On Fri 20-08-21 18:44:56, Theodore Ts'o wrote:
> On Fri, Aug 20, 2021 at 09:46:56PM +0200, Jan Kara wrote:
> >
> > diff --git a/lib/support/mkquota.c b/lib/support/mkquota.c
> > index fbc3833aee98..34ab632fb81c 100644
> > --- a/lib/support/mkquota.c
> > +++ b/lib/support/mkquota.c
> > @@ -569,14 +569,14 @@ static int scan_dquots_callback(struct dquot *dquot, void *cb_data)
> > */
> > static errcode_t quota_read_all_dquots(struct quota_handle *qh,
> > quota_ctx_t qctx,
> > - int update_limits EXT2FS_ATTR((unused)))
> > + int update_limits)
> > {
> > struct scan_dquots_data scan_data;
> >
> > scan_data.quota_dict = qctx->quota_dict[qh->qh_type];
> > scan_data.check_consistency = 0;
> > - scan_data.update_limits = 0;
> > - scan_data.update_usage = 1;
> > + scan_data.update_limits = update_limits;
> > + scan_data.update_usage = 0;
> >
> > return qh->qh_ops->scan_dquots(qh, scan_dquots_callback, &scan_data);
> > }
>
> This change, while it is correct for tune2fs, is breaking e2fsck's
> f_orphquot test. The root cause is that e2fsck_read_all_quotas() in
> e2fsck/super.c is calling quota_update_limits(), where it had wanted
> to be reading the quota usage data, not the limits.
>
> I think what we need to do is to take quota_read_all_dquots(), which
> is only used by quota_update_limits(), and then fodl it into
> quota_update_limits(). And then add a flags parameter to indicate
> whether we want to be reading the limits or the usage, and then rename
> quota_update_limits() to quota_read_all_dquots().
>
> Does that make sense to you? (One of the reasons why the quota
> functions are all in libsupport is precisely because I knew these
> functions still needed more polishing.)
Yes, makes sense. I'll send patches.
Honza
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR
