[MDEV-14440] Server crash in in handler::ha_external_lock or Assertion `inited==RND' failed in handler::ha_rnd_end upon SELECT from partitioned table Created: 2017-11-19  Updated: 2020-08-25  Resolved: 2019-01-18

Status: Closed
Project: MariaDB Server
Component/s: Optimizer, Storage Engine - InnoDB
Affects Version/s: 10.0, 10.1, 10.2, 10.3
Fix Version/s: 10.4.2, 10.1.38, 10.0.38, 10.2.22, 10.3.13

Type: Bug Priority: Critical
Reporter: Elena Stepanova Assignee: Varun Gupta (Inactive)
Resolution: Fixed Votes: 0
Labels: affects-tests

Issue Links:
Duplicate
duplicates MDEV-16516 [Draft] Assertion `inited==RND' faile... Closed
duplicates MDEV-17839 Crash (pure virtual method called) in... Closed
Relates
relates to MDEV-17839 Crash (pure virtual method called) in... Closed
relates to MDEV-16241 Assertion `inited==RND' failed in ha... Closed

 Description   

The test case is sporadic, run with --repeat=N (with high enough N). Can also try ASAN, it doesn't add anything to the result, but changes the dynamics._

# Run with --repeat=N
 
 
 
--source include/have_innodb.inc
 
--source include/have_partition.inc
 
 
 
SET @stats.save= @@innodb_stats_persistent;
 
SET GLOBAL innodb_stats_persistent= ON;
 
 
 
CREATE TABLE t1 (a INT) ENGINE=InnoDB;
 
CREATE TABLE t2 (
 
  pk INT AUTO_INCREMENT,
 
  i INT,
 
  t1 TIME,
 
  t2 TIME,
 
  d1 DATETIME,
 
  d2 DATETIME,
 
  v1 VARCHAR(1),
 
  v2 VARCHAR(1),
 
  PRIMARY KEY (pk),
 
  KEY (i)
 
) ENGINE=InnoDB;
 
 
 
BEGIN;
 
SELECT * FROM t1;
 
 
 
--connect (con1,localhost,root,,test)
 
 
 
ALTER TABLE t2 PARTITION BY KEY() PARTITIONS 5;
 
 
 
INSERT INTO t2 (i,t1,t2,d1,d2,v1,v2) VALUES
 
  (7, '01:39:40', '01:39:40', '1900-01-01 00:00:00', '1900-01-01 00:00:00', 'w', 'w'),
 
  (2, '18:14:11', '18:14:11', '1980-03-03 15:06:11', '1980-03-03 15:06:11', 'e', 'e'),
 
  (2, '03:28:55', '03:28:55', '2025-06-03 03:14:33', '2025-06-03 03:14:33', 'a', 'a'),
 
  (4, '15:00:03', '15:00:03', '1900-01-01 00:00:00', '1900-01-01 00:00:00', 'r', 'r'),
 
  (6, '17:10:27', '17:10:27', '2014-04-18 18:00:26', '2014-04-18 18:00:26', 'k', 'k'),
 
  (6, '09:50:22', '09:50:22', '2026-10-20 00:00:00', '2026-10-20 00:00:00', 'c', 'c'),
 
  (8, '00:00:00', '00:00:00', '2010-07-24 13:01:33', '2010-07-24 13:01:33', 'k', 'k'),
 
  (4, '09:51:13', '09:51:13', '2006-04-14 03:38:00', '2006-04-14 03:38:00', 'x', 'x'),
 
  (2, '11:25:28', '11:25:28', '2012-12-12 12:12:12', '2012-12-12 12:12:12', 'j', 'j'),
 
  (8, '11:11:11', '11:11:11', '1973-07-16 03:25:43', '1973-07-16 03:25:43', 't', 't'),
 
  (1, '11:14:24', '11:14:24', '1981-12-12 08:27:15', '1981-12-12 08:27:15', 'n', 'n'),
 
