X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=io_u.c;h=5b9d483dcd330b69ff0ddff4f86cb9512247cf31;hp=acc1a7b1dcb9f4ab6ac2c56a1552247c905ab41e;hb=78d55e721268e2fb9bea707529db89bacf9964ef;hpb=002fe73409d1e3d5e7dfe2885f75885bfaf506bc

diff --git a/io_u.c b/io_u.c
index acc1a7b1..5b9d483d 100644
--- a/io_u.c
+++ b/io_u.c
@@ -104,7 +104,7 @@ static int __get_next_rand_offset(struct thread_data *td, struct fio_file *f,
 
 		dprint(FD_RANDOM, "off rand %llu\n", (unsigned long long) r);
 
-		*b = (lastb - 1) * (r / ((uint64_t) rmax + 1.0));
+		*b = lastb * (r / ((uint64_t) rmax + 1.0));
 	} else {
 		uint64_t off = 0;
 
@@ -273,7 +273,7 @@ static int get_next_seq_offset(struct thread_data *td, struct fio_file *f,
 {
 	assert(ddir_rw(ddir));
 
-	if (f->last_pos >= f->io_size + get_start_offset(td) && td->o.time_based)
+	if (f->last_pos >= f->io_size + get_start_offset(td, f) && td->o.time_based)
 		f->last_pos = f->last_pos - f->io_size;
 
 	if (f->last_pos < f->real_file_size) {
@@ -415,7 +415,7 @@ static inline int io_u_fits(struct thread_data *td, struct io_u *io_u,
 {
 	struct fio_file *f = io_u->file;
 
-	return io_u->offset + buflen <= f->io_size + get_start_offset(td);
+	return io_u->offset + buflen <= f->io_size + get_start_offset(td, f);
 }
 
 static unsigned int __get_next_buflen(struct thread_data *td, struct io_u *io_u,
@@ -426,12 +426,10 @@ static unsigned int __get_next_buflen(struct thread_data *td, struct io_u *io_u,
 	unsigned int minbs, maxbs;
 	unsigned long r, rand_max;
 
-	assert(ddir_rw(io_u->ddir));
+	assert(ddir_rw(ddir));
 
 	if (td->o.bs_is_seq_rand)
 		ddir = is_random ? DDIR_WRITE: DDIR_READ;
-	else
-		ddir = io_u->ddir;
 
 	minbs = td->o.min_bs[ddir];
 	maxbs = td->o.max_bs[ddir];
@@ -680,7 +678,7 @@ static void set_rw_ddir(struct thread_data *td, struct io_u *io_u)
 
 void put_file_log(struct thread_data *td, struct fio_file *f)
 {
-	int ret = put_file(td, f);
+	unsigned int ret = put_file(td, f);
 
 	if (ret)
 		td_verror(td, ret, "file close");
@@ -1093,13 +1091,16 @@ static struct fio_file *__get_next_file(struct thread_data *td)
 	td->file_service_file = f;
 	td->file_service_left = td->file_service_nr - 1;
 out:
-	dprint(FD_FILE, "get_next_file: %p [%s]\n", f, f->file_name);
+	if (f)
+		dprint(FD_FILE, "get_next_file: %p [%s]\n", f, f->file_name);
+	else
+		dprint(FD_FILE, "get_next_file: NULL\n");
 	return f;
 }
 
 static struct fio_file *get_next_file(struct thread_data *td)
 {
-	if (!(td->flags & TD_F_PROFILE_OPS)) {
+	if (td->flags & TD_F_PROFILE_OPS) {
 		struct prof_io_ops *ops = &td->prof_io_ops;
 
 		if (ops->get_next_file)
@@ -1162,6 +1163,10 @@ static int __lat_target_failed(struct thread_data *td)
 		return 1;
 
 	td->latency_qd_high = td->latency_qd;
+
+	if (td->latency_qd == td->latency_qd_low)
+		td->latency_qd_low--;
+
 	td->latency_qd = (td->latency_qd + td->latency_qd_low) / 2;
 
 	dprint(FD_RATE, "Ramped down: %d %d %d\n", td->latency_qd_low, td->latency_qd, td->latency_qd_high);
@@ -1186,6 +1191,8 @@ static int lat_target_failed(struct thread_data *td)
 
 void lat_target_init(struct thread_data *td)
 {
+	td->latency_end_run = 0;
+
 	if (td->o.latency_target) {
 		dprint(FD_RATE, "Latency target=%llu\n", td->o.latency_target);
 		fio_gettime(&td->latency_ts, NULL);
@@ -1197,9 +1204,16 @@ void lat_target_init(struct thread_data *td)
 		td->latency_qd = td->o.iodepth;
 }
 
+void lat_target_reset(struct thread_data *td)
+{
+	if (!td->latency_end_run)
+		lat_target_init(td);
+}
+
 static void lat_target_success(struct thread_data *td)
 {
 	const unsigned int qd = td->latency_qd;
+	struct thread_options *o = &td->o;
 
 	td->latency_qd_low = td->latency_qd;
 
@@ -1208,20 +1222,32 @@ static void lat_target_success(struct thread_data *td)
 	 * of bisecting from highest possible queue depth. If we have set
 	 * a limit other than td->o.iodepth, bisect between that.
 	 */
-	if (td->latency_qd_high != td->o.iodepth)
+	if (td->latency_qd_high != o->iodepth)
 		td->latency_qd = (td->latency_qd + td->latency_qd_high) / 2;
 	else
 		td->latency_qd *= 2;
 
