Re: [PATCH 6/9] xfs_repair: check inode btree block counters in AGI

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Wed, Oct 28, 2020 at 01:29:59PM -0400, Brian Foster wrote:
> On Mon, Oct 26, 2020 at 04:33:44PM -0700, Darrick J. Wong wrote:
> > From: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> > 
> > Make sure that both inode btree block counters in the AGI are correct.
> > 
> > Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx>
> > ---
> >  repair/scan.c |   38 +++++++++++++++++++++++++++++++++++---
> >  1 file changed, 35 insertions(+), 3 deletions(-)
> > 
> > 
> > diff --git a/repair/scan.c b/repair/scan.c
> > index 2a38ae5197c6..c826af7dee86 100644
> > --- a/repair/scan.c
> > +++ b/repair/scan.c
> ...
> > @@ -2022,6 +2029,17 @@ scan_inobt(
> >  			return;
> >  	}
> >  
> > +	switch (magic) {
> > +	case XFS_FIBT_MAGIC:
> > +	case XFS_FIBT_CRC_MAGIC:
> > +		ipriv->fino_blocks++;
> > +		break;
> > +	case XFS_IBT_MAGIC:
> > +	case XFS_IBT_CRC_MAGIC:
> > +		ipriv->ino_blocks++;
> > +		break;
> > +	}
> > +
> 
> Is this intentionally not folded into the earlier magic switch
> statement?

I'd originally thought that we wouldn't want to count the block if it's
obviously bad, but the tree and the agi counter are supposed to be
consistent with each other, so yes, we should always bump the counter
when we are pointed towards a block.

> >  	/*
> >  	 * check for btree blocks multiply claimed, any unknown/free state
> >  	 * is ok in the bitmap block.
> ...
> > @@ -2393,6 +2414,17 @@ validate_agi(
> >  		}
> >  	}
> >  
> > +	if (xfs_sb_version_hasinobtcounts(&mp->m_sb)) {
> > +		if (be32_to_cpu(agi->agi_iblocks) != priv.ino_blocks)
> > +			do_warn(_("bad inobt block count %u, saw %u\n"),
> > +					priv.ino_blocks,
> > +					be32_to_cpu(agi->agi_iblocks));
> 
> These two params are backwards (here and below), no?

Oops.  Good catch!

--D

> Brian
> 
> > +		if (be32_to_cpu(agi->agi_fblocks) != priv.fino_blocks)
> > +			do_warn(_("bad finobt block count %u, saw %u\n"),
> > +					priv.fino_blocks,
> > +					be32_to_cpu(agi->agi_fblocks));
> > +	}
> > +
> >  	if (be32_to_cpu(agi->agi_count) != agcnts->agicount) {
> >  		do_warn(_("agi_count %u, counted %u in ag %u\n"),
> >  			 be32_to_cpu(agi->agi_count), agcnts->agicount, agno);
> > 
> 



[Index of Archives]     [XFS Filesystem Development (older mail)]     [Linux Filesystem Development]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux RAID]     [Linux SCSI]


  Powered by Linux