USER_CONNECTIONS 0 > show create user u3@h; > CREATE USER for u3@h > -CREATE USER 'u3'@'h' IDENTIFIED BY PASSWORD '*975B2CD4FF9AE554FE8AD33168FBFC326D2021DD' ACCOUNT UNLOCK > +CREATE USER 'u3'@'h' IDENTIFIED BY PASSWORD '*975B2CD4FF9AE554FE8AD33168FBFC326D2021DD' PASSWORD EXPIRE DEFAULT ACCOUNT UNLOCK > show create user u4@h; > CREATE USER for u4@h > -CREATE USER 'u4'@'h' IDENTIFIED BY PASSWORD '*975B2CD4FF9AE554FE8AD33168FBFC326D2021DD' ACCOUNT UNLOCK > +CREATE USER 'u4'@'h' IDENTIFIED BY PASSWORD '*975B2CD4FF9AE554FE8AD33168FBFC326D2021DD' PASSWORD EXPIRE DEFAULT ACCOUNT UNLOCK > show create user u5@h; > ERROR 28000: Can't find any matching row in the user table > show create user u6@h; > CREATE USER for u6@h > -CREATE USER 'u6'@'h' IDENTIFIED BY PASSWORD '78a302dd267f6044' ACCOUNT UNLOCK > +CREATE USER 'u6'@'h' IDENTIFIED BY PASSWORD '78a302dd267f6044' PASSWORD EXPIRE DEFAULT ACCOUNT UNLOCK > show create user u7@h; > CREATE USER for u7@h > -CREATE USER 'u7'@'h' IDENTIFIED BY PASSWORD '78a302dd267f6044' ACCOUNT UNLOCK > +CREATE USER 'u7'@'h' IDENTIFIED BY PASSWORD '78a302dd267f6044' PASSWORD EXPIRE DEFAULT ACCOUNT UNLOCK > show create user u8@h; > ERROR 28000: Can't find any matching row in the user table > grant select on *.* to u1@h; > diff --git a/mysql-test/main/password_expiration_dbug.test b/mysql-test/main/password_expiration_dbug.test > new file mode 100644 > index 00000000000..99cbf5a6efb > --- /dev/null > +++ b/mysql-test/main/password_expiration_dbug.test > @@ -0,0 +1,57 @@ > +# > +# Test password expiration INTERVAL and default_password_lifetime options > +# > + > +--source include/have_debug.inc > +--source include/not_embedded.inc > + > +set @old_dbug=@@global.debug_dbug; > +set global debug_dbug="+d,password_expiration_interval_sec"; > + > +--echo # > +--echo # PASSWORD EXPIRE DEFAULT should use the default_password_lifetime > +--echo # system var to set the number of days till expiration > +--echo # > +set global disconnect_on_expired_password=ON; > +set global default_password_lifetime=1; > +create user user1@localhost password expire default; > +--sleep 2 I don't like that. all sleep-based tests have proven to be very unreliable in buildbot :( I don't have any good ideas, you cannot change the global server time from the test. May be, mess with a timezone? Like, set the expiration for a few minutes (like you do) and change server's timezone to move the time forward? Might work. Depends on what time you use for expiration - local or UTC. > +--replace_result $MASTER_MYPORT MYSQL_PORT $MASTER_MYSOCK MYSQL_SOCK > +--error ER_MUST_CHANGE_PASSWORD_LOGIN > +connect(con1,localhost,user1); > +drop user user1@localhost; > + > +--echo # > +--echo # PASSWORD EXPIRE INTERVAL should expire a client's password after > +--echo # X seconds and not before > +--echo # > +set global disconnect_on_expired_password=ON; > +create user user1@localhost password expire interval 2 day; > +--replace_result $MASTER_MYPORT MYSQL_PORT $MASTER_MYSOCK MYSQL_SOCK > +connect(con1,localhost,user1); > +disconnect con1; > +connection default; > +--sleep 3 > +--replace_result $MASTER_MYPORT MYSQL_PORT $MASTER_MYSOCK MYSQL_SOCK > +--error ER_MUST_CHANGE_PASSWORD_LOGIN > +connect(con1,localhost,user1); > +drop user user1@localhost; > + > +--echo # > +--echo # PASSWORD EXPIRE NEVER should override the other policies and never > +--echo # expire a client's password > +--echo # > +set global disconnect_on_expired_password=ON; > +create user user1@localhost password expire interval 2 day; > +alter user user1@localhost password expire never; > +--sleep 3 > +--replace_result $MASTER_MYPORT MYSQL_PORT $MASTER_MYSOCK MYSQL_SOCK > +connect(con1,localhost,user1); > +disconnect