-	if (td->latency_qd > td->o.iodepth)
-		td->latency_qd = td->o.iodepth;
+	if (td->latency_qd > o->iodepth)
+		td->latency_qd = o->iodepth;
 
 	dprint(FD_RATE, "Ramped up: %d %d %d\n", td->latency_qd_low, td->latency_qd, td->latency_qd_high);
+
 	/*
-	 * Same as last one, we are done
+	 * Same as last one, we are done. Let it run a latency cycle, so
+	 * we get only the results from the targeted depth.
 	 */
-	if (td->latency_qd == qd)
-		td->done = 1;
+	if (td->latency_qd == qd) {
+		if (td->latency_end_run) {
+			dprint(FD_RATE, "We are done\n");
+			td->done = 1;
+		} else {
+			dprint(FD_RATE, "Quiesce and final run\n");
+			io_u_quiesce(td);
+			td->latency_end_run = 1;
+			reset_all_stats(td);
+			reset_io_stats(td);
+		}
+	}
 
 	lat_new_cycle(td);
 }
@@ -1269,7 +1295,7 @@ int queue_full(struct thread_data *td)
 
 struct io_u *__get_io_u(struct thread_data *td)
 {
-	struct io_u *io_u;
+	struct io_u *io_u = NULL;
 
 	td_io_u_lock(td);
 
@@ -1279,9 +1305,9 @@ again:
 	else if (!queue_full(td)) {
 		io_u = io_u_qpop(&td->io_u_freelist);
 
+		io_u->file = NULL;
 		io_u->buflen = 0;
 		io_u->resid = 0;
-		io_u->file = NULL;
 		io_u->end_io = NULL;
 	}
 
@@ -1462,7 +1488,8 @@ struct io_u *get_io_u(struct thread_data *td)
 			if (td->flags & TD_F_REFILL_BUFFERS) {
 				io_u_fill_buffer(td, io_u,
 					io_u->xfer_buflen, io_u->xfer_buflen);
-			} else if (td->flags & TD_F_SCRAMBLE_BUFFERS)
+			} else if ((td->flags & TD_F_SCRAMBLE_BUFFERS) &&
+				   !(td->flags & TD_F_COMPRESS))
 				do_scramble = 1;
 			if (td->flags & TD_F_VER_NONE) {
 				populate_verify_io_u(td, io_u);
@@ -1521,20 +1548,26 @@ void io_u_log_error(struct thread_data *td, struct io_u *io_u)
 		td_verror(td, io_u->error, "io_u error");
 }
 
+static inline int gtod_reduce(struct thread_data *td)
+{
+	return td->o.disable_clat && td->o.disable_lat && td->o.disable_slat
+		&& td->o.disable_bw;
+}
+
 static void account_io_completion(struct thread_data *td, struct io_u *io_u,
 				  struct io_completion_data *icd,
 				  const enum fio_ddir idx, unsigned int bytes)
 {
 	unsigned long lusec = 0;
 
-	if (!td->o.disable_clat || !td->o.disable_bw)
+	if (!gtod_reduce(td))
 		lusec = utime_since(&io_u->issue_time, &icd->time);
 
 	if (!td->o.disable_lat) {
 		unsigned long tusec;
 
 		tusec = utime_since(&io_u->start_time, &icd->time);
-		add_lat_sample(td, idx, tusec, bytes);
+		add_lat_sample(td, idx, tusec, bytes, io_u->offset);
 
 		if (td->flags & TD_F_PROFILE_OPS) {
 			struct prof_io_ops *ops = &td->prof_io_ops;
@@ -1552,17 +1585,15 @@ static void account_io_completion(struct thread_data *td, struct io_u *io_u,
 	}
 
 	if (!td->o.disable_clat) {
-		add_clat_sample(td, idx, lusec, bytes);
+		add_clat_sample(td, idx, lusec, bytes, io_u->offset);
 		io_u_mark_latency(td, lusec);
 	}
 
 	if (!td->o.disable_bw)
 		add_bw_sample(td, idx, bytes, &icd->time);
 
-	add_iops_sample(td, idx, bytes, &icd->time);
-
-	if (td->o.number_ios && !--td->o.number_ios)
-		td->done = 1;
+	if (!gtod_reduce(td))
+		add_iops_sample(td, idx, bytes, &icd->time);
 }
 
 static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)
@@ -1591,8 +1622,15 @@ static void io_completed(struct thread_data *td, struct io_u *io_u,
 	 * Mark IO ok to verify
 	 */
 	if (io_u->ipo) {
-		io_u->ipo->flags &= ~IP_F_IN_FLIGHT;
-		write_barrier();
+		/*
+		 * Remove errored entry from the verification list
+		 */
+		if (io_u->error)
+			unlog_io_piece(td, io_u);
+		else {
+			io_u->ipo->flags &= ~IP_F_IN_FLIGHT;
+			write_barrier();
+		}
 	}
 
 	td_io_u_unlock(td);
@@ -1680,7 +1718,8 @@ static void init_icd(struct thread_data *td, struct io_completion_data *icd,
 		     int nr)
 {
 	int ddir;
-	if (!td->o.disable_clat || !td->o.disable_bw)
+
+	if (!gtod_reduce(td))
 		fio_gettime(&icd->time, NULL);
 
 	icd->nr = nr;
@@ -1784,7 +1823,8 @@ void io_u_queued(struct thread_data *td, struct io_u *io_u)
 		unsigned long slat_time;
 
 		slat_time = utime_since(&io_u->start_time, &io_u->issue_time);
-		add_slat_sample(td, io_u->ddir, slat_time, io_u->xfer_buflen);
+		add_slat_sample(td, io_u->ddir, slat_time, io_u->xfer_buflen,
+				io_u->offset);
 	}
 }