  (6, '04:00:47', '04:00:47', '2022-11-25 18:24:30', '2022-11-25 18:24:30', 'z', 'z'),
 
  (3, '21:22:12', '21:22:12', '1991-07-28 18:41:31', '1991-07-28 18:41:31', 'u', 'u'),
 
  (2, '14:29:11', '14:29:11', '2034-09-25 22:51:04', '2034-09-25 22:51:04', 'a', 'a'),
 
  (3, '16:23:38', '16:23:38', '1900-01-01 00:00:00', '1900-01-01 00:00:00', 'g', 'g'),
 
  (4, '20:06:35', '20:06:35', '2011-01-11 07:21:24', '2011-01-11 07:21:24', 'f', 'f'),
 
  (6, '19:33:10', '19:33:10', '1979-09-15 17:25:04', '1979-09-15 17:25:04', 'p', 'p'),
 
  (0, '08:51:41', '08:51:41', '2008-01-07 21:23:06', '2008-01-07 21:23:06', 'm', 'm'),
 
  (8, '22:40:20', '22:40:20', '1974-11-12 05:56:02', '1974-11-12 05:56:02', 't', 't'),
 
  (5, '13:00:50', '13:00:50', '2000-08-27 08:22:13', '2000-08-27 08:22:13', 'i', 'i'),
 
  (4, '22:22:22', '12:12:12', '2015-02-27 00:00:00', '2015-02-27 00:00:00', 'h', 'h'),
 
  (2, '12:54:06', '12:54:06', '1996-09-27 12:06:03', '1996-09-27 12:06:03', 'd', 'd'),
 
  (3, '22:47:40', '22:47:40', '2029-07-13 16:52:43', '2029-07-13 16:52:43', 'b', 'b'),
 
  (7, '03:01:56', '03:01:56', '1973-10-13 00:00:00', '1973-10-13 00:00:00', 'o', 'o'),
 
  (1, '12:35:21', '12:35:21', '2025-12-02 00:57:31', '2025-12-02 00:57:31', 'j', 'j'),
 
  (1, '00:00:00', '00:00:00', '1978-03-12 09:48:05', '1978-03-12 09:48:05', 's', 's');
 
 
 
--connection default
 
--error 0,ER_TABLE_DEF_CHANGED
 
SELECT * FROM t2 WHERE pk != 0 AND i = 0;
 
 
 
DROP TABLE t1, t2;
 
SET GLOBAL innodb_stats_persistent= @stats.save;

10.0 6aff5fa27a

 
mysqld: /data/src/10.0/sql/handler.h:2672: int handler::ha_rnd_end(): Assertion `inited==RND' failed.
 
180330 23:55:41 [ERROR] mysqld got signal 6 ;
 
 
 
#7  0x00007f91585c0ee2 in __assert_fail () from /lib/x86_64-linux-gnu/libc.so.6
 
#8  0x00000000005e47a7 in handler::ha_rnd_end (this=0x7f9144883888) at /data/src/10.0/sql/handler.h:2672
 
#9  0x000000000094be7f in QUICK_ROR_INTERSECT_SELECT::~QUICK_ROR_INTERSECT_SELECT (this=0x7f9144859d80, __in_chrg=<optimized out>) at /data/src/10.0/sql/opt_range.cc:2288
 
#10 0x000000000094bedc in QUICK_ROR_INTERSECT_SELECT::~QUICK_ROR_INTERSECT_SELECT (this=0x7f9144859d80, __in_chrg=<optimized out>) at /data/src/10.0/sql/opt_range.cc:2290
 
#11 0x000000000094a8c7 in SQL_SELECT::cleanup (this=0x7f914491eb88) at /data/src/10.0/sql/opt_range.cc:1800
 
#12 0x000000000094a94e in SQL_SELECT::~SQL_SELECT (this=0x7f914491eb88, __in_chrg=<optimized out>) at /data/src/10.0/sql/opt_range.cc:1814
 
#13 0x000000000069e810 in st_join_table::cleanup (this=0x7f914491e260) at /data/src/10.0/sql/sql_select.cc:11449
 