con1; > +connection default; > +drop user user1@localhost; > + > +set global debug_dbug=@old_dbug; > +set global disconnect_on_expired_password=default; > +set global default_password_lifetime=default; > + > diff --git a/mysql-test/main/system_mysql_db_fix40123.result b/mysql-test/main/system_mysql_db_fix40123.result > index 3903dbdad94..a96a33ce54f 100644 > --- a/mysql-test/main/system_mysql_db_fix40123.result > +++ b/mysql-test/main/system_mysql_db_fix40123.result > @@ -108,6 +108,8 @@ user CREATE TABLE `user` ( > `plugin` char(64) CHARACTER SET latin1 NOT NULL DEFAULT '', > `authentication_string` text COLLATE utf8_bin NOT NULL, > `password_expired` enum('N','Y') CHARACTER SET utf8 NOT NULL DEFAULT 'N', > + `password_last_changed` timestamp NOT NULL DEFAULT current_timestamp(), > + `password_lifetime` smallint(5) unsigned NOT NULL DEFAULT 0, I've fixed it, so when you rebase they should go away, make sure to rerun these tests. > `account_locked` enum('N','Y') CHARACTER SET utf8 NOT NULL DEFAULT 'N', > `is_role` enum('N','Y') CHARACTER SET utf8 NOT NULL DEFAULT 'N', > `default_role` char(80) COLLATE utf8_bin NOT NULL DEFAULT '', > diff --git a/scripts/mysql_system_tables_fix.sql b/scripts/mysql_system_tables_fix.sql > index 04a9bd9785a..e8af55993f5 100644 > --- a/scripts/mysql_system_tables_fix.sql > +++ b/scripts/mysql_system_tables_fix.sql > @@ -805,6 +807,9 @@ IF 'BASE TABLE' = (select table_type from information_schema.tables where table_ > 'max_statement_time', max_statement_time, > 'plugin', if(plugin>'',plugin,if(length(password)=16,'mysql_old_password','mysql_native_password')), > 'authentication_string', if(plugin>'',authentication_string,password), > + 'password_expired', 'Y'=password_expired, Why do you need an explicit password_expired value? I'd just use password_last_changed=0 for that. It'll work automatically (changed too back in the past so it makes it expired) without an explicit check for password_expired==1 > + 'password_last_changed', UNIX_TIMESTAMP(password_last_changed), > + 'password_lifetime', ifnull(password_lifetime, -1), > 'account_locked', 'Y'=account_locked, > 'default_role', default_role, > 'is_role', 'Y'=is_role)) as Priv > diff --git a/sql/share/errmsg-utf8.txt b/sql/share/errmsg-utf8.txt > index 4536abd2907..e0db27165c4 100644 > --- a/sql/share/errmsg-utf8.txt > +++ b/sql/share/errmsg-utf8.txt > @@ -6940,8 +6940,9 @@ ER_NOT_VALID_PASSWORD > eng "Your password does not satisfy the current policy requirements" > > ER_MUST_CHANGE_PASSWORD > - eng "You must SET PASSWORD before executing this statement" > + eng "You must change your password using either ALTER USER or SET PASSWORD before executing this statement" > bgn "Трябва първо да си смените паролата със SET PASSWORD за да можете да изпълните тази команда" > + rum "Trebuie sa iti schimbi parola folosind ALTER USER sau SET PASSWORD inainte de a executa aceasta comanda" Let's keep the old SET PASSWORD message. It seems that ALTER USER is more appropriate for DBA, for account administration, not for mere users. > > ER_FK_NO_INDEX_CHILD > eng "Failed to add the foreign key constaint. Missing index for constraint '%s' in the foreign table '%s'" > diff --git a/sql/sql_acl.cc b/sql/sql_acl.cc > index 983686d51c7..f0a8ffb1a60 100644 > --- a/sql/sql_acl.cc > +++ b/sql/sql_acl.cc > @@ -634,7 +638,7 @@ static ACL_ROLE *find_acl_role(const char *user); > static ROLE_GRANT_PAIR *find_role_grant_pair(const LEX_CSTRING *u, const LEX_CSTRING *h, const LEX_CSTRING *r); > static ACL_USER_BASE *find_acl_user_base(const char *user, const char *host); > static bool update_user_table_password(THD *, const User_table&, > - const ACL_USER &); > + ACL_USER &); use only