On Wed, Oct 09, 2019 at 04:46:36PM -0500, Eric Sandeen wrote:
> On 9/25/19 4:34 PM, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> >
> > Fix all the places in the per-thread counter functions either we fail to
> > check for runtime errors or fail to communicate them properly to
> > callers. Then fix all the callers to report the error messages instead
> > of hiding them.
> >
> > Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> > ---
> > scrub/counter.c | 33 ++++++++++++++++++---------------
> > scrub/counter.h | 6 +++---
> > scrub/phase3.c | 23 +++++++++++++++++------
> > scrub/progress.c | 12 +++++++++---
> > scrub/read_verify.c | 9 ++++++---
> > 5 files changed, 53 insertions(+), 30 deletions(-)
>
> ...
>
> > @@ -282,5 +282,8 @@ uint64_t
> > read_verify_bytes(
> > struct read_verify_pool *rvp)
> > {
> > - return ptcounter_value(rvp->verified_bytes);
> > + uint64_t ret;
> > +
> > + ptcounter_value(rvp->verified_bytes, &ret);
> > + return ret;
> > }
>
> IMHO this is a confusing use of "ret" which is normally return status but
> here it is the sum? And errors are ignored? We just get a ret ("sum") of zero?
For now, yes. The patch "xfs_scrub: fix read-verify pool error
communication problems" in the next series will fix a bunch of error
handling problems in the read_verify.c functions.
(I'm trying only to change the direct ptvar.c callers in this patch...)
--D
