Cleanup end IO handling
[fio.git] / fio.c
diff --git a/fio.c b/fio.c
index 5358af2c8a8dc7bea5234c329bdf39087679a15f..78deaf66a732b6e92434cf0ccef52b66b62d5a71 100644 (file)
--- a/fio.c
+++ b/fio.c
@@ -163,20 +163,14 @@ static struct fio_file *get_next_file(struct thread_data *td)
  */
 static void cleanup_pending_aio(struct thread_data *td)
 {
-       struct timespec ts = { .tv_sec = 0, .tv_nsec = 0};
        struct list_head *entry, *n;
-       struct io_completion_data icd;
        struct io_u *io_u;
        int r;
 
        /*
         * get immediately available events, if any
         */
-       r = td_io_getevents(td, 0, td->cur_depth, &ts);
-       if (r > 0) {
-               icd.nr = r;
-               ios_completed(td, &icd);
-       }
+       io_u_queued_complete(td, 0, NULL);
 
        /*
         * now cancel remaining active events
@@ -191,13 +185,8 @@ static void cleanup_pending_aio(struct thread_data *td)
                }
        }
 
-       if (td->cur_depth) {
-               r = td_io_getevents(td, td->cur_depth, td->cur_depth, NULL);
-               if (r > 0) {
-                       icd.nr = r;
-                       ios_completed(td, &icd);
-               }
-       }
+       if (td->cur_depth)
+               io_u_queued_complete(td, td->cur_depth, NULL);
 }
 
 /*
@@ -207,7 +196,6 @@ static void cleanup_pending_aio(struct thread_data *td)
 static int fio_io_sync(struct thread_data *td, struct fio_file *f)
 {
        struct io_u *io_u = __get_io_u(td);
-       struct io_completion_data icd;
        int ret;
 
        if (!io_u)
@@ -227,27 +215,15 @@ static int fio_io_sync(struct thread_data *td, struct fio_file *f)
                put_io_u(td, io_u);
                return 1;
        } else if (ret == FIO_Q_QUEUED) {
-               ret = td_io_getevents(td, 1, td->cur_depth, NULL);
-               if (ret < 0) {
-                       td_verror(td, ret);
+               if (io_u_queued_complete(td, 1, NULL))
                        return 1;
-               }
-
-               icd.nr = ret;
-               ios_completed(td, &icd);
-               if (icd.error) {
-                       td_verror(td, icd.error);
-                       return 1;
-               }
        } else if (ret == FIO_Q_COMPLETED) {
                if (io_u->error) {
                        td_verror(td, io_u->error);
                        return 1;
                }
 
-               init_icd(&icd);
-               io_completed(td, io_u, &icd);
-               put_io_u(td, io_u);
+               io_u_sync_complete(td, io_u, NULL);
        }
 
        return 0;
@@ -261,7 +237,7 @@ static void do_verify(struct thread_data *td)
 {
        struct fio_file *f;
        struct io_u *io_u;
-       int ret, i;
+       int ret, i, min_events;
 
        /*
         * sync io first and invalidate cache, to make sure we really
@@ -295,7 +271,7 @@ requeue:
                switch (ret) {
                case FIO_Q_COMPLETED:
                        if (io_u->error)
-                               ret = io_u->error;
+                               ret = -io_u->error;
                        if (io_u->xfer_buflen != io_u->resid && io_u->resid) {
                                int bytes = io_u->xfer_buflen - io_u->resid;
 
@@ -303,32 +279,33 @@ requeue:
                                io_u->xfer_buf += bytes;
                                goto requeue;
                        }
-                       if (do_io_u_verify(td, &io_u)) {
-                               ret = -EIO;
+                       if (io_u_sync_complete(td, io_u, verify_io_u))
                                break;
-                       }
                        continue;
                case FIO_Q_QUEUED:
                        break;
                default:
                        assert(ret < 0);
-                       td_verror(td, ret);
+                       td_verror(td, -ret);
                        break;
                }
 
-               /*
-                * We get here for a queued request, in the future we
-                * want to later make this take full advantage of
-                * keeping IO in flight while verifying others.
-                */
-               ret = td_io_getevents(td, 1, 1, NULL);
                if (ret < 0)
                        break;
 