const references or pointers. So, this one should become a pointer. > static bool acl_load(THD *thd, const Grant_tables& grant_tables); > static inline void get_grantor(THD *thd, char* grantor); > static bool add_role_user_mapping(const char *uname, const char *hname, const char *rname); > @@ -2705,6 +2806,7 @@ bool acl_getroot(Security_context *sctx, const char *user, const char *host, > sctx->master_access= 0; > sctx->db_access= 0; > *sctx->priv_user= *sctx->priv_host= *sctx->priv_role= 0; > + sctx->set_password_expired(false); I'd rather call init() here. Future-proof. > > if (host[0]) // User, not Role > { > @@ -3457,6 +3565,17 @@ bool change_password(THD *thd, LEX_USER *user) > goto end; > } > > + if (sctx->password_expired() && > + (strcmp(acl_user->user.str, sctx->priv_user) || > + my_strcasecmp(system_charset_info, > + acl_user->host.hostname, > + sctx->priv_host))) > + { > + mysql_mutex_unlock(&acl_cache->lock); > + my_error(ER_MUST_CHANGE_PASSWORD, MYF(0)); > + goto end; > + } > + technically, check_alter_user() should've done that already, I'm not sure you need to repeat the check here. > if (update_user_table_password(thd, tables.user_table(), *acl_user)) > { > mysql_mutex_unlock(&acl_cache->lock); /* purecov: deadcode */ > @@ -3930,6 +4049,12 @@ static bool update_user_table_password(THD *thd, const User_table& user_table, > user_table.set_auth(user.plugin.str, user.plugin.length, > user.auth_string.str, user.auth_string.length); > > + /* Update the persistent password expired state of user */ > + user_table.set_password_expired(false); > + thd->set_time(); Hmm, why? This is done at the start of every statement, why do you repeat it here? > + Timeval now= thd->query_start_timeval(); Just simply m_time_t now= thd->query_start(); Don't use query_start_timeval() or query_start_sec_part() unless you _really_ need microsecond precision. They have a side-effect of server remembering that you asked for microseconds, so they'll be logged in a binary log. I've now removed THD::query_start_timeval(), it's too error-prone > + user_table.set_password_last_changed(now.tv_sec); > + > if (unlikely(error= table->file->ha_update_row(table->record[1], > table->record[0])) && > error != HA_ERR_RECORD_IS_THE_SAME) > @@ -3937,6 +4062,19 @@ static bool update_user_table_password(THD *thd, const User_table& user_table, > table->file->print_error(error,MYF(0)); /* purecov: deadcode */ > DBUG_RETURN(1); > } > + > + /* Update the acl password expired state of user */ > + user.password_last_changed= now.tv_sec; > + user.password_expired= false; > + > + /* If user is the connected user, reset the password expired field on sctx > + and allow the user to exit sandbox mode */ > + if (!strcmp(user.user.str, thd->security_ctx->priv_user) && > + !my_strcasecmp(system_charset_info, > + user.host.hostname, > + thd->security_ctx->priv_host)) this should be a method in ACL_USER or in Security_context - it's used too often. > + thd->security_ctx->set_password_expired(false); > + > DBUG_RETURN(0); > } > > @@ -4153,6 +4291,68 @@ static int replace_user_table(THD *thd, const User_table &user_table, > user_table.set_account_locked(lock_value); > new_acl_user.account_locked= lock_value; > } > + > + thd->set_time(); > + Timeval now= thd->query_start_timeval(); same as above > + if (!old_row_exists) > + { > + user_table.set_password_last_changed(now.tv_sec); > + new_acl_user.password_last_changed= now.tv_sec; > + new_acl_user.use_default_pass_lifetime= true; > + } > + > + /* Unexpire the user password */ > + if (combo->is_changing_password) Is it needed? Couldn't you just check if the password was