#14 0x000000000069f842 in JOIN::cleanup (this=0x7f91449a50d0, full=true) at /data/src/10.0/sql/sql_select.cc:11878
 
#15 0x000000000069f456 in JOIN::join_free (this=0x7f91449a50d0) at /data/src/10.0/sql/sql_select.cc:11765
 
#16 0x00000000006ad209 in do_select (join=0x7f91449a50d0, fields=0x7f914d341210, table=0x0, procedure=0x0) at /data/src/10.0/sql/sql_select.cc:17668
 
#17 0x0000000000689e15 in JOIN::exec_inner (this=0x7f91449a50d0) at /data/src/10.0/sql/sql_select.cc:3108
 
#18 0x00000000006872d2 in JOIN::exec (this=0x7f91449a50d0) at /data/src/10.0/sql/sql_select.cc:2394
 
#19 0x000000000068a674 in mysql_select (thd=0x7f914d33d070, rref_pointer_array=0x7f914d3413a0, tables=0x7f91449a42a0, wild_num=1, fields=..., conds=0x7f91449a4ed8, og_num=0, order=0x0, group=0x0, having=0x0, proc_param=0x0, select_options=2148797184, result=0x7f91449a50b0, unit=0x7f914d340a08, select_lex=0x7f914d3410f8) at /data/src/10.0/sql/sql_select.cc:3333
 
#20 0x00000000006808d2 in handle_select (thd=0x7f914d33d070, lex=0x7f914d340940, result=0x7f91449a50b0, setup_tables_done_option=0) at /data/src/10.0/sql/sql_select.cc:377
 
#21 0x000000000065505f in execute_sqlcom_select (thd=0x7f914d33d070, all_tables=0x7f91449a42a0) at /data/src/10.0/sql/sql_parse.cc:5293
 
#22 0x000000000064d5f8 in mysql_execute_command (thd=0x7f914d33d070) at /data/src/10.0/sql/sql_parse.cc:2553
 
#23 0x0000000000657ce0 in mysql_parse (thd=0x7f914d33d070, rawbuf=0x7f91449a4088 "SELECT * FROM t2 WHERE pk != 0 AND i = 0", length=40, parser_state=0x7f915a686640) at /data/src/10.0/sql/sql_parse.cc:6569
 
#24 0x000000000064a81f in dispatch_command (command=COM_QUERY, thd=0x7f914d33d070, packet=0x7f914d382071 "", packet_length=40) at /data/src/10.0/sql/sql_parse.cc:1296
 
#25 0x0000000000649b1f in do_command (thd=0x7f914d33d070) at /data/src/10.0/sql/sql_parse.cc:999
 
#26 0x0000000000769cf2 in do_handle_one_connection (thd_arg=0x7f914d33d070) at /data/src/10.0/sql/sql_connect.cc:1377
 
#27 0x0000000000769a64 in handle_one_connection (arg=0x7f914d33d070) at /data/src/10.0/sql/sql_connect.cc:1292
 
#28 0x0000000000aca5dc in pfs_spawn_thread (arg=0x7f914d282270) at /data/src/10.0/storage/perfschema/pfs.cc:1861
 
#29 0x00007f915a2c4494 in start_thread (arg=0x7f915a687700) at pthread_create.c:333
 
#30 0x00007f915867d93f in clone () from /lib/x86_64-linux-gnu/libc.so.6

If the same test is repeated long enough on a non-debug version, it eventually causes a crash in bitmap_get_first_set:

10.2 RelWithDebInfo 861038f2e8

 
#2  <signal handler called>
 
#3  bitmap_get_first_set (map=0x7fd3040e8a18) at /data/src/10.2/mysys/my_bitmap.c:634
 
#4  0x000055bfc7c3a26f in ha_partition::external_lock (this=0x7fd3040e8430, thd=0x7fd3080009a8, lock_type=2) at /data/src/10.2/sql/ha_partition.cc:3780
 
