X-Git-Url: https://git.kernel.dk/?p=fio.git;a=blobdiff_plain;f=io_u.c;h=612057d15762dfa13e7e48bc48e99658f9e8880a;hp=8e27708731c7290abc4e468000ec045d856e1241;hb=effd6ff0d9f701a64d6bc46509eedfdcfd12f553;hpb=0cae66f6df9eb63d4896c29d7998d4aaf964fea5

diff --git a/io_u.c b/io_u.c
index 8e277087..612057d1 100644
--- a/io_u.c
+++ b/io_u.c
@@ -223,7 +223,7 @@ static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
 	if (!flist_empty(&td->next_rand_list)) {
 		struct rand_off *r;
 fetch:
-		r = flist_entry(td->next_rand_list.next, struct rand_off, list);
+		r = flist_first_entry(&td->next_rand_list, struct rand_off, list);
 		flist_del(&r->list);
 		*b = r->off;
 		free(r);
@@ -271,20 +271,32 @@ static int get_next_rand_block(struct thread_data *td, struct fio_file *f,
 static int get_next_seq_offset(struct thread_data *td, struct fio_file *f,
 			       enum fio_ddir ddir, uint64_t *offset)
 {
+	struct thread_options *o = &td->o;
+
 	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) &&
+	    o->time_based)
 		f->last_pos = f->last_pos - f->io_size;
 
 	if (f->last_pos < f->real_file_size) {
 		uint64_t pos;
 
-		if (f->last_pos == f->file_offset && td->o.ddir_seq_add < 0)
+		if (f->last_pos == f->file_offset && o->ddir_seq_add < 0)
 			f->last_pos = f->real_file_size;
 
 		pos = f->last_pos - f->file_offset;
-		if (pos)
-			pos += td->o.ddir_seq_add;
+		if (pos && o->ddir_seq_add) {
+			pos += o->ddir_seq_add;
+
+			/*
+			 * If we reach beyond the end of the file
+			 * with holed IO, wrap around to the
+			 * beginning again.
+			 */
+			if (pos >= f->real_file_size)
+				pos = f->file_offset;
+		}
 
 		*offset = pos;
 		return 0;
@@ -415,7 +427,7 @@ 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 +438,10 @@ 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 +690,7 @@ 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");
@@ -690,10 +700,10 @@ void put_io_u(struct thread_data *td, struct io_u *io_u)
 {
 	td_io_u_lock(td);
 
-	if (io_u->file && !(io_u->flags & IO_U_F_FREE_DEF))
+	if (io_u->file && !(io_u->flags & IO_U_F_NO_FILE_PUT))
 		put_file_log(td, io_u->file);
+
 	io_u->file = NULL;
-	io_u->flags &= ~IO_U_F_FREE_DEF;
 	io_u->flags |= IO_U_F_FREE;
 
 	if (io_u->flags & IO_U_F_IN_CUR_DEPTH)
@@ -750,9 +760,17 @@ static int fill_io_u(struct thread_data *td, struct io_u *io_u)
 	 * See if it's time to switch to a new zone
 	 */
 	if (td->zone_bytes >= td->o.zone_size && td->o.zone_skip) {
+		struct fio_file *f = io_u->file;
+
 		td->zone_bytes = 0;
-		io_u->file->file_offset += td->o.zone_range + td->o.zone_skip;
-		io_u->file->last_pos = io_u->file->file_offset;
+		f->file_offset += td->o.zone_range + td->o.zone_skip;
+
+		/*
+		 * Wrap from the beginning, if we exceed the file size
+		 */
+		if (f->file_offset >= f->real_file_size)
+			f->file_offset = f->real_file_size - f->file_offset;
+		f->last_pos = f->file_offset;
 		td->io_skip_bytes += td->o.zone_skip;
 	}
 
@@ -1102,7 +1120,7 @@ out:
 
