On Thu, 2021-09-30 at 09:02 +0900, Shin'ichiro Kawasaki wrote:
> Test case #46 in t/zbd/test-zbd-support fails when it is repeated
> hundreds of times on null_blk zoned devices. The test case uses libaio
> IO engine to run 8 random write jobs on 4 sequential write required
> zones. When all of the 4 zones get almost full but still open for
> in-flight writes, the helper function zbd_convert_to_open_zone() fails
> to get an opened zone for next write. This results in unexpected job
> termination.
>
> To avoid the unexpected job termination, retry the steps in
> zbd_convert_to_open_zone(). Before retry, call io_u_quiesce() to ensure
> that the in-flight writes get completed.
>
> To prevent infinite loop by the retry, retry only when any IOs are
> in-flight or in-flight IOs get completed. To check in-flight IO count
> of
> all jobs, add a new helper function any_io_in_flight().
>
> Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@xxxxxxx>
Looks good,
Reviewed-by: Dmitry Fomichev <dmitry.fomichev@xxxxxxx>
> ---
> zbd.c | 34 ++++++++++++++++++++++++++++++++++
> 1 file changed, 34 insertions(+)
>
> diff --git a/zbd.c b/zbd.c
> index 64415d2b..c0b0b81c 100644
> --- a/zbd.c
> +++ b/zbd.c
> @@ -1204,6 +1204,19 @@ static uint32_t pick_random_zone_idx(const
> struct fio_file *f,
> f->io_size;
> }
>
> +static bool any_io_in_flight(void)
> +{
> + struct thread_data *td;
> + int i;
> +
> + for_each_td(td, i) {
> + if (td->io_u_in_flight)
> + return true;
> + }
> +
> + return false;
> +}
> +
> /*
> * Modify the offset of an I/O unit that does not refer to an open
> zone such
> * that it refers to an open zone. Close an open zone and open a new
> zone if
> @@ -1223,6 +1236,8 @@ static struct fio_zone_info
> *zbd_convert_to_open_zone(struct thread_data *td,
> uint32_t zone_idx, new_zone_idx;
> int i;
> bool wait_zone_close;
> + bool in_flight;
> + bool should_retry = true;
>
> assert(is_valid_offset(f, io_u->offset));
>
> @@ -1337,6 +1352,7 @@ open_other_zone:
> io_u_quiesce(td);
> }
>
> +retry:
> /* Zone 'z' is full, so try to open a new zone. */
> for (i = f->io_size / zbdi->zone_size; i > 0; i--) {
> zone_idx++;
> @@ -1376,6 +1392,24 @@ open_other_zone:
> goto out;
> pthread_mutex_lock(&zbdi->mutex);
> }
> +
> + /*
> + * When any I/O is in-flight or when all I/Os in-flight get
> completed,
> + * the I/Os might have closed zones then retry the steps to
> open a zone.
> + * Before retry, call io_u_quiesce() to complete in-flight
> writes.
> + */
> + in_flight = any_io_in_flight();
> + if (in_flight || should_retry) {
> + dprint(FD_ZBD, "%s(%s): wait zone close and retry open
> zones\n",
> + __func__, f->file_name);
> + pthread_mutex_unlock(&zbdi->mutex);
> + zone_unlock(z);
> + io_u_quiesce(td);
> + zone_lock(td, f, z);
> + should_retry = in_flight;
> + goto retry;
> + }
> +
> pthread_mutex_unlock(&zbdi->mutex);
> zone_unlock(z);
> dprint(FD_ZBD, "%s(%s): did not open another zone\n", __func__,
