Re: [PATCH 3/3] xfs_repair: try to fill the AGFL before we fix the freelist

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Tue, Jul 07, 2020 at 08:59:06AM -0400, Brian Foster wrote:
> On Thu, Jul 02, 2020 at 08:27:09AM -0700, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@oracle.com>
> > 
> > In commit 9851fd79bfb1, we added a slight amount of slack to the free
> > space btrees being reconstructed so that the initial fix_freelist call
> > (which is run against a totally empty AGFL) would never have to split
> > either free space btree in order to populate the free list.
> > 
> > The new btree bulk loading code in xfs_repair can re-create this
> > situation because it can set the slack values to zero if the filesystem
> > is very full.  However, these days repair has the infrastructure needed
> > to ensure that overestimations of the btree block counts end up on the
> > AGFL or get freed back into the filesystem at the end of phase 5.
> > 
> > Fix this problem by reserving extra blocks in the bnobt reservation, and
> > checking that there are enough overages in the bnobt/cntbt fakeroots to
> > populate the AGFL with the minimum number of blocks it needs to handle a
> > split in the bno/cnt/rmap btrees.
> > 
> > Note that we reserve blocks for the new bnobt/cntbt/AGFL at the very end
> > of the reservation steps in phase 5, so the extra allocation should not
> > cause repair to fail if it can't find blocks for btrees.
> > 
> > Fixes: 9851fd79bfb1 ("repair: AGFL rebuild fails if btree split required")
> > Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
> > ---
> >  repair/agbtree.c |   51 ++++++++++++++++++++++++++++++++++++++++++++-------
> >  1 file changed, 44 insertions(+), 7 deletions(-)
> > 
> > 
> > diff --git a/repair/agbtree.c b/repair/agbtree.c
> > index de8015ec..9f64d54b 100644
> > --- a/repair/agbtree.c
> > +++ b/repair/agbtree.c
> ...
> > @@ -268,16 +288,33 @@ _("Unable to compute free space by length btree geometry, error %d.\n"), -error)
> >  				 btr_cnt->bload.nr_blocks;
> >  
> >  		/* We don't need any more blocks, so we're done. */
> > -		if (delta_bno >= 0 && delta_cnt >= 0) {
> > +		if (delta_bno >= 0 && delta_cnt >= 0 &&
> > +		    delta_bno + delta_cnt >= agfl_goal) {
> >  			*extra_blocks = delta_bno + delta_cnt;
> >  			break;
> >  		}
> >  
> >  		/* Allocate however many more blocks we need this time. */
> > -		if (delta_bno < 0)
> > +		if (delta_bno < 0) {
> >  			reserve_btblocks(sc->mp, agno, btr_bno, -delta_bno);
> > -		if (delta_cnt < 0)
> > +			delta_bno = 0;
> > +		}
> > +		if (delta_cnt < 0) {
> >  			reserve_btblocks(sc->mp, agno, btr_cnt, -delta_cnt);
> > +			delta_cnt = 0;
> > +		}
> > +
> > +		/*
> > +		 * Try to fill the bnobt cursor with extra blocks to populate
> > +		 * the AGFL.  If we don't get all the blocks we want, stop
> > +		 * trying to fill the AGFL because the AG is totally out of
> > +		 * space.
> > +		 */
> > +		agfl_wanted = agfl_goal - (delta_bno + delta_cnt);
> > +		if (agfl_wanted > 0 &&
> > +		    agfl_wanted != reserve_agblocks(sc->mp, agno, btr_bno,
> > +						    agfl_wanted))
> > +			agfl_goal = 0;
> 
> Nit: can we split off the function call so it's not embedded in the if
> condition? With that tweak:

It occurs to me that we don't care how much we fall short of the
requested allocation.  I could change reserve_agblocks to return true if
it got all the blocks it was asked to get, and then that becomes:

		if (agfl_wanted > 0 &&
		    !reserve_agblocks(sc->mp, agno, btr_bno, agfl_wanted)
			agfl_goal = 0;

How does that sound?

--D

> Reviewed-by: Brian Foster <bfoster@redhat.com>
> 
> >  
> >  		/* Ok, now how many free space records do we have? */
> >  		*nr_extents = count_bno_extents_blocks(agno, &num_freeblocks);
> > 
>