 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)
@@ -1283,7 +1301,7 @@ void lat_target_check(struct thread_data *td)
  * If latency target is enabled, we might be ramping up or down and not
  * using the full queue depth available.
  */
-int queue_full(struct thread_data *td)
+int queue_full(const struct thread_data *td)
 {
 	const int qempty = io_u_qempty(&td->io_u_freelist);
 
@@ -1315,9 +1333,9 @@ again:
 
 	if (io_u) {
 		assert(io_u->flags & IO_U_F_FREE);
-		io_u->flags &= ~(IO_U_F_FREE | IO_U_F_FREE_DEF);
-		io_u->flags &= ~(IO_U_F_TRIMMED | IO_U_F_BARRIER);
-		io_u->flags &= ~IO_U_F_VER_LIST;
+		io_u->flags &= ~(IO_U_F_FREE | IO_U_F_NO_FILE_PUT |
+				 IO_U_F_TRIMMED | IO_U_F_BARRIER |
+				 IO_U_F_VER_LIST);
 
 		io_u->error = 0;
 		io_u->acct_ddir = -1;
@@ -1489,8 +1507,10 @@ struct io_u *get_io_u(struct thread_data *td)
 		if (io_u->ddir == DDIR_WRITE) {
 			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)
+					td->o.min_bs[DDIR_WRITE],
+					io_u->xfer_buflen);
+			} 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);
@@ -1529,21 +1549,16 @@ err_put:
 void io_u_log_error(struct thread_data *td, struct io_u *io_u)
 {
 	enum error_type_bit eb = td_error_type(io_u->ddir, io_u->error);
-	const char *msg[] = { "read", "write", "sync", "datasync",
-				"sync_file_range", "wait", "trim" };
 
 	if (td_non_fatal_error(td, eb, io_u->error) && !td->o.error_dump)
 		return;
 
-	log_err("fio: io_u error");
-
-	if (io_u->file)
-		log_err(" on file %s", io_u->file->file_name);
-
-	log_err(": %s\n", strerror(io_u->error));
-
-	log_err("     %s offset=%llu, buflen=%lu\n", msg[io_u->ddir],
-					io_u->offset, io_u->xfer_buflen);
+	log_err("fio: io_u error%s%s: %s: %s offset=%llu, buflen=%lu\n",
+		io_u->file ? " on file " : "",
+		io_u->file ? io_u->file->file_name : "",
+		strerror(io_u->error),
+		io_ddir_name(io_u->ddir),
+		io_u->offset, io_u->xfer_buflen);
 
 	if (!td->error)
 		td_verror(td, io_u->error, "io_u error");
@@ -1568,7 +1583,7 @@ static void account_io_completion(struct thread_data *td, struct io_u *io_u,
 		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;
@@ -1586,7 +1601,7 @@ 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);
 	}
 
@@ -1595,9 +1610,6 @@ static void account_io_completion(struct thread_data *td, struct io_u *io_u,
 
 	if (!gtod_reduce(td))
 		add_iops_sample(td, idx, bytes, &icd->time);
-
-	if (td->o.number_ios && !--td->o.number_ios)
-		td->done = 1;
 }
 
 static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)
@@ -1611,10 +1623,12 @@ static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)
 	return remainder * 1000000 / bps + secs * 1000000;
 }
 
-static void io_completed(struct thread_data *td, struct io_u *io_u,
+static void io_completed(struct thread_data *td, struct io_u **io_u_ptr,
 			 struct io_completion_data *icd)
 {
-	struct fio_file *f;
+	struct io_u *io_u = *io_u_ptr;
+	enum fio_ddir ddir = io_u->ddir;
+	struct fio_file *f = io_u->file;
 
 	dprint_io_u(io_u, "io complete");
 
@@ -1626,15 +1640,21 @@ 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);
 
