|
I encountered this very same issue when testing MDEV-12545. With that patch, it always repeats.
In the reported case, just like on my local system, we have this hang:
Thread 28 (Thread 0x7f0385b91700 (LWP 24959)):
|
#0 0x00007f038cfc7c1d in nanosleep () at ../sysdeps/unix/syscall-template.S:84
|
#1 0x000056159e95764c in os_thread_sleep (tm=tm@entry=20000) at /source/storage/innobase/os/os0thread.cc:220
|
#2 0x000056159eb16ccb in fil_space_crypt_close_tablespace (space=space@entry=0x7f033c035cd0) at /source/storage/innobase/fil/fil0crypt.cc:2359
|
#3 0x000056159eb05d32 in fil_check_pending_operations (id=id@entry=6, operation=operation@entry=FIL_OPERATION_DELETE, space=space@entry=0x7f0385b89938, path=path@entry=0x7f0385b89930) at /source/storage/innobase/fil/fil0fil.cc:2940
|
#4 0x000056159eb1047c in fil_delete_tablespace (id=id@entry=6, buf_remove=buf_remove@entry=BUF_REMOVE_FLUSH_NO_WRITE) at /source/storage/innobase/fil/fil0fil.cc:3078
|
#5 0x000056159e9b7192 in row_drop_single_table_tablespace (table_flags=33, filepath=0x7f033c033fe0 "./test/t1.ibd", tablename=<optimized out>, space_id=6) at /source/storage/innobase/row/row0mysql.cc:3566
|
#6 row_drop_table_for_mysql (name=name@entry=0x7f0385b8a4d0 "test/t1", trx=trx@entry=0x7f038673bb28, drop_db=<optimized out>, create_failed=create_failed@entry=0, nonatomic=<optimized out>, nonatomic@entry=true) at /source/storage/innobase/row/row0mysql.cc:4088
|
I got it for a different table, but that does not matter.
I believe that this is the relevant fix:
@@ -2174,6 +2168,8 @@ DECLARE_THREAD(fil_crypt_thread)(
|
/* If space is marked as stopping, release
|
space and stop rotation. */
|
if (thr.space->is_stopping()) {
|
+ fil_crypt_complete_rotate_space(
|
+ &new_state, &thr);
|
fil_space_release(thr.space);
|
thr.space = NULL;
|
break;
|
I would also do some other cleanup in the same fix, removing some unnecessary null-pointer checks.
I believe that this could hang in 10.1 as well.
|
10.2-encryption.mtr.log.bz2