[...] > > Vincent gets these too, but pasting it here for Shin'ichiro: > > Please find the latest report on new defect(s) introduced to axboe/fio found with Coverity Scan. > > 1 new defect(s) introduced to axboe/fio found with Coverity Scan. > 1 defect(s), reported by Coverity Scan earlier, were marked fixed in the recent build analyzed by Coverity Scan. > > New defect(s) Reported-by: Coverity Scan > Showing 1 of 1 defect(s) > > > ** CID 462265: (ATOMICITY) > /zbd.c: 1554 in zbd_convert_to_write_zone() > /zbd.c: 1554 in zbd_convert_to_write_zone() > > > ________________________________________________________________________________________________________ > *** CID 462265: (ATOMICITY) > /zbd.c: 1554 in zbd_convert_to_write_zone() > 1548 "%s(%s): wait zone write and retry write target zone selection\n", > 1549 __func__, f->file_name); > 1550 pthread_mutex_unlock(&zbdi->mutex); > 1551 zone_unlock(z); > 1552 io_u_quiesce(td); > 1553 zone_lock(td, f, z); > >>> CID 462265: (ATOMICITY) > >>> Using an unreliable value of "in_flight" inside the second locked section. If the data that "in_flight" depends on was changed by another thread, this use might be incorrect. > 1554 should_retry = in_flight; > 1555 goto retry; > 1556 } > 1557 > 1558 pthread_mutex_unlock(&zbdi->mutex); > 1559 > /zbd.c: 1554 in zbd_convert_to_write_zone() > 1548 "%s(%s): wait zone write and retry write target zone selection\n", > 1549 __func__, f->file_name); > 1550 pthread_mutex_unlock(&zbdi->mutex); > 1551 zone_unlock(z); > 1552 io_u_quiesce(td); > 1553 zone_lock(td, f, z); > >>> CID 462265: (ATOMICITY) > >>> Using an unreliable value of "in_flight" inside the second locked section. If the data that "in_flight" depends on was changed by another thread, this use might be incorrect. > 1554 should_retry = in_flight; > 1555 goto retry; > 1556 } > 1557 > 1558 pthread_mutex_unlock(&zbdi->mutex); > 1559 Thanks for the notice. I think the defect report is false-positive. The variable in_flight is on the stack and thread local, so no other thread can change it across the zone_unlock() and zone_lock(). I guess and hope the following patch will suppress the report. May I ask Vincent or Jens to confirm it? If its formal patch helps, please let me know. diff --git a/zbd.c b/zbd.c index 9455140a..7fcf1ec4 100644 --- a/zbd.c +++ b/zbd.c @@ -1547,11 +1547,11 @@ retry: dprint(FD_ZBD, "%s(%s): wait zone write and retry write target zone selection\n", __func__, f->file_name); + should_retry = in_flight; pthread_mutex_unlock(&zbdi->mutex); zone_unlock(z); io_u_quiesce(td); zone_lock(td, f, z); - should_retry = in_flight; goto retry; }
