X-Git-Url: https://git.kernel.dk/?a=blobdiff_plain;ds=sidebyside;f=backend.c;h=950333034e749154eb082e8ab3169c2794276aef;hb=8f380de25d7c9753223069d1536e344f818309e9;hp=7de6f654d153b2f2798450ed4298a6d2c6763a2d;hpb=a39fb9ea87afae396a882961fe65d87e548a1a1b;p=fio.git

diff --git a/backend.c b/backend.c
index 7de6f654..95033303 100644
--- a/backend.c
+++ b/backend.c
@@ -441,6 +441,12 @@ static int wait_for_completions(struct thread_data *td, struct timeval *time)
 	int min_evts = 0;
 	int ret;
 
+	if (td->flags & TD_F_REGROW_LOGS) {
+		ret = io_u_quiesce(td);
+		regrow_logs(td);
+		return ret;
+	}
+
 	/*
 	 * if the queue is full, we MUST reap at least 1 event
 	 */
@@ -518,6 +524,9 @@ sync_done:
 				break;
 		}
 
+		if (td->flags & TD_F_REGROW_LOGS)
+			regrow_logs(td);
+
 		/*
 		 * when doing I/O (not when verifying),
 		 * check for any errors that are to be ignored
@@ -594,6 +603,15 @@ static void do_verify(struct thread_data *td, uint64_t verify_bytes)
 	if (td->error)
 		return;
 
+	/*
+	 * verify_state needs to be reset before verification
+	 * proceeds so that expected random seeds match actual
+	 * random seeds in headers. The main loop will reset
+	 * all random number generators if randrepeat is set.
+	 */
+	if (!td->o.rand_repeatable)
+		td_fill_verify_state_seed(td);
+
 	td_set_runstate(td, TD_VERIFYING);
 
 	io_u = NULL;
@@ -1419,7 +1437,6 @@ static void *thread_main(void *data)
 	struct thread_data *td = fd->td;
 	struct thread_options *o = &td->o;
 	struct sk_out *sk_out = fd->sk_out;
-	pthread_condattr_t attr;
 	int clear_state;
 	int ret;
 
@@ -1444,12 +1461,18 @@ static void *thread_main(void *data)
 	INIT_FLIST_HEAD(&td->verify_list);
 	INIT_FLIST_HEAD(&td->trim_list);
 	INIT_FLIST_HEAD(&td->next_rand_list);
-	pthread_mutex_init(&td->io_u_lock, NULL);
 	td->io_hist_tree = RB_ROOT;
 
-	pthread_condattr_init(&attr);
-	pthread_cond_init(&td->verify_cond, &attr);
-	pthread_cond_init(&td->free_cond, &attr);
+	ret = mutex_cond_init_pshared(&td->io_u_lock, &td->free_cond);
+	if (ret) {
+		td_verror(td, ret, "mutex_cond_init_pshared");
+		goto err;
+	}
+	ret = cond_init_pshared(&td->verify_cond);
+	if (ret) {
+		td_verror(td, ret, "mutex_cond_pshared");
+		goto err;
+	}
 
 	td_set_runstate(td, TD_INITIALIZED);
 	dprint(FD_MUTEX, "up startup_mutex\n");
@@ -1471,6 +1494,14 @@ static void *thread_main(void *data)
 		goto err;
 	}
 
+	/*
+	 * Do this early, we don't want the compress threads to be limited
+	 * to the same CPUs as the IO workers. So do this before we set
+	 * any potential CPU affinity
+	 */
+	if (iolog_compress_init(td, sk_out))
+		goto err;
+
 	/*
 	 * If we have a gettimeofday() thread, make sure we exclude that
 	 * thread from this job
@@ -1605,9 +1636,6 @@ static void *thread_main(void *data)
 			goto err;
 	}
 
-	if (iolog_compress_init(td, sk_out))
-		goto err;
-
 	fio_verify_init(td);
 
 	if (rate_submit_init(td, sk_out))
@@ -1705,6 +1733,8 @@ static void *thread_main(void *data)
 			break;
 	}
 
+	td_set_runstate(td, TD_FINISHING);
+
 	update_rusage_stat(td);
 	td->ts.total_run_time = mtime_since_now(&td->epoch);
 	td->ts.io_bytes[DDIR_READ] = td->io_bytes[DDIR_READ];
@@ -1813,8 +1843,9 @@ static int fork_main(struct sk_out *sk_out, int shmid, int offset)
 
 static void dump_td_info(struct thread_data *td)
 {
-	log_err("fio: job '%s' hasn't exited in %lu seconds, it appears to "
-		"be stuck. Doing forceful exit of this job.\n", td->o.name,
+	log_err("fio: job '%s' (state=%d) hasn't exited in %lu seconds, it "
+		"appears to be stuck. Doing forceful exit of this job.\n",
+			td->o.name, td->runstate,
 			(unsigned long) time_since_now(&td->terminate_time));
 }
 
@@ -1900,6 +1931,7 @@ static void reap_threads(unsigned int *nr_running, unsigned int *t_rate,
 		 * move on.
 		 */
 		if (td->terminate &&
+		    td->runstate < TD_FSYNCING &&
 		    time_since_now(&td->terminate_time) >= FIO_REAP_TIMEOUT) {
 			dump_td_info(td);
 			td_set_runstate(td, TD_REAPED);