DragonFlyBSD

A pthread_t variable doesn't need to be initialized with NULL prior to calling pthread_create(3).
This also isn't portable since pthread_t is an opaque type (i.e. may not be a pointer, and in fact it's not on Linux/glibc).

1. grep "thread = NULL" sbin/hammer2 rI
   sbin/hammer2/subs.c: pthread_t thread = NULL;
   sbin/hammer2/cmd_service.c: thread = NULL;
   sbin/hammer2/cmd_service.c: thread = NULL;
   sbin/hammer2/cmd_service.c: thread = NULL;
   sbin/hammer2/cmd_service.c: conf>thread = NULL;
   sbin/hammer2/cmd_service.c: conf->iocom_thread = NULL;
   sbin/hammer2/cmd_service.c: thread = NULL;

It looks like above exists for this conditional below.
If this could be rewritten without testing pthread_t itself (it shouldn't be too), NULL initialization could be removed.

369                 if (conf->thread == NULL) {
 370                         fprintf(stderr, "VOLCONF THREAD STARTED\n");
 371                         pthread_cond_init(&conf->cond, NULL);
 372                         pthread_create(&conf->thread, NULL,
 373                                        hammer2_volconf_thread, (void *)conf);
 374                 }