Merge branch 'master' of ssh://brick.kernel.dk/data/git/fio

[fio.git] / io_u.c

diff --git a/io_u.c b/io_u.c
index 9488adeeffe43983a03a0b8089eb400afe7db96c..8a033485e7cf9eee2adbb893ab2d13bd2695dde0 100644 (file)
--- a/io_u.c
+++ b/io_u.c
@@ -114,6 +114,9 @@ static unsigned long long last_block(struct thread_data *td, struct fio_file *f,
        if (max_size > f->real_file_size)
                max_size = f->real_file_size;
 
+       if (td->o.zone_range)
+               max_size = td->o.zone_range;
+
        max_blocks = max_size / (unsigned long long) td->o.ba[ddir];
        if (!max_blocks)
                return 0;
@@ -158,7 +161,7 @@ static int get_next_free_block(struct thread_data *td, struct fio_file *f,
 static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
                                enum fio_ddir ddir, unsigned long long *b)
 {
-       unsigned long long r, lastb;
+       unsigned long long rmax, r, lastb;
        int loops = 5;
 
        lastb = last_block(td, f, ddir);
@@ -168,14 +171,14 @@ static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
        if (f->failed_rands >= 200)
                goto ffz;
 
+       rmax = td->o.use_os_rand ? OS_RAND_MAX : FRAND_MAX;
        do {
-               if (td->o.use_os_rand) {
+               if (td->o.use_os_rand)
                        r = os_random_long(&td->random_state);
-                       *b = (lastb - 1) * (r / ((unsigned long long) OS_RAND_MAX + 1.0));
-               } else {
+               else
                        r = __rand(&td->__random_state);
-                       *b = (lastb - 1) * (r / ((unsigned long long) FRAND_MAX + 1.0));
-               }
+
+               *b = (lastb - 1) * (r / ((unsigned long long) rmax + 1.0));
 
                dprint(FD_RANDOM, "off rand %llu\n", r);
 
@@ -207,7 +210,7 @@ static int get_next_rand_offset(struct thread_data *td, struct fio_file *f,
        loops = 10;
        do {
                f->last_free_lookup = (f->num_maps - 1) *
-                                       (r / (OS_RAND_MAX + 1.0));
+                                       (r / ((unsigned long long) rmax + 1.0));
                if (!get_next_free_block(td, f, ddir, b))
                        goto ret;
 
@@ -374,6 +377,12 @@ static unsigned int __get_next_buflen(struct thread_data *td, struct io_u *io_u)
        if (minbs == maxbs)
                return minbs;
 
+       /*
+        * If we can't satisfy the min block size from here, then fail
+        */
+       if (!io_u_fits(td, io_u, minbs))
+               return 0;
+
        if (td->o.use_os_rand)
                rand_max = OS_RAND_MAX;
        else
@@ -493,12 +502,17 @@ static enum fio_ddir rate_ddir(struct thread_data *td, enum fio_ddir ddir)
 
        /*
         * We are going to sleep, ensure that we flush anything pending as
-        * not to skew our latency numbers
+        * not to skew our latency numbers.
+        *
+        * Changed to only monitor 'in flight' requests here instead of the
+        * td->cur_depth, b/c td->cur_depth does not accurately represent
+        * io's that have been actually submitted to an async engine,
+        * and cur_depth is meaningless for sync engines.
         */
-       if (td->cur_depth) {
+       if (td->io_u_in_flight) {
                int fio_unused ret;
 
-               ret = io_u_queued_complete(td, td->cur_depth, NULL);
+               ret = io_u_queued_complete(td, td->io_u_in_flight, NULL);
        }
 
        fio_gettime(&t, NULL);
@@ -597,13 +611,12 @@ void put_io_u(struct thread_data *td, struct io_u *io_u)
 {
        td_io_u_lock(td);
 
-       io_u->flags |= IO_U_F_FREE;
-       io_u->flags &= ~IO_U_F_FREE_DEF;
-
-       if (io_u->file)
+       if (io_u->file && !(io_u->flags & IO_U_F_FREE_DEF))
                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)
                td->cur_depth--;
        flist_del_init(&io_u->list);
@@ -657,7 +670,8 @@ static int fill_io_u(struct thread_data *td, struct io_u *io_u)
         */
        if (td->zone_bytes >= td->o.zone_size) {
                td->zone_bytes = 0;
-               io_u->file->last_pos += td->o.zone_skip;
+               io_u->file->file_offset += td->o.zone_range + td->o.zone_skip;
+               io_u->file->last_pos = io_u->file->file_offset;
                td->io_skip_bytes += td->o.zone_skip;
        }
 
@@ -1139,6 +1153,7 @@ static void small_content_scramble(struct io_u *io_u)
 
        p = io_u->xfer_buf;
        boffset = io_u->offset;
+       io_u->buf_filled_len = 0;
 
        for (i = 0; i < nr_blocks; i++) {
                /*
@@ -1297,6 +1312,16 @@ static void account_io_completion(struct thread_data *td, struct io_u *io_u,
        add_iops_sample(td, idx, &icd->time);
 }
 
+static long long usec_for_io(struct thread_data *td, enum fio_ddir ddir)
+{
+       unsigned long long secs, remainder, bps, bytes;
+       bytes = td->this_io_bytes[ddir];
+       bps = td->rate_bps[ddir];
+       secs = bytes / bps;
+       remainder = bytes % bps;
+       return remainder * 1000000 / bps + secs * 1000000;
+}
+
 static void io_completed(struct thread_data *td, struct io_u *io_u,
                         struct io_completion_data *icd)
 {
@@ -1355,14 +1380,12 @@ static void io_completed(struct thread_data *td, struct io_u *io_u,
 
                        if (__should_check_rate(td, idx)) {
                                td->rate_pending_usleep[idx] =
-                                       ((td->this_io_bytes[idx] *
-                                         td->rate_nsec_cycle[idx]) / 1000 -
+                                       (usec_for_io(td, idx) -
                                         utime_since_now(&td->start));
                        }
-                       if (__should_check_rate(td, idx ^ 1))
+                       if (__should_check_rate(td, odx))
                                td->rate_pending_usleep[odx] =
-                                       ((td->this_io_bytes[odx] *
-                                         td->rate_nsec_cycle[odx]) / 1000 -
+                                       (usec_for_io(td, odx) -
                                         utime_since_now(&td->start));
                }
 
@@ -1382,8 +1405,8 @@ static void io_completed(struct thread_data *td, struct io_u *io_u,
                icd->error = io_u->error;
                io_u_log_error(td, io_u);
        }
-       if (td->o.continue_on_error && icd->error &&
-           td_non_fatal_error(icd->error)) {
+       if (icd->error && td_non_fatal_error(icd->error) &&
+           (td->o.continue_on_error & td_error_type(io_u->ddir, icd->error))) {
                /*
                 * If there is a non_fatal error, then add to the error count
                 * and clear all the errors.