#5  0x000055bfc778245c in handler::ha_external_lock (this=0x7fd3040e8430, thd=thd@entry=0x7fd3080009a8, lock_type=lock_type@entry=2) at /data/src/10.2/sql/handler.cc:5887
 
#6  0x000055bfc784bdd8 in unlock_external (thd=0x7fd3080009a8, table=0x7fd308010008, count=<optimized out>) at /data/src/10.2/sql/lock.cc:703
 
#7  0x000055bfc784c590 in mysql_unlock_read_tables (thd=<optimized out>, sql_lock=0x7fd30800fff0) at /data/src/10.2/sql/lock.cc:480
 
#8  0x000055bfc7621187 in JOIN::join_free (this=this@entry=0x7fd308010030) at /data/src/10.2/sql/sql_select.cc:12213
 
#9  0x000055bfc763a33c in do_select (procedure=<optimized out>, join=0x7fd308010030) at /data/src/10.2/sql/sql_select.cc:18352
 
#10 JOIN::exec_inner (this=this@entry=0x7fd308010030) at /data/src/10.2/sql/sql_select.cc:3621
 
#11 0x000055bfc763a619 in JOIN::exec (this=this@entry=0x7fd308010030) at /data/src/10.2/sql/sql_select.cc:3416
 
#12 0x000055bfc763a75b in mysql_select (thd=thd@entry=0x7fd3080009a8, tables=0x7fd30800f208, wild_num=1, fields=..., conds=0x7fd30800fe28, og_num=0, order=0x0, group=0x0, having=0x0, proc_param=0x0, select_options=2148797184, result=0x7fd308010010, unit=0x7fd3080043c8, select_lex=0x7fd308004b00) at /data/src/10.2/sql/sql_select.cc:3816
 
#13 0x000055bfc763b166 in handle_select (thd=thd@entry=0x7fd3080009a8, lex=lex@entry=0x7fd308004300, result=result@entry=0x7fd308010010, setup_tables_done_option=setup_tables_done_option@entry=0) at /data/src/10.2/sql/sql_select.cc:376
 
#14 0x000055bfc75293ef in execute_sqlcom_select (thd=0x7fd3080009a8, all_tables=0x7fd30800f208) at /data/src/10.2/sql/sql_parse.cc:6479
 
#15 0x000055bfc75e566e in mysql_execute_command (thd=0x7fd3080009a8) at /data/src/10.2/sql/sql_parse.cc:3485
 
#16 0x000055bfc75eb3da in mysql_parse (thd=0x7fd3080009a8, rawbuf=<optimized out>, length=<optimized out>, parser_state=<optimized out>, is_com_multi=<optimized out>, is_next_command=<optimized out>) at /data/src/10.2/sql/sql_parse.cc:8013
 
#17 0x000055bfc75eef54 in dispatch_command (command=command@entry=COM_QUERY, thd=thd@entry=0x7fd3080009a8, packet=packet@entry=0x7fd308006ce9 "", packet_length=packet_length@entry=40, is_com_multi=is_com_multi@entry=false, is_next_command=is_next_command@entry=false) at /data/src/10.2/sql/sql_parse.cc:1824
 
#18 0x000055bfc75ef989 in do_command (thd=0x7fd3080009a8) at /data/src/10.2/sql/sql_parse.cc:1378
 
#19 0x000055bfc76b8834 in do_handle_one_connection (connect=connect@entry=0x55bfca85ad78) at /data/src/10.2/sql/sql_connect.cc:1335
 
#20 0x000055bfc76b89d4 in handle_one_connection (arg=arg@entry=0x55bfca85ad78) at /data/src/10.2/sql/sql_connect.cc:1241
 
#21 0x000055bfc797c914 in pfs_spawn_thread (arg=0x55bfca81d3d8) at /data/src/10.2/storage/perfschema/pfs.cc:1862
 
#22 0x00007fd35c491494 in start_thread (arg=0x7fd34c5f3700) at pthread_create.c:333
 
#23 0x00007fd35aa9293f in clone () from /lib/x86_64-linux-gnu/libc.so.6



 Comments   
Comment by Varun Gupta (Inactive) [ 2018-04-04 ]

In the function QUICK_RANGE_SELECT::init_ror_merged_scan we create a seperate handler if the handler in head->file cannot be reused.

