malloc memory still needs to be aligned, if we are doing direct IO

[fio.git] / fio.c

diff --git a/fio.c b/fio.c
index 0c297f7e4cc450a6984764baed64fcdff14a9726..ead118ec35a0d02572a5366f4fe065a904c0cf05 100644 (file)
--- a/fio.c
+++ b/fio.c
@@ -614,7 +614,6 @@ static void fill_io_buf(struct thread_data *td, struct io_u *io_u, int max_bs)
 
 static int init_io_u(struct thread_data *td)
 {
-       unsigned long long buf_size;
        struct io_u *io_u;
        unsigned int max_bs;
        int i, max_units;
@@ -626,19 +625,15 @@ static int init_io_u(struct thread_data *td)
                max_units = td->o.iodepth;
 
        max_bs = max(td->o.max_bs[DDIR_READ], td->o.max_bs[DDIR_WRITE]);
-       buf_size = (unsigned long long) max_bs * (unsigned long long) max_units;
-       buf_size += page_mask;
-       if (buf_size != (size_t) buf_size) {
-               log_err("fio: IO memory too large. Reduce max_bs or iodepth\n");
-               return 1;
-       }
-
-       td->orig_buffer_size = buf_size;
+       td->orig_buffer_size = (unsigned long long) max_bs * (unsigned long long) max_units;
 
        if (td->o.mem_type == MEM_SHMHUGE || td->o.mem_type == MEM_MMAPHUGE)
                td->orig_buffer_size = (td->orig_buffer_size + td->o.hugepage_size - 1) & ~(td->o.hugepage_size - 1);
-       else if (td->orig_buffer_size & page_mask)
-               td->orig_buffer_size = (td->orig_buffer_size + page_mask) & ~page_mask;
+
+       if (td->orig_buffer_size != (size_t) td->orig_buffer_size) {
+               log_err("fio: IO memory too large. Reduce max_bs or iodepth\n");
+               return 1;
+       }
 
        if (allocate_io_mem(td))
                return 1;
@@ -721,6 +716,24 @@ static int switch_ioscheduler(struct thread_data *td)
        return 0;
 }
 
+static int keep_running(struct thread_data *td)
+{
+       unsigned long long io_done;
+
+       if (td->o.time_based)
+               return 1;
+       if (td->o.loops) {
+               td->o.loops--;
+               return 1;
+       }
+
+       io_done = td->io_bytes[DDIR_READ] + td->io_bytes[DDIR_WRITE];
+       if (io_done < td->o.size)
+               return 1;
+
+       return 0;
+}
+
 static int clear_io_state(struct thread_data *td)
 {
        struct fio_file *f;
@@ -736,11 +749,14 @@ static int clear_io_state(struct thread_data *td)
 
        td->last_was_sync = 0;
 
+       td->nr_done_files = 0;
+
        for_each_file(td, f, i)
                td_io_close_file(td, f);
 
        ret = 0;
        for_each_file(td, f, i) {
+               f->flags &= ~FIO_FILE_DONE;
                ret = td_io_open_file(td, f);
                if (ret)
                        break;
@@ -792,7 +808,7 @@ static void *thread_main(void *data)
        if (init_io_u(td))
                goto err;
 
-       if (fio_setaffinity(td) == -1) {
+       if (td->o.cpumask_set && fio_setaffinity(td) == -1) {
                td_verror(td, errno, "cpu_set_affinity");
                goto err;
        }
@@ -835,7 +851,7 @@ static void *thread_main(void *data)
 
        runtime[0] = runtime[1] = 0;
        clear_state = 0;
-       while (td->o.time_based || td->o.loops--) {
+       while (keep_running(td)) {
                fio_gettime(&td->start, NULL);
                memcpy(&td->ts.stat_sample_time, &td->start, sizeof(td->start));