> -----Original Message----- > From: fio-owner@xxxxxxxxxxxxxxx [mailto:fio-owner@xxxxxxxxxxxxxxx] On > Behalf Of John OCallaghan > Sent: Friday, 15 August, 2014 11:55 AM > To: fio@xxxxxxxxxxxxxxx > Subject: Re: runtime and status-interval > > > > FYI I found the problem in fio/backend.c. Seems to be working fine now. > > > The fix I put together is this: > > if (td_read(td) && td->io_bytes[DDIR_READ]) { > elapsed = utime_since_now(&td->start); > td->ts.runtime[DDIR_READ] += elapsed/1000; > } > if (td_write(td) && td->io_bytes[DDIR_WRITE]) { > elapsed = utime_since_now(&td->start); > td->ts.runtime[DDIR_WRITE] += elapsed/1000; > } > if (td_trim(td) && td->io_bytes[DDIR_TRIM]) { > elapsed = utime_since_now(&td->start); > td->ts.runtime[DDIR_TRIM] += elapsed/1000; > } > > The problem is is that runtime is holding milliseconds but > utime_since_now() returns microseconds. I added the "/1000" to make the > conversion. > > My change is a simple fix but quite possibly there are other places > within fio making the same mistake? These lines could just call mtime_since_now rather than utime_since_now. There's one more a few lines down: fio_gettime(&td->start, NULL); do_verify(td, verify_bytes); td->ts.runtime[DDIR_READ] += utime_since_now(&td->start); --- Rob Elliott HP Server Storage ��.n��������+%������w��{.n�������^n�r������&��z�ޗ�zf���h���~����������_��+v���)ߣ�
