On Wed, Feb 01, 2012 at 09:56:01AM +1100, Dave Chinner wrote:
> On Sat, Jan 28, 2012 at 01:55:01PM +0300, Dan Carpenter wrote:
> > Smatch complains because we check "commit_lsn" for NULL inconsistently.
> > fs/xfs/xfs_log_cil.c +705 xfs_log_commit_cil(43) error: we previously
> > assumed 'commit_lsn' could be null (see line 688)
> > xfs_log_commit_cil() is only called from one place, and "commit_lsn" is
> > a valid pointer, so I've removed the NULL check.
> >
> > Signed-off-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>
> >
> > diff --git a/fs/xfs/xfs_log_cil.c b/fs/xfs/xfs_log_cil.c
> > index d4fadbe..d9c4652 100644
> > --- a/fs/xfs/xfs_log_cil.c
> > +++ b/fs/xfs/xfs_log_cil.c
> > @@ -685,8 +685,7 @@ xfs_log_commit_cil(
> >
> > /* lock out background commit */
> > down_read(&log->l_cilp->xc_ctx_lock);
> > - if (commit_lsn)
> > - *commit_lsn = log->l_cilp->xc_ctx->sequence;
> > + *commit_lsn = log->l_cilp->xc_ctx->sequence;
> >
> > xlog_cil_insert_items(log, log_vector, tp->t_ticket);
>
> There's a set of reviewed patches (for 3.3) that change all this
> code. The null check might still be there, but that needs to be
> checked.
>
> On that note, Ben, can you get all of the reviewed patches that are
> currently outstanding into the for-next branch of the oss tree? We
> need this done sooner rather than later so that all our current
> testing during development is done with those patches applied. it
> also helps us to avoid conflicts between patchsets that touch the
> same code (which I have quite a few of at the moment)....
I'll have it resolved ASAP.
-Ben
|