So when we create the handler we set the new handler here:

   (file= head->file->clone(head->s->normalized_path.str, alloc)

and then we set 

   free_file= TRUE

It is defined as

bool free_file;   /* TRUE <=> this->file is "owned" by this quick select */

so when we create a new handler for this quick select and there is some failure and we abort and reset the handler with the previous one , we don't change this free_file flag.

Comment by Varun Gupta (Inactive) [ 2018-04-04 ]

Patch:
http://lists.askmonty.org/pipermail/commits/2018-April/012276.html

Comment by Elena Stepanova [ 2018-11-28 ]

Raised to critical because not only does it badly affect tests, but also turns out to cause a non-debug crash.

Comment by Varun Gupta (Inactive) [ 2018-11-29 ]

With the help of the optimizer trace, I see

"analyzing_roworder_intersect": {
 
                        "intersecting_indexes": [
 
                          {
 
                            "index": "i",
 
                            "index_scan_cost": 2.0647,
 
                            "cumulateed_index_scan_cost": 2.0647,
 
                            "disk_sweep_cost": 1,
 
                            "cumulative_total_cost": 3.0647,
 
                            "usable": true,
 
                            "matching_rows_now": 795,
 
                            "intersect_covering_with_this_index": false,
 
                            "chosen": true
 
                          }
 
                        ],
 
                        "clustered_pk": {
 
                          "index_scan_cost": 1.59,
 
                          "cumulateed_index_scan_cost": 3.6547,
 
                          "disk_sweep_cost": 1,
 
                          "clustered_pk_added_to_intersect": false,
 
                          "cause": "cost"
 
                        },
 
                        "chosen": false,
 
                        "cause": "too_few_indexes_to_merge"
 
                      },

We don't pick the cpk scan, hence we should not be doing the index_merge_intersect as there is only one index scan to do. We don't use the cpk to do an index scan.

Comment by Varun Gupta (Inactive) [ 2018-11-29 ]

Also I think the above patch is trying to solve the assert during the execution , the better solution would be to not pick ROR intersect in such a case.

I think psergey patch in 10.3 with commit f2c418079def should be backported to earlier versions

Comment by Varun Gupta (Inactive) [ 2019-01-14 ]

Here is the test case without partitioned tables

--source include/have_innodb.inc
 
set @optimizer_switch_save= @@optimizer_switch;
 
set optimizer_switch='index_merge_sort_intersection=off';
 
create table t0 (a int)engine=innodb;
 
insert into t0 values (0),(1),(2),(3),(4),(5),(6),(7),(8),(9);
 
create table t1 (
 
  a int, b int, c int,
 
  key(a),key(b),key(c)
 
)engine=innodb;
 
insert into t1
 
select A.a+10*B.a, A.a+10*B.a, A.a+10*B.a+100*C.a
 
from t0 A, t0 B, t0 C, t0 D where D.a<5;
 
set @@global.debug_dbug="+d,ha_index_init_fail";
 
select * from t1 where a=10 and b=10;
 
DROP TABLE t0,t1;

Comment by Sergei Petrunia [ 2019-01-17 ]
  • The testcase in the patch doesn't fire for me. varun, is it possible to use the testcase from the previous comment?
  • As far as I understand, the problem in this bug is orthogonal to whatever the commit f2c418079def is solving.
Comment by Sergei Petrunia [ 2019-01-17 ]

varun, it is ok to push the fix for this bug with the above "testcase without partitioned tables".

Comment by Sergei Petrunia [ 2019-01-17 ]

I think it is also ok to backport f2c418079def to the earlier versions.

Generated at Thu Feb 08 08:13:34 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.