| Bug #9626 | Several serious errors reported by Valgrind in latest 5.0 bk tree | ||
|---|---|---|---|
| Submitted: | 4 Apr 2005 20:59 | Modified: | 7 Jun 2005 8:50 |
| Reporter: | Guilhem Bichot | Email Updates: | |
| Status: | Closed | Impact on me: | |
| Category: | MySQL Server | Severity: | S3 (Non-critical) |
| Version: | 5.0-bk | OS: | Linux (linux) |
| Assigned to: | Magnus Blåudd | CPU Architecture: | Any |
[6 Apr 2005 9:47]
Martin Skold
The following warnings is due to the thread for query cache maintenance in ndbcluster not being exited properly at server shutdown (this is not a show stopper leak, but will be looked at ASAP). ==18453== 39012 (656 direct, 38356 indirect) bytes in 1 blocks are definitely lo <monty> st in loss record 9 of 12 <monty> ==18453== at 0x1B9042A4: malloc (vg_replace_malloc.c:130) <monty> ==18453== by 0x84AC6CE: NdbMem_Allocate (NdbMem.c:39) <monty> ==18453== by 0x84B91D3: operator new(unsigned) (new.cpp:13) <monty> ==18453== by 0x8464C23: Ndb::setup(Ndb_cluster_connection*, char const*, char <monty> const*) (Ndbinit.cpp:53) <monty> ==18453== by 0x8464B6F: Ndb::Ndb(Ndb_cluster_connection*, char const*, char c <monty> onst*) (Ndbinit.cpp:43) <monty> ==18453== by 0x82B167D: ndb_util_thread_func (ha_ndbcluster.cc:5796) <monty> ==18453== by 0x4274898B: start_thread (in /lib/tls/libpthread-0.61.so) <monty> and in some other test (not sure which): <monty> ==18589== 612 bytes in 9 blocks are possibly lost in loss record 6 of 8 <monty> ==18589== at 0x1B904BD5: calloc (vg_replace_malloc.c:175) <monty> ==18589== by 0x1B8F269C: _dl_allocate_tls (in /lib/ld-2.3.3.so) <monty> ==18589== by 0x42749746: allocate_stack (in /lib/tls/libpthread-0.61.so) <monty> ==18589== by 0x42748BB7: pthread_create@@GLIBC_2.1 (in /lib/tls/libpthread-0. <monty> 61.so) <monty> ==18589== by 0x84AC4E7: NdbThread_Create (NdbThread.c:104) <monty> ==18589== by 0x84BBB20: TransporterFacade::init(unsigned, ndb_mgm_configurati <monty> on const*) (TransporterFacade.cpp:544) <monty> ==18589== by 0x84BB31D: TransporterFacade::start_instance(int, ndb_mgm_config <monty> uration const*) (TransporterFacade.cpp:345) <monty> ==18589== by 0x848B9F4: Ndb_cluster_connection::connect(int, int, int) (ndb_c <monty> luster_connection.cpp:496) <monty> ==18589== by 0x82AE911: ndbcluster_init() (ha_ndbcluster.cc:4619) <monty> ==18589== by 0x8285CC5: ha_init() (handler.cc:397) <monty> ==18589== by 0x81CA3F2: init_server_components() (mysqld.cc:2790) <monty> ==18589== by 0x81CA944: main (mysqld.cc:3095)
[6 Apr 2005 10:28]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/internals/23705
[7 Apr 2005 11:09]
Michael Widenius
Have now fixed all warnings, except - Errors reported by pthread_create (added local suppress file to get rid of these) - Errors reported by libz (added local suppress file to get rid of these) - Errors in ndb tests (the ndb team is working on these) - Not freed memory in InnoDB (Created a new bug report for this, #9673) - The warning from rr_cmp() as this was ok. (Will soon push a fix where this doesn't show up when compiling for valgrind) The above is pushed into 5.0.4 tree
[13 May 2005 11:26]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/internals/24857
[30 May 2005 14:54]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/internals/25066
[2 Jun 2005 12:34]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/internals/25511
[2 Jun 2005 12:35]
Bugs System
A patch for this bug has been committed. After review, it may be pushed to the relevant source trees for release in the next version. You can access the patch from: http://lists.mysql.com/internals/25512
[7 Jun 2005 8:50]
Magnus Blåudd
Pushed fixes to mysql-5.0-ndb. Wrote three new more specific bug reports for remaining valgrind warnings.
Description: You need Valgrind *2.4.0* to test this; 2.2.0 segfaults and hangs at various tests. Then pull 5.0 and: BUILD/compile-pentium-valgrind-max ./mysql-test-run --valgrind --force Most warnings are about unfreed memory allocated by pthread_create(), but there are really serious ones, drowned in the middle of the Valgrind log: ==18453== at 0x8502AF7: heap_create (hp_create.c:171) <cut> ==10212== at 0x818A0B7: Item_char_typecast::val_str(String*) (sql_string.h:91) <cut> ==10212== at 0x853A5B5: my_vsnprintf (my_vsnprintf.c:91) <any other ones to look at> In the "Files" section of this bug report I attach the var/log/master.err (slave.err has only pthread_create() warnings). To help find the "non-pthread_create" warnings out of it, do this grep: for i in "^Warning:" "^Error:" "^==.* at 0x" do grep -n "$i" var/log/master.err | grep -v calloc > filt_warnings done filt_warnings will contain the serious warnings with a line number for you to get back to the total error message in var/log/master.err. How to repeat: See description.