Percona crashes when doing multi-connection update against it (PPC, xenial and above)

[Mario Splivalo]

percona-xtradb-cluster-server-5.6 will segfault on PPC when there is multi-user (multi-connection) update against the database.

The installation proceeds without an issue, and running single-connection sysbench benchmark completes fine. However when number of connections increase percona will segfault almost immediately.

This is excerpt from the error.log file:
Thread pointer: 0x1002c0f1080
Attempting backtrace. You can use the following information to find out
where mysqld died. If you see no messages after this, something went
terribly wrong...
stack_bottom = 3ffe87cbe6a0 thread_stack 0x30000
/usr/sbin/mysqld(my_print_stacktrace+0x50)[0x10673270]
/usr/sbin/mysqld(handle_fatal_signal+0x4e4)[0x102ed454]
[0x3fff80ab04d8]
/usr/sbin/mysqld[0x107b5428]
[0x3fff3800fc68]
/usr/sbin/mysqld[0x107b5c84]
/usr/sbin/mysqld[0x107ba72c]
/usr/sbin/mysqld[0x10849f40]
/usr/sbin/mysqld[0x1084f69c]
/usr/sbin/mysqld[0x1077fac0]
/usr/sbin/mysqld(_ZN7handler14index_read_mapEPhPKhm16ha_rkey_function+0xb4)[0x101d3344]
/usr/sbin/mysqld(_ZN7handler16read_range_firstEPK12st_key_rangeS2_bb+0x170)[0x101c5370]
/usr/sbin/mysqld(_ZN7handler21multi_range_read_nextEPPc+0xbc)[0x101bd0bc]
/usr/sbin/mysqld(_ZN11ha_innobase21multi_range_read_nextEPPc+0x24)[0x10763694]
/usr/sbin/mysqld(_ZN18QUICK_RANGE_SELECT8get_nextEv+0x7c)[0x1052114c]
/usr/sbin/mysqld[0x1054f090]
/usr/sbin/mysqld(_Z12mysql_updateP3THDP10TABLE_LISTR4ListI4ItemES6_PS4_jP8st_ordery15enum_duplicatesbPySB_+0xd0c)[0x1043dadc]
/usr/sbin/mysqld(_Z21mysql_execute_commandP3THD+0x1820)[0x103a9d70]
/usr/sbin/mysqld(_ZN18Prepared_statement7executeEP6Stringb+0x3e4)[0x103cc854]
/usr/sbin/mysqld[0x103ccb7c]
/usr/sbin/mysqld(_Z19mysqld_stmt_executeP3THDPcj+0x314)[0x103cd4c4]
/usr/sbin/mysqld(_Z16dispatch_command19enum_server_commandP3THDPcj+0xe3c)[0x103b2b3c]
/usr/sbin/mysqld(_Z10do_commandP3THD+0x458)[0x103b5308]
/usr/sbin/mysqld(_Z24do_handle_one_connectionP3THD+0x4a8)[0x10366f48]
/usr/sbin/mysqld(handle_one_connection+0x74)[0x10367004]
/usr/sbin/mysqld(pfs_spawn_thread+0x134)[0x106d9914]
/lib/powerpc64le-linux-gnu/libpthread.so.0(+0x8070)[0x3fff80648070]
/lib/powerpc64le-linux-gnu/libc.so.6(clone+0x98)[0x3fff80263230]

I am attaching the error.log as well as my.cnf to this bug report too.

This bug is similar to Bug #1657256, but not the same. As codebase is different (percona-xtradb-cluster-server-5.5 uses InnoDB version 5.5, where this one, percona-xtradb-cluster-server-5.6 uses InnoDB version 5.6).

This bug was created to separate SRU effort as fix for percona-xtradb-cluster-server-5.5 (in Trusty) is ready, and this one will require a bit more effort.