specified? > + { > + user_table.set_password_expired(false); > + user_table.set_password_last_changed(now.tv_sec); > + new_acl_user.password_last_changed= now.tv_sec; > + new_acl_user.password_expired= false; > + > + /* If combo is the connected user, reset the password expired field on sctx > + and allow the user to exit sandbox mode */ > + if (!strcmp(combo->user.str, thd->security_ctx->priv_user) && > + !my_strcasecmp(system_charset_info, > + combo->host.str, > + thd->security_ctx->priv_host)) > + thd->security_ctx->set_password_expired(false); I tend to think (again) that ALTER USER is for DBA's, and the user should use SET PASSWORD. So, let's not allow ALTER USER in the sandbox mode and, consequently, you shouldn't need the check above either. For example, ALTER USER can specify MAX_QUERIES_PER_HOUR. Surely, it's for DBA to specify the limit, it doesn't make any sense to allow every user to overwrite his own limits. And even for authentication, think of PAM. A user can change his own password, but not /etc/pam.d/login for example. Similarly here a user should be able to change his password, but not the authentication plugin. > + } > + > + switch (lex->account_options.password_expire) { > + case PASSWORD_EXPIRE_UNSPECIFIED: > + break; > + case PASSWORD_EXPIRE_NOW: > + user_table.set_password_expired(true); > + new_acl_user.password_expired= true; > + new_acl_user.use_default_pass_lifetime= false; why do you change use_default_pass_lifetime? > + break; > + case PASSWORD_EXPIRE_NEVER: > + user_table.set_password_lifetime(0); > + new_acl_user.password_lifetime= 0; > + new_acl_user.use_default_pass_lifetime= false; > + break; > + case PASSWORD_EXPIRE_DEFAULT: > + user_table.set_password_lifetime_null(); > + new_acl_user.use_default_pass_lifetime= true; > + new_acl_user.password_lifetime= -1; why two variables to say the same thing? > + break; > + case PASSWORD_EXPIRE_INTERVAL: > + longlong interval= lex->account_options.num_expiration_days; > + Timestamp exp_tstamp(new_acl_user.password_last_changed, 0); > + int warn= 0; > + exp_tstamp.add(3600 * 24 * interval, &warn); I'm not sure why you're doing it. If the user sets too many days here - fine, why would that be a problem? > + if (warn & MYSQL_TIME_WARN_OUT_OF_RANGE) > + { > + char num[MAX_BIGINT_WIDTH + 1]; > + my_snprintf(num, sizeof(num), "%lu", interval); > + my_error(ER_WRONG_VALUE, MYF(0), "DAY", num); > + goto end; > + } > + user_table.set_password_lifetime(interval); > + new_acl_user.password_lifetime= interval; > + new_acl_user.use_default_pass_lifetime= false; > + break; > + } > } > > if (old_row_exists) > @@ -10548,6 +10764,39 @@ int mysql_alter_user(THD* thd, List<LEX_USER> &users_list) > > LEX_USER *tmp_lex_user; > List_iterator<LEX_USER> users_list_iterator(users_list); > + bool password_expired= thd->security_ctx->password_expired(); > + bool own_password_changed= false; > + > + /* Precheck the users list to see if the password of the connected > + user is being changed */ > + if (password_expired) > + { // TODO: check bug with identified by '' not blanking the password > + Security_context *sctx= thd->security_ctx; > + while ((tmp_lex_user= users_list_iterator++)) > + { > + LEX_USER* lex_user= get_current_user(thd, tmp_lex_user, false); > + > + bool is_self= !strcmp(lex_user->user.str, sctx->priv_user) && > + !my_strcasecmp(system_charset_info, > + lex_user->host.str, > + sctx->priv_host); > + if (lex_user->is_changing_password && is_self) > + { > + own_password_changed= true; > + break; > + } > + } > + > + if (!own_password_changed) > + { > + mysql_mutex_unlock(&acl_cache->lock); > + mysql_rwlock_unlock(&LOCK_grant); > + my_error(ER_MUST_CHANGE_PASSWORD, MYF(0)); already commented earlier. Just do ER_MUST_CHANGE_PASSWORD always if in a sandbox mode > + DBUG_RETURN(1); > + } > + users_list_iterator.init(users_list); > + } > + > while ((tmp_lex_user= users_list_iterator++)) > { > LEX_USER* lex_user= get_current_user(thd, tmp_lex_user, false); > @@ -13294,6 +13543,44 @@ static void handle_password_errors(const char *user, const char *hostname, PASSW > #endif > } > > +bool check_password_lifetime(THD *thd, const ACL_USER *acl_user) > +{ > + /* the password should never expire */ > + if (!acl_user->password_lifetime) > + return false; > + > + longlong interval= acl_user->password_lifetime; > + if (acl_user->use_default_pass_lifetime) > + { > + mysql_mutex_lock(&LOCK_global_system_variables); > + interval= default_password_lifetime; > + mysql_mutex_unlock(&LOCK_global_system_variables); I'd skip the mutex lock here. The worst that can happen, if someone would be logging in when default_password_lifetime is just being modified - that this someone might be classified incorrectly, allowed in with a expired password or not allowed with a non-expired. Which is a small price to pay to avoid every single user (because most users never change default settings) taking yet another mutex lock on every single login. > + > + /* default global policy applies, and that is password never expires */ > + if (!interval) > + return false; > + } > + > + thd->set_time(); > + Timeval tval_now= thd->query_start_timeval(); and again > + Timestamp tstamp_now(tval_now); > + Timestamp expiration_tstamp(acl_user->password_last_changed, 0); > + int interval_sec= 3600 * 24 * interval; > + int warn= 0; > + > + /* this helps test set a testable password lifetime in seconds not days */ > + DBUG_EXECUTE_IF("password_expiration_interval_sec", { interval_sec= interval; }); > + > + expiration_tstamp.add(interval_sec, &warn); > + > + DBUG_ASSERT(!(warn & MYSQL_TIME_WARN_OUT_OF_RANGE)); > + > + if (tstamp_now.cmp(expiration_tstamp) >= 0) { > + return true; > + } why not to write simply if ((thd->query_start() - acl_user->password_last_changed)/24/3600 > interval) this keeps all the code in one (or two) lines, instead of spreading it over two Timestamp::Timestamp constructors, Timestamp::add() and Timestamp::cmp() methods. And it doesn't have a problem with large interval values, as a bonus. > + > + return false; > +} > > /** > Perform the handshake, authorize the client and update thd sctx variables. > @@ -13512,6 +13799,25 @@ bool acl_authenticate(THD *thd, uint com_change_user_pkt_len) > DBUG_RETURN(1); > } > > + bool disconnect_password_expired= true; > + bool client_can_handle_exp_pass= thd->client_capabilities & > + CLIENT_CAN_HANDLE_EXPIRED_PASSWORDS; > + bool password_lifetime_due= check_password_lifetime(thd, acl_user); > + mysql_mutex_lock(&LOCK_global_system_variables); > + disconnect_password_expired= disconnect_on_expired_password; > + mysql_mutex_unlock(&LOCK_global_system_variables); same as above > + > + if (!client_can_handle_exp_pass && disconnect_password_expired && > + (acl_user->password_expired || password_lifetime_due)) by now you've locked the mutex twice, did some calculations to see if the password expired... and only then you've checked acl_user->password_expired. > + { > + status_var_increment(denied_connections); > + my_error(ER_MUST_CHANGE_PASSWORD_LOGIN, MYF(0)); > + DBUG_RETURN(1); > + } > + > + sctx->set_password_expired(acl_user->password_expired || > + password_lifetime_due); > + > /* > Don't allow the user to connect if he has done too many queries. > As we are testing max_user_connections == 0 here, it means that we > diff --git a/sql/sql_class.h b/sql/sql_class.h > index 69fabee708c..57b043fb8e4 100644 > --- a/sql/sql_class.h > +++ b/sql/sql_class.h > @@ -1340,6 +1342,11 @@ class Security_context { > restore_security_context(THD *thd, Security_context *backup); > #endif > bool user_matches(Security_context *); > + > + bool password_expired() > + { return pass_expired; } > + void set_password_expired(bool val) > + { pass_expired= val; } Better remove them, this class doesn't have other trivial setters/getters either. > }; > > > diff --git a/sql/sql_const.h b/sql/sql_const.h > index 82f3b9c21f2..3d888c25a9f 100644 > --- a/sql/sql_const.h > +++ b/sql/sql_const.h > @@ -275,6 +275,7 @@ > #define DELAYED_QUEUE_SIZE 1000 > #define DELAYED_WAIT_TIMEOUT (5*60) /**< Wait for delayed insert */ > #define MAX_CONNECT_ERRORS 100 ///< errors before disabling host > +#define DEFAULT_PASSWORD_LIFETIME 0 ///< days before passwords expire no need to create a name for that > > #define LONG_TIMEOUT ((ulong) 3600L*24L*365L) > > diff --git a/sql/sql_parse.cc b/sql/sql_parse.cc > index 65b52b5b5da..35e9e869b49 100644 > --- a/sql/sql_parse.cc > +++ b/sql/sql_parse.cc > @@ -1687,6 +1687,17 @@ bool dispatch_command(enum enum_server_command command, THD *thd, > thd->get_stmt_da()->set_skip_flush(); > } > > + if (unlikely(thd->security_ctx->password_expired() && > + command != COM_QUERY && > + command != COM_STMT_CLOSE && > + command != COM_STMT_SEND_LONG_DATA && > + command != COM_PING && > + command != COM_QUIT)) Why do you allow COM_STMT_CLOSE and COM_STMT_SEND_LONG_DATA? > + { > + my_error(ER_MUST_CHANGE_PASSWORD, MYF(0)); > + goto dispatch_end; > + } > + > switch (command) { > case COM_INIT_DB: > { > @@ -3249,6 +3260,14 @@ mysql_execute_command(THD *thd) > #endif > DBUG_ENTER("mysql_execute_command"); > > + if (thd->security_ctx->password_expired() && > + lex->sql_command != SQLCOM_SET_OPTION && > + lex->sql_command != SQLCOM_ALTER_USER) SQLCOM_SET_OPTION would be enough > + { > + my_error(ER_MUST_CHANGE_PASSWORD, MYF(0)); > + DBUG_RETURN(1); > + } > + > DBUG_ASSERT(thd->transaction.stmt.is_empty() || thd->in_sub_stmt); > /* > Each statement or replication event which might produce deadlock > diff --git a/sql/sql_prepare.cc b/sql/sql_prepare.cc > index cb822fc2e98..e621bd6519e 100644 > --- a/sql/sql_prepare.cc > +++ b/sql/sql_prepare.cc > @@ -4188,14 +4188,11 @@ Prepared_statement::execute_loop(String *expanded_query, > if (set_parameters(expanded_query, packet, packet_end)) > return TRUE; > > -#ifdef NOT_YET_FROM_MYSQL_5_6 > - if (unlikely(thd->security_ctx->password_expired && > - !lex->is_change_password)) > + if (unlikely(thd->security_ctx->password_expired())) Really? How can one end up here, if you don't allow COM_STMT_PREPARE and COM_STMT_EXECUTE and SQLCOM_PREPARE/SQLCOM_EXECUTE ? > { > my_error(ER_MUST_CHANGE_PASSWORD, MYF(0)); > return true; > } > -#endif > > reexecute: > // Make sure that reprepare() did not create any new Items. > @@ -4361,15 +4358,12 @@ Prepared_statement::execute_bulk_loop(String *expanded_query, > } > read_types= FALSE; > > -#ifdef NOT_YET_FROM_MYSQL_5_6 > - if (unlikely(thd->security_ctx->password_expired && > - !lex->is_change_password)) > + if (unlikely(thd->security_ctx->password_expired())) Same > { > my_error(ER_MUST_CHANGE_PASSWORD, MYF(0)); > thd->set_bulk_execution(0); > return true; > } > -#endif > > // iterations changed by set_bulk_parameters > while ((iterations || start_param) && !error && !thd->is_error()) > diff --git a/sql/sql_type.cc b/sql/sql_type.cc > index 96a73a85267..e06272e4564 100644 > --- a/sql/sql_type.cc > +++ b/sql/sql_type.cc > @@ -618,20 +618,35 @@ Sec9 & Sec9::round(uint dec) > return *this; > } > > +void Timestamp::set_out_of_range(uint sec, uint usec, int *warn) > +{ > + *warn|= MYSQL_TIME_WARN_OUT_OF_RANGE; > + tv_sec= sec; > + tv_usec= usec; > +} > > void Timestamp::round_or_set_max(uint