-               assert(ret == 1);
-               io_u = td->io_ops->event(td, 0);
+               /*
+                * if we can queue more, do so. but check if there are
+                * completed io_u's first.
+                */
+               min_events = 0;
+               if (queue_full(td))
+                       min_events = 1;
 
-               if (do_io_u_verify(td, &io_u))
+               /*
+                * Reap required number of io units, if any, and do the
+                * verification on them through the callback handler
+                */
+               if (io_u_queued_complete(td, min_events, verify_io_u))
                        break;
        }
 
@@ -372,7 +349,6 @@ static void do_cpuio(struct thread_data *td)
  */
 static void do_io(struct thread_data *td)
 {
-       struct io_completion_data icd;
        struct timeval s;
        unsigned long usec;
        struct fio_file *f;
@@ -381,7 +357,8 @@ static void do_io(struct thread_data *td)
        td_set_runstate(td, TD_RUNNING);
 
        while ((td->this_io_bytes[0] + td->this_io_bytes[1]) < td->io_size) {
-               struct timespec *timeout;
+               struct timeval comp_time;
+               long bytes_done = 0;
                int min_evts = 0;
                struct io_u *io_u;
 
@@ -397,6 +374,11 @@ static void do_io(struct thread_data *td)
                        break;
 
                memcpy(&s, &io_u->start_time, sizeof(s));
+
+               if (runtime_exceeded(td, &s)) {
+                       put_io_u(td, io_u);
+                       break;
+               }
 requeue:
                ret = td_io_queue(td, io_u);
 
@@ -413,9 +395,8 @@ requeue:
                                io_u->xfer_buf += bytes;
                                goto requeue;
                        }
-                       init_icd(&icd);
-                       io_completed(td, io_u, &icd);
-                       put_io_u(td, io_u);
+                       fio_gettime(&comp_time, NULL);
+                       bytes_done = io_u_sync_complete(td, io_u, NULL);
                        break;
                case FIO_Q_QUEUED:
                        break;
@@ -430,53 +411,39 @@ requeue:
 
                add_slat_sample(td, io_u->ddir, mtime_since(&io_u->start_time, &io_u->issue_time));
 
+               /*
+                * See if we need to complete some commands
+                */
                if (ret == FIO_Q_QUEUED) {
-                       if (td->cur_depth < td->iodepth) {
-                               struct timespec ts;
-
-                               ts.tv_sec = 0;
-                               ts.tv_nsec = 0;
-                               timeout = &ts;
-                               min_evts = 0;
-                       } else {
-                               timeout = NULL;
+                       min_evts = 0;
+                       if (queue_full(td))
                                min_evts = 1;
-                       }
 
-                       ret = td_io_getevents(td, min_evts, td->cur_depth, timeout);
-                       if (ret < 0) {
-                               td_verror(td, ret);
+                       fio_gettime(&comp_time, NULL);
+                       bytes_done = io_u_queued_complete(td, min_evts, NULL);
+                       if (bytes_done < 0)
                                break;
-                       } else if (!ret)
-                               continue;
-
-                       icd.nr = ret;
-                       ios_completed(td, &icd);
-                       if (icd.error) {
-                               td_verror(td, icd.error);
-                               break;
-                       }
                }
 
+               if (!bytes_done)
+                       continue;
+
                /*
                 * the rate is batched for now, it should work for batches
                 * of completions except the very first one which may look
                 * a little bursty
                 */
-               usec = utime_since(&s, &icd.time);
+               usec = utime_since(&s, &comp_time);
 
-               rate_throttle(td, usec, icd.bytes_done[td->ddir], td->ddir);
+               rate_throttle(td, usec, bytes_done, td->ddir);
 
-               if (check_min_rate(td, &icd.time)) {
+               if (check_min_rate(td, &comp_time)) {
                        if (exitall_on_terminate)
                                terminate_threads(td->groupid, 0);
                        td_verror(td, ENODATA);
                        break;
                }
 
-               if (runtime_exceeded(td, &icd.time))
-                       break;
-
                if (td->thinktime) {
                        unsigned long long b;