-	if (ddir_sync(io_u->ddir)) {
+	if (ddir_sync(ddir)) {
 		td->last_was_sync = 1;
-		f = io_u->file;
 		if (f) {
 			f->first_write = -1ULL;
 			f->last_write = -1ULL;
@@ -1643,52 +1663,51 @@ static void io_completed(struct thread_data *td, struct io_u *io_u,
 	}
 
 	td->last_was_sync = 0;
-	td->last_ddir = io_u->ddir;
+	td->last_ddir = ddir;
 
-	if (!io_u->error && ddir_rw(io_u->ddir)) {
+	if (!io_u->error && ddir_rw(ddir)) {
 		unsigned int bytes = io_u->buflen - io_u->resid;
-		const enum fio_ddir idx = io_u->ddir;
-		const enum fio_ddir odx = io_u->ddir ^ 1;
+		const enum fio_ddir oddir = ddir ^ 1;
 		int ret;
 
-		td->io_blocks[idx]++;
-		td->this_io_blocks[idx]++;
-		td->io_bytes[idx] += bytes;
+		td->io_blocks[ddir]++;
+		td->this_io_blocks[ddir]++;
+		td->io_bytes[ddir] += bytes;
 
 		if (!(io_u->flags & IO_U_F_VER_LIST))
-			td->this_io_bytes[idx] += bytes;
-
-		if (idx == DDIR_WRITE) {
-			f = io_u->file;
-			if (f) {
-				if (f->first_write == -1ULL ||
-				    io_u->offset < f->first_write)
-					f->first_write = io_u->offset;
-				if (f->last_write == -1ULL ||
-				    ((io_u->offset + bytes) > f->last_write))
-					f->last_write = io_u->offset + bytes;
-			}
+			td->this_io_bytes[ddir] += bytes;
+
+		if (ddir == DDIR_WRITE && f) {
+			if (f->first_write == -1ULL ||
+			    io_u->offset < f->first_write)
+				f->first_write = io_u->offset;
+			if (f->last_write == -1ULL ||
+			    ((io_u->offset + bytes) > f->last_write))
+				f->last_write = io_u->offset + bytes;
 		}
 
 		if (ramp_time_over(td) && (td->runstate == TD_RUNNING ||
 					   td->runstate == TD_VERIFYING)) {
-			account_io_completion(td, io_u, icd, idx, bytes);
+			account_io_completion(td, io_u, icd, ddir, bytes);
 
-			if (__should_check_rate(td, idx)) {
-				td->rate_pending_usleep[idx] =
-					(usec_for_io(td, idx) -
+			if (__should_check_rate(td, ddir)) {
+				td->rate_pending_usleep[ddir] =
+					(usec_for_io(td, ddir) -
 					 utime_since_now(&td->start));
 			}
-			if (idx != DDIR_TRIM && __should_check_rate(td, odx))
-				td->rate_pending_usleep[odx] =
-					(usec_for_io(td, odx) -
+			if (ddir != DDIR_TRIM &&
+			    __should_check_rate(td, oddir)) {
+				td->rate_pending_usleep[oddir] =
+					(usec_for_io(td, oddir) -
 					 utime_since_now(&td->start));
+			}
 		}
 
-		icd->bytes_done[idx] += bytes;
+		icd->bytes_done[ddir] += bytes;
 
 		if (io_u->end_io) {
-			ret = io_u->end_io(td, io_u);
+			ret = io_u->end_io(td, io_u_ptr);
+			io_u = *io_u_ptr;
 			if (ret && !icd->error)
 				icd->error = ret;
 		}
@@ -1697,9 +1716,11 @@ static void io_completed(struct thread_data *td, struct io_u *io_u,
 		io_u_log_error(td, io_u);
 	}
 	if (icd->error) {
-		enum error_type_bit eb = td_error_type(io_u->ddir, icd->error);
+		enum error_type_bit eb = td_error_type(ddir, icd->error);
+
 		if (!td_non_fatal_error(td, eb, icd->error))
 			return;
+
 		/*
 		 * If there is a non_fatal error, then add to the error count
 		 * and clear all the errors.
@@ -1707,7 +1728,8 @@ static void io_completed(struct thread_data *td, struct io_u *io_u,
 		update_error_count(td, icd->error);
 		td_clear_error(td);
 		icd->error = 0;
-		io_u->error = 0;
+		if (io_u)
+			io_u->error = 0;
 	}
 }
 