dec, int *warn) > { > DBUG_ASSERT(dec <= TIME_SECOND_PART_DIGITS); > if (add_nanoseconds_usec(msec_round_add[dec]) && > tv_sec++ >= TIMESTAMP_MAX_VALUE) > - { > - tv_sec= TIMESTAMP_MAX_VALUE; > - tv_usec= TIME_MAX_SECOND_PART; > - *warn|= MYSQL_TIME_WARN_OUT_OF_RANGE; > - } > + set_out_of_range(TIMESTAMP_MAX_VALUE, TIME_MAX_SECOND_PART, warn); > + > my_timeval_trunc(this, dec); > } > > +Timestamp& Timestamp::add(int sec, int *warn) > +{ > + longlong res= tv_sec + sec; > + if (res >= TIMESTAMP_MAX_VALUE) > + set_out_of_range(TIMESTAMP_MAX_VALUE, TIME_MAX_SECOND_PART, warn); > + else if (res < 0) > + set_out_of_range(0, 0, warn); > + else > + tv_sec= (my_time_t) res; > + > + return *this; > +} you don't need Timestamp class at all, nor any changes here. > > bool Temporal::add_nanoseconds_with_round(THD *thd, int *warn, > date_conv_mode_t mode, > diff --git a/sql/sql_yacc.yy b/sql/sql_yacc.yy > index 8f42b18c176..919aae6858f 100644 > --- a/sql/sql_yacc.yy > +++ b/sql/sql_yacc.yy > @@ -8030,6 +8032,34 @@ opt_account_option: > { > Lex->account_options.account_locked= ACCOUNTLOCK_UNLOCKED; > } > + | PASSWORD_SYM EXPIRE_SYM opt_password_expire > + { } > + ; > +opt_password_expire: > + /* Nothing */ > + { > + Lex->account_options.password_expire= PASSWORD_EXPIRE_NOW; > + } > + | NEVER_SYM > + { > + Lex->account_options.password_expire= PASSWORD_EXPIRE_NEVER; > + } > + | DEFAULT > + { > + Lex->account_options.password_expire= PASSWORD_EXPIRE_DEFAULT; > + } > + | INTERVAL_SYM real_ulong_num DAY_SYM use NUM instead of real_ulong_num it'll make a number within UINT_MAX32 range which is what you need here anyway, no reason to limit it to UINT_MAX16. > + { > + if (unlikely($2 == 0 || $2 > UINT_MAX16)) > + { > + char num[MAX_BIGINT_WIDTH + 1]; > + my_snprintf(num, sizeof(num), "%lu", $2); > + my_yyabort_error((ER_WRONG_VALUE, MYF(0), "DAY", num)); > + } > + > + Lex->account_options.password_expire= PASSWORD_EXPIRE_INTERVAL; > + Lex->account_options.num_expiration_days= $2; > + } > ; > > ev_alter_on_schedule_completion: > diff --git a/sql/sys_vars.cc b/sql/sys_vars.cc > index 37b3d65e20a..ff521c10d9b 100644 > --- a/sql/sys_vars.cc > +++ b/sql/sys_vars.cc > @@ -1507,6 +1507,25 @@ static Sys_var_ulong Sys_max_connections( > DEFAULT(MAX_CONNECTIONS_DEFAULT), BLOCK_SIZE(1), NO_MUTEX_GUARD, > NOT_IN_BINLOG, ON_CHECK(0), ON_UPDATE(fix_max_connections)); > > +static Sys_var_ulong Sys_default_password_lifetime( > + "default_password_lifetime", > + "This defines the global password expiration policy. Defaults to 0, " > + "meaning automatic password expiration is disabled. If the value is a " > + "positive integer N, the passwords must be changed every N days. This " > + "behavior can be overriden using the password expiration options in " > + "ALTER USER.", Don't say "Defaults to 0", Just "0 means password expiration is disabled". > + GLOBAL_VAR(default_password_lifetime), CMD_LINE(REQUIRED_ARG), > + VALID_RANGE(0, UINT_MAX), DEFAULT(DEFAULT_PASSWORD_LIFETIME), > + BLOCK_SIZE(1)); > + > +static Sys_var_mybool Sys_disconnect_on_expired_password( > + "disconnect_on_expired_password", > + "This variable controls how the server handles clients with expired " > + "passwords. If enabled, the server disconnects the client, otherwise " > + "the server puts the client in sandbox mode.", Not exactly. "How the server handles clients _that don't know about the sandbox mode_" > + GLOBAL_VAR(disconnect_on_expired_password), CMD_LINE(OPT_ARG), > + DEFAULT(FALSE)); > + > static Sys_var_ulong Sys_max_connect_errors( > "max_connect_errors", > "If there is more than this number of interrupted connections from " Regards, Sergei Chief Architect MariaDB and security(a)mariadb.org

1 0