Add code to detect a task that exited prior to up'ing the startup mutex

[fio.git] / fio.c

diff --git a/fio.c b/fio.c
index 70d4b2b166f8208f6081c8544d2173c5ecfeee94..fa5deabb6842ed0c527c07e76f68c9c93560e5cc 100644 (file)
--- a/fio.c
+++ b/fio.c
@@ -42,7 +42,8 @@
 
 unsigned long page_mask;
 unsigned long page_size;
-#define ALIGN(buf)     \
+
+#define PAGE_ALIGN(buf)        \
        (char *) (((unsigned long) (buf) + page_mask) & ~page_mask)
 
 int groupid = 0;
@@ -186,6 +187,9 @@ static int __check_min_rate(struct thread_data *td, struct timeval *now,
        unsigned int rate_iops = 0;
        unsigned int rate_iops_min = 0;
 
+       if (!td->o.ratemin[ddir] && !td->o.rate_iops_min[ddir])
+               return 0;
+
        /*
         * allow a 2 second settle period in the beginning
         */
@@ -369,6 +373,43 @@ static inline void update_tv_cache(struct thread_data *td)
                fio_gettime(&td->tv_cache, NULL);
 }
 
+static int break_on_this_error(struct thread_data *td, int *retptr)
+{
+       int ret = *retptr;
+
+       if (ret < 0 || td->error) {
+               int err;
+
+               if (!td->o.continue_on_error)
+                       return 1;
+
+               if (ret < 0)
+                       err = -ret;
+               else
+                       err = td->error;
+
+               update_error_count(td, err);
+
+               if (td_non_fatal_error(err)) {
+                       /*
+                        * Continue with the I/Os in case of
+                        * a non fatal error.
+                        */
+                       td_clear_error(td);
+                       *retptr = 0;
+                       return 0;
+               } else {
+                       /*
+                        * Stop the I/O in case of a fatal
+                        * error.
+                        */
+                       return 1;
+               }
+       }
+
+       return 0;
+}
+
 /*
  * The main verify engine. Runs over the writes we previously submitted,
  * reads the blocks back in, and checks the crc/md5 of the data.
@@ -402,18 +443,17 @@ static void do_verify(struct thread_data *td)
        while (!td->terminate) {
                int ret2, full;
 
-               io_u = __get_io_u(td);
-               if (!io_u)
-                       break;
-
                update_tv_cache(td);
 
                if (runtime_exceeded(td, &td->tv_cache)) {
-                       put_io_u(td, io_u);
                        td->terminate = 1;
                        break;
                }
 
+               io_u = __get_io_u(td);
+               if (!io_u)
+                       break;
+
                if (get_next_verify(td, io_u)) {
                        put_io_u(td, io_u);
                        break;
@@ -429,9 +469,10 @@ static void do_verify(struct thread_data *td)
                ret = td_io_queue(td, io_u);
                switch (ret) {
                case FIO_Q_COMPLETED:
-                       if (io_u->error)
+                       if (io_u->error) {
                                ret = -io_u->error;
-                       else if (io_u->resid) {
+                               clear_io_u(td, io_u);
+                       } else if (io_u->resid) {
                                int bytes = io_u->xfer_buflen - io_u->resid;
                                struct fio_file *f = io_u->file;
 
@@ -475,7 +516,7 @@ sync_done:
                        break;
                }
 
-               if (ret < 0 || td->error)
+               if (break_on_this_error(td, &ret))
                        break;
 
                /*
@@ -539,18 +580,17 @@ static void do_io(struct thread_data *td)
                if (td->terminate)
                        break;
 
-               io_u = get_io_u(td);
-               if (!io_u)
-                       break;
-
                update_tv_cache(td);
 
                if (runtime_exceeded(td, &td->tv_cache)) {
-                       put_io_u(td, io_u);
                        td->terminate = 1;
                        break;
                }
 
+               io_u = get_io_u(td);
+               if (!io_u)
+                       break;
+
                /*
                 * Add verification end_io handler, if asked to verify
                 * a previously written file.
@@ -566,9 +606,10 @@ static void do_io(struct thread_data *td)
                ret = td_io_queue(td, io_u);
                switch (ret) {
                case FIO_Q_COMPLETED:
-                       if (io_u->error)
+                       if (io_u->error) {
                                ret = -io_u->error;
-                       else if (io_u->resid) {
+                               clear_io_u(td, io_u);
+                       } else if (io_u->resid) {
                                int bytes = io_u->xfer_buflen - io_u->resid;
                                struct fio_file *f = io_u->file;
 
@@ -623,7 +664,7 @@ sync_done:
                        break;
                }
 
-               if (ret < 0 || td->error)
+               if (break_on_this_error(td, &ret))
                        break;
 
                /*
@@ -750,8 +791,8 @@ static int init_io_u(struct thread_data *td)
        if (allocate_io_mem(td))
                return 1;
 
-       if (td->o.odirect)
-               p = ALIGN(td->orig_buffer);
+       if (td->o.mem_align)
+               p = PAGE_ALIGN(td->orig_buffer) + td->o.mem_align;
        else
                p = td->orig_buffer;
 
@@ -771,9 +812,11 @@ static int init_io_u(struct thread_data *td)
                io_u = ptr;
                memset(io_u, 0, sizeof(*io_u));
                INIT_FLIST_HEAD(&io_u->list);
+               dprint(FD_MEM, "io_u alloc %p, index %u\n", io_u, i);
 
                if (!(td->io_ops->flags & FIO_NOIO)) {
                        io_u->buf = p + max_bs * i;
+                       dprint(FD_MEM, "io_u %p, mem %p\n", io_u, io_u->buf);
 
                        if (td_write(td) && !td->o.refill_buffers)
                                io_u_fill_buffer(td, io_u, max_bs);
@@ -1368,8 +1411,10 @@ static void run_threads(void)
                         * its own files. so close them, if we opened them
                         * for creation
                         */
-                       for_each_file(td, f, i)
-                               td_io_close_file(td, f);
+                       for_each_file(td, f, i) {
+                               if (fio_file_open(f))
+                                       td_io_close_file(td, f);
+                       }
                }
 
                init_disk_util(td);
@@ -1447,7 +1492,13 @@ static void run_threads(void)
                                        *fio_debug_jobp = pid;
                        }
                        dprint(FD_MUTEX, "wait on startup_mutex\n");
-                       fio_mutex_down(startup_mutex);
+                       if (fio_mutex_down_timeout(startup_mutex, 10)) {
+                               log_err("fio: job startup hung? exiting.\n");
+                               terminate_threads(TERMINATE_ALL);
+                               fio_abort = 1;
+                               nr_started--;
+                               break;
+                       }
                        dprint(FD_MUTEX, "done waiting on startup_mutex\n");
                }