MySQL Bugs: #70490: Suppression is too strict on some systems

Bug #70490	Suppression is too strict on some systems
Submitted:	2 Oct 2013 12:24	Modified:	29 Jun 2017 12:32
Reporter:	Sergei Glushchenko	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server: Security: Encryption	Severity:	S3 (Non-critical)
Version:	5.6.12, 5.7	OS:	Linux
Assigned to:		CPU Architecture:	Any
Tags:	openssl, valgrind

Description:
valgrind.supp has an suppression which looks like

{
   OpenSSL still reachable.
   Memcheck:Leak
   fun:malloc
   fun:CRYPTO_malloc
   fun:sk_new
   fun:load_builtin_compressions
   fun:SSL_COMP_get_compression_methods
   fun:SSL_library_init
}

It however doesn't work everywhere. We saw valgrind errors as following:

==32338==
==32338== HEAP SUMMARY:
==32338== in use at exit: 75,876,624 bytes in 832 blocks
==32338== total heap usage: 32,498 allocs, 31,666 frees, 130,387,498 bytes allocated
==32338==
==32338== 32 bytes in 1 blocks are still reachable in loss record 1 of 296
==32338== at 0x4A069EE: malloc (vg_replace_malloc.c:270)
==32338== by 0x3417C5D9CD: CRYPTO_malloc (in /usr/lib64/libcrypto.so.1.0.0)
==32338== by 0x3417CB6F5E: sk_new (in /usr/lib64/libcrypto.so.1.0.0)
==32338== by 0x341C83D0E4: ??? (in /usr/lib64/libssl.so.1.0.0)
==32338== by 0x341C83D308: SSL_COMP_get_compression_methods (in /usr/lib64/libssl.so.1.0.0)
==32338== by 0x341C844651: SSL_library_init (in /usr/lib64/libssl.so.1.0.0)
==32338== by 0xC798DB: ssl_start (viosslfactories.c:155)
==32338== by 0x625FD9: init_ssl() (mysqld.cc:4310)
==32338== by 0x630170: mysqld_main(int, char**) (mysqld.cc:5509)
==32338== by 0x340C41ECDC: (below main) (in /lib64/libc-2.12.so)
==32338==
==32338== 32 bytes in 1 blocks are still reachable in loss record 2 of 296
==32338== at 0x4A069EE: malloc (vg_replace_malloc.c:270)
==32338== by 0x3417C5D9CD: CRYPTO_malloc (in /usr/lib64/libcrypto.so.1.0.0)
==32338== by 0x3417CB6F7C: sk_new (in /usr/lib64/libcrypto.so.1.0.0)
==32338== by 0x341C83D0E4: ??? (in /usr/lib64/libssl.so.1.0.0)
==32338== by 0x341C83D308: SSL_COMP_get_compression_methods (in /usr/lib64/libssl.so.1.0.0)
==32338== by 0x341C844651: SSL_library_init (in /usr/lib64/libssl.so.1.0.0)
==32338== by 0xC798DB: ssl_start (viosslfactories.c:155)
==32338== by 0x625FD9: init_ssl() (mysqld.cc:4310)
==32338== by 0x630170: mysqld_main(int, char**) (mysqld.cc:5509)
==32338== by 0x340C41ECDC: (below main) (in /lib64/libc-2.12.so)

Originally reported as https://bugs.launchpad.net/percona-server/+bug/1205196.

How to repeat:
Probably depends on environment.

Suggested fix:
Make suppression less strict.

Thank you for the report.

Please provide test case, demonstrating the problem.

Sveta, can we proceed without test case please?

Roel,

we can. But, please, at least specify which tests and on which platform did you run when you saw such errors?

No feedback was provided for this bug for over a month, so it is
being suspended automatically. If you are able to provide the
information that was originally requested, please do so and change
the status of the bug back to "Open".

This is still an issue. Server 5.7.9, built with -DWITH_SSL=system -DWITH_VALGRIND=ON, example testcase main.mysql_client_test, will produce on Ubuntu Trusty LTS and Ubuntu Wily the following:

32 bytes in 1 blocks are still reachable in loss record 2 of 4
   at 0x4C2BBCF: malloc (vg_replace_malloc.c:299)
   by 0x553DD87: CRYPTO_malloc (in /lib/x86_64-linux-gnu/libcrypto.so.1.0.0)
   by 0x55F555E: sk_new (in /lib/x86_64-linux-gnu/libcrypto.so.1.0.0)
   by 0x52B5FF9: ??? (in /lib/x86_64-linux-gnu/libssl.so.1.0.0)
   by 0x52B80F8: SSL_COMP_get_compression_methods (in /lib/x86_64-linux-gnu/libssl.so.1.0.0)
   by 0x52BD7E2: SSL_library_init (in /lib/x86_64-linux-gnu/libssl.so.1.0.0)
   by 0x1DF2F67: ssl_start (viosslfactories.c:367)
   by 0xE8DBB7: init_ssl() (mysqld.cc:3323)
   by 0xE90B9D: mysqld_main(int, char**) (mysqld.cc:4725)
   by 0xE87EE5: main (main.cc:25)

The suppression for this is 

{
   OpenSSL still reachable.
   Memcheck:Leak
   fun:malloc
   fun:CRYPTO_malloc
   fun:sk_new
   fun:load_builtin_compressions
   fun:SSL_COMP_get_compression_methods
   fun:SSL_library_init
}

Note that sk_new is ??? above. This hinders any Valgrind testing on these platforms at least with -DWITH_SSL=system.
You have also observed this issue while verifying e.g. bug 78999.

Err "load_builtin_compressions" not "sk_new"

Hi Laurynas,

I am going to try to repeat this on Oracle Linux, with latest 5.7, since I do not use Ubuntu Linux.

Would that be OK ???

I verified it with 5.7.18:

---------------

orker[1] Valgrind report from /home/smilivoj/ssl/mysql-5.7.18/mysql-test/var/log/mysqld.1.err after tests:
 main.mysql_client_test
--------------------------------------------------------------------------
HEAP SUMMARY:
    in use at exit: 64 bytes in 2 blocks
  total heap usage: 60,271 allocs, 60,269 frees, 153,093,454 bytes allocated

32 bytes in 1 blocks are still reachable in loss record 1 of 2
   at 0x4C29BFD: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
   by 0x596C2D2: CRYPTO_malloc (in /usr/lib64/libcrypto.so.1.0.1e)
   by 0x59E4B2E: sk_new (in /usr/lib64/libcrypto.so.1.0.1e)
   by 0x56DE879: ??? (in /usr/lib64/libssl.so.1.0.1e)
   by 0x56E0828: SSL_COMP_get_compression_methods (in /usr/lib64/libssl.so.1.0.1e)
   by 0x56E61C8: SSL_library_init (in /usr/lib64/libssl.so.1.0.1e)
   by 0x1D9461A: ssl_start (viosslfactories.c:422)
   by 0xE645EB: init_ssl() (mysqld.cc:3381)
   by 0xE674BE: mysqld_main(int, char**) (mysqld.cc:4834)
   by 0xE5EA2F: main (main.cc:25)

32 bytes in 1 blocks are still reachable in loss record 2 of 2
   at 0x4C29BFD: malloc (in /usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
   by 0x596C2D2: CRYPTO_malloc (in /usr/lib64/libcrypto.so.1.0.1e)
   by 0x59E4B4C: sk_new (in /usr/lib64/libcrypto.so.1.0.1e)
   by 0x56DE879: ??? (in /usr/lib64/libssl.so.1.0.1e)
   by 0x56E0828: SSL_COMP_get_compression_methods (in /usr/lib64/libssl.so.1.0.1e)
   by 0x56E61C8: SSL_library_init (in /usr/lib64/libssl.so.1.0.1e)
   by 0x1D9461A: ssl_start (viosslfactories.c:422)
   by 0xE645EB: init_ssl() (mysqld.cc:3381)
   by 0xE674BE: mysqld_main(int, char**) (mysqld.cc:4834)
   by 0xE5EA2F: main (main.cc:25)

LEAK SUMMARY:
   definitely lost: 0 bytes in 0 blocks
   indirectly lost: 0 bytes in 0 blocks
     possibly lost: 0 bytes in 0 blocks
   still reachable: 64 bytes in 2 blocks
        suppressed: 0 bytes in 0 blocks

For counts of detected and suppressed errors, rerun with: -v
ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 2 from 2)

valgrind_report                          [ fail ]
        Test ended at 2017-05-02 20:30:50

Valgrind reported failures at shutdown, see above

--------------------------------------------------------------------------
The servers were restarted 0 times
Spent 0.000 of 31 seconds executing testcases

Completed: Failed 1/1 tests, 0.00% were successful.

Thanks Sinisa!

Posted by developer:
 
Approved on RB

Posted by developer:
 
Fixed in 5.6.38, 5.7.20, 8.0.3.

Work was done for test suite. No changelog entry needed.