Bug 2229999

Description of problem:
The following issue was found while trying to test BZ 2116948 LMDB import is too slow
but configuring a database size too small to hosts the users

After adding a new test case:
diff --git a/dirsrvtests/tests/suites/import/import_test.py b/dirsrvtests/tests/suites/import/import_test.py
index 84c8cf290..ee71e0bea 100644
--- a/dirsrvtests/tests/suites/import/import_test.py
+++ b/dirsrvtests/tests/suites/import/import_test.py
@@ -22,6 +22,7 @@ from lib389.tasks import ImportTask
 from lib389.index import Indexes
 from lib389.monitor import Monitor
 from lib389.backend import Backends
+from lib389.config import LMDB_LDBMConfig
 from lib389.config import LDBMConfig
 from lib389.utils import ds_is_newer, get_default_db_lib
 from lib389.idm.user import UserAccount
@@ -550,6 +551,15 @@ def test_import_wrong_file_path(topo):
         dbtasks_ldif2db(topo.standalone, log, args)
     assert "The LDIF file does not exist" in str(e.value)

+def test_crash_on_ldif2db_with_lmdb(topo, _import_clean):
+    BIG_MAP_SIZE = 20 * 1024 * 1024 * 1024
+    if get_default_db_lib() == "mdb":
+        handler = LMDB_LDBMConfig(topo.standalone)
+        mapsize = BIG_MAP_SIZE
+        log.info(f'Set lmdb map size to {mapsize}.')
+        handler.replace('nsslapd-mdb-max-size', str(mapsize))
+        topo.standalone.restart()
+    _import_offline(topo, 10_000_000)

 if __name__ == '__main__':
     # Run isolated


While running it with mdb, it crashes
NSSLAPD_DB_LIB=mdb py.test -v import_test.py::test_crash_on_ldif2db_with_lmdb

core was not generated, but I attached gdb during the test and then generated the core.

Thread 7 "ns-slapd" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7f9269bfa640 (LWP 3924)]
__strncmp_avx2_rtm () at ../sysdeps/x86_64/multiarch/strcmp-avx2.S:284
284             VMOVU   (%rdi), %ymm0
(gdb) bt
#0  __strncmp_avx2_rtm () at ../sysdeps/x86_64/multiarch/strcmp-avx2.S:284
#1  0x00007f976d18fb7d in dbmdb_import_prepare_worker_entry (wqelmnt=0x55cd73989a40)
    at ldap/servers/slapd/back-ldbm/db-mdb/mdb_import_threads.c:1347
#2  0x00007f976d1958ce in dbmdb_import_worker (param=<optimized out>)
    at ldap/servers/slapd/back-ldbm/db-mdb/mdb_import_threads.c:3191
#3  0x00007f9770ab4c34 in _pt_root (arg=0x55cd73973dc0) at pthreads/../../../../nspr/pr/src/pthreads/ptthread.c:201
#4  0x00007f977089f822 in start_thread (arg=<optimized out>) at pthread_create.c:443
#5  0x00007f977083f450 in clone3 () at ../sysdeps/unix/sysv/linux/x86_64/clone3.S:81

Full backtrace is attached, core is available at https://drive.google.com/file/d/1bGYpP0JuifLKVGdXRVMYj-SkuWCvlnXZ/view?usp=sharing


Version-Release number of selected component (if applicable):

How reproducible:
  Always

Steps to Reproduce:
1. See test case in the description

Actual results:
ns-slapd crashes


Expected results:
ns-slapd should fail without crashing.

Additional info:
Database size should be at least 50 Gb for 10 000 000 users

Crash is caused by a double free when freeing the import pipe line resources.