@@ -1735,9 +1757,9 @@ static void ios_completed(struct thread_data *td,
 	for (i = 0; i < icd->nr; i++) {
 		io_u = td->io_ops->event(td, i);
 
-		io_completed(td, io_u, icd);
+		io_completed(td, &io_u, icd);
 
-		if (!(io_u->flags & IO_U_F_FREE_DEF))
+		if (io_u)
 			put_io_u(td, io_u);
 	}
 }
@@ -1751,9 +1773,9 @@ int io_u_sync_complete(struct thread_data *td, struct io_u *io_u,
 	struct io_completion_data icd;
 
 	init_icd(td, &icd, 1);
-	io_completed(td, io_u, &icd);
+	io_completed(td, &io_u, &icd);
 
-	if (!(io_u->flags & IO_U_F_FREE_DEF))
+	if (io_u)
 		put_io_u(td, io_u);
 
 	if (icd.error) {
@@ -1786,6 +1808,8 @@ int io_u_queued_complete(struct thread_data *td, int min_evts,
 
 	if (!min_evts)
 		tvp = &ts;
+	else if (min_evts > td->cur_depth)
+		min_evts = td->cur_depth;
 
 	ret = td_io_getevents(td, min_evts, td->o.iodepth_batch_complete, tvp);
 	if (ret < 0) {
@@ -1820,10 +1844,39 @@ 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);
 	}
 }
 
+/*
+ * See if we should reuse the last seed, if dedupe is enabled
+ */
+static struct frand_state *get_buf_state(struct thread_data *td)
+{
+	unsigned int v;
+	unsigned long r;
+
+	if (!td->o.dedupe_percentage)
+		return &td->buf_state;
+	else if (td->o.dedupe_percentage == 100)
+		return &td->buf_state_prev;
+
+	r = __rand(&td->dedupe_state);
+	v = 1 + (int) (100.0 * (r / (FRAND_MAX + 1.0)));
+
+	if (v <= td->o.dedupe_percentage)
+		return &td->buf_state_prev;
+
+	return &td->buf_state;
+}
+
+static void save_buf_state(struct thread_data *td, struct frand_state *rs)
+{
+	if (rs == &td->buf_state)
+		frand_copy(&td->buf_state_prev, rs);
+}
+
 void fill_io_buffer(struct thread_data *td, void *buf, unsigned int min_write,
 		    unsigned int max_bs)
 {
@@ -1831,18 +1884,30 @@ void fill_io_buffer(struct thread_data *td, void *buf, unsigned int min_write,
 		fill_buffer_pattern(td, buf, max_bs);
 	else if (!td->o.zero_buffers) {
 		unsigned int perc = td->o.compress_percentage;
+		struct frand_state *rs;
+		unsigned int left = max_bs;
+
+		do {
+			rs = get_buf_state(td);
+
+			min_write = min(min_write, left);
+
+			if (perc) {
+				unsigned int seg = min_write;
 
-		if (perc) {
-			unsigned int seg = min_write;
+				seg = min(min_write, td->o.compress_chunk);
+				if (!seg)
+					seg = min_write;
 
-			seg = min(min_write, td->o.compress_chunk);
-			if (!seg)
-				seg = min_write;
+				fill_random_buf_percentage(rs, buf, perc, seg,
+								min_write);
+			} else
+				fill_random_buf(rs, buf, min_write);
 
-			fill_random_buf_percentage(&td->buf_state, buf,
-						perc, seg, max_bs);
-		} else
-			fill_random_buf(&td->buf_state, buf, max_bs);
+			buf += min_write;
+			left -= min_write;
+			save_buf_state(td, rs);
+		} while (left);
 	} else
 		memset(buf, 0, max_bs);
 }