Description
We are seeing on average a 100x slowdown in SQL queries-per-second when using UBSAN's print_stacktrace=1 on UB+ASAN MariaDB builds.
For example, when using this SQL in the mariadb CLI on MariaDB server 10.5 or 10.6
INSTALL SONAME 'ha_rocksdb';
CREATE TABLE Ｔ８ (Ｃ１ char(5),INDEX(Ｃ１)) DEFAULT CHARSET = utf8 ENGINE=RocksDB COMMENT='abcde';
The second query will take about 90 seconds with print_stacktrace=1 in UBSAN_OPTIONS and about 4-5 seconds without.
On generic testruns we measure how many queries are executed within a given timeframe, for example 30 seconds, and we are seeing that with print_stacktrace=1 the number of queries executed is about 1/100th of the amount of queries without print_stacktrace=1 while using a wide variety of queries.
The overhead of print_stacktrace=1 thus seems to be about a 100x slowdown, at least when it comes to MariaDB testing.
Compiler: Clang 18.1.3 (1), LLVM 18
OS: Ubuntu 24.04 LTS
Build: ASAN+UBSAN
Settings made:
export UBSAN_OPTIONS=suppressions=${HOME}/mariadb-qa/UBSAN.filter:print_stacktrace=1:report_error_type=1
export ASAN_OPTIONS=suppressions=${HOME}/mariadb-qa/ASAN.filter:quarantine_size_mb=512:atexit=0:detect_invalid_pointer_pairs=3:dump_instruction_bytes=1:abort_on_error=1:allocator_may_return_null=1
ulimits (if relevant):
$ ulimit -a
real-time non-blocking time (microseconds, -R) unlimited
core file size (blocks, -c) unlimited
data seg size (kbytes, -d) unlimited
scheduling priority (-e) 0
file size (blocks, -f) unlimited
pending signals (-i) unlimited
max locked memory (kbytes, -l) unlimited
max memory size (kbytes, -m) unlimited
open files (-n) 1048576
pipe size (512 bytes, -p) 8
POSIX message queues (bytes, -q) unlimited
real-time priority (-r) 0
stack size (kbytes, -s) 400000
cpu time (seconds, -t) unlimited
max user processes (-u) unlimited
virtual memory (kbytes, -v) unlimited
file locks (-x) unlimited
Please let me know if there is any fix / solution or workaround, as full stacks are required.