On Sun, Oct 09, 2011 at 07:45:29PM -0400, Christoph Hellwig wrote:
> On Mon, Oct 10, 2011 at 10:11:47AM +1100, Dave Chinner wrote:
> > From: Dave Chinner <dchinner@xxxxxxxxxx>
> >
> > Fix a potential prefetch read problem due to the first loop
> > execution of pf_batch_read potentially not initialising the fsbno
> > varaible:
>
> variable.
>
> >
> > ==10177== Thread 6:
> > ==10177== Conditional jump or move depends on uninitialised value(s)
> > ==10177== at 0x8079CAB: pf_batch_read (prefetch.c:408)
> > ==10177== by 0x6A2996D: clone (clone.S:130)
> > ==10177==
> >
> > Signed-off-by: Dave Chinner <dchinner@xxxxxxxxxx>
> > ---
> > repair/prefetch.c | 2 +-
> > 1 files changed, 1 insertions(+), 1 deletions(-)
> >
> > diff --git a/repair/prefetch.c b/repair/prefetch.c
> > index d2fdf90..da074a8 100644
> > --- a/repair/prefetch.c
> > +++ b/repair/prefetch.c
> > @@ -397,7 +397,7 @@ pf_batch_read(
> > int len, size;
> > int i;
> > int inode_bufs;
> > - unsigned long fsbno;
> > + unsigned long fsbno = 0;
>
> Call me stupid, but I can't see how this could actually be a real
> life issue. The first thing we do in the loop is to to write
> to fsbno in btree_find. I'm fine adding this to shut up warnins,
> but I can't see a real issue.
If btree_find() fails to find the key being looked up, it returns
without having initialised fsbno.
Cheers,
Dave.
--
Dave Chinner
david@xxxxxxxxxxxxx
|