On Tue, Jul 14, 2020 at 02:46:43PM -0700, Darrick J. Wong wrote: > From: Darrick J. Wong <darrick.wong@xxxxxxxxxx> > > The onstack inode in xfs_check's process_inode is a potential landmine > since it's not a /real/ incore inode. The upcoming 5.8 merge will make > this messier wrt inode forks, so just remove the onstack inode and > reference the ondisk fields directly. This also reduces the amount of > thinking that I have to do w.r.t. future libxfs porting efforts. > > Signed-off-by: Darrick J. Wong <darrick.wong@xxxxxxxxxx> Comparing this to my version here: http://git.infradead.org/users/hch/xfsprogs.git/commitdiff/791d7d324290dbb83be7bf35fe15f9898f5df1c1 > mode_t mode; > + uint16_t diflags; > + uint64_t diflags2 = 0; > + xfs_nlink_t nlink; > + xfs_dqid_t uid; > + xfs_dqid_t gid; > + xfs_dqid_t prid; Not sure we really need the local variables, as they are mostly just used once except for error messages.. > + if (dip->di_version == 1) { > + nlink = be16_to_cpu(dip->di_onlink); > + prid = 0; > + } else { > + nlink = be32_to_cpu(dip->di_nlink); > + prid = (xfs_dqid_t)be16_to_cpu(dip->di_projid_hi) << 16 | > + be16_to_cpu(dip->di_projid_lo); > + } I mad the assumption that we don't support v1 inodes anymore, but it appears we actually do. So we might need to keep these two. > if (isfree) { > - if (xino.i_d.di_nblocks != 0) { > + if (be64_to_cpu(dip->di_nblocks) != 0) { No need to byte swap for a comparism with 0. > - if ((unsigned int)xino.i_d.di_aformat > XFS_DINODE_FMT_BTREE) { > + if ((unsigned int)dip->di_aformat > XFS_DINODE_FMT_BTREE) { No need for the (pre-existing) cast here. > + fmtnames[(int)dip->di_aformat], Same here.
