Re: [PATCH v3 05/26] xfs: Hold inode locks in xfs_rename

[Allison Henderson <allison.henderson@oracle.com>]

On Fri, 2022-09-23 at 12:21 -0700, Darrick J. Wong wrote:
> On Wed, Sep 21, 2022 at 10:44:37PM -0700,
> allison.henderson@oracle.com wrote:
> > From: Allison Henderson <allison.henderson@oracle.com>
> > 
> > Modify xfs_rename to hold all inode locks across a rename operation
> > We will need this later when we add parent pointers
> > 
> > Signed-off-by: Allison Henderson <allison.henderson@oracle.com>
> > ---
> >  fs/xfs/xfs_inode.c | 35 ++++++++++++++++++++++-------------
> >  1 file changed, 22 insertions(+), 13 deletions(-)
> > 
> > diff --git a/fs/xfs/xfs_inode.c b/fs/xfs/xfs_inode.c
> > index 9a3174a8f895..4bfa4a1579f0 100644
> > --- a/fs/xfs/xfs_inode.c
> > +++ b/fs/xfs/xfs_inode.c
> > @@ -2837,18 +2837,16 @@ xfs_rename(
> >         xfs_lock_inodes(inodes, num_inodes, XFS_ILOCK_EXCL);
> >  
> >         /*
> > -        * Join all the inodes to the transaction. From this point
> > on,
> > -        * we can rely on either trans_commit or trans_cancel to
> > unlock
> > -        * them.
> > +        * Join all the inodes to the transaction.
> >          */
> > -       xfs_trans_ijoin(tp, src_dp, XFS_ILOCK_EXCL);
> > +       xfs_trans_ijoin(tp, src_dp, 0);
> >         if (new_parent)
> > -               xfs_trans_ijoin(tp, target_dp, XFS_ILOCK_EXCL);
> > -       xfs_trans_ijoin(tp, src_ip, XFS_ILOCK_EXCL);
> > +               xfs_trans_ijoin(tp, target_dp, 0);
> > +       xfs_trans_ijoin(tp, src_ip, 0);
> >         if (target_ip)
> > -               xfs_trans_ijoin(tp, target_ip, XFS_ILOCK_EXCL);
> > +               xfs_trans_ijoin(tp, target_ip, 0);
> >         if (wip)
> > -               xfs_trans_ijoin(tp, wip, XFS_ILOCK_EXCL);
> > +               xfs_trans_ijoin(tp, wip, 0);
> >  
> >         /*
> >          * If we are using project inheritance, we only allow
> > renames
> > @@ -2862,10 +2860,12 @@ xfs_rename(
> >         }
> >  
> >         /* RENAME_EXCHANGE is unique from here on. */
> > -       if (flags & RENAME_EXCHANGE)
> > -               return xfs_cross_rename(tp, src_dp, src_name,
> > src_ip,
> > +       if (flags & RENAME_EXCHANGE) {
> > +               error = xfs_cross_rename(tp, src_dp, src_name,
> > src_ip,
> >                                         target_dp, target_name,
> > target_ip,
> >                                         spaceres);
> > +               goto out_unlock;
> > +       }
> >  
> >         /*
> >          * Try to reserve quota to handle an expansion of the
> > target directory.
> > @@ -3090,12 +3090,21 @@ xfs_rename(
> >                 xfs_trans_log_inode(tp, target_dp, XFS_ILOG_CORE);
> >  
> >         error = xfs_finish_rename(tp);
> > -       if (wip)
> > -               xfs_irele(wip);
> > -       return error;
> > +
> > +       goto out_unlock;
> >  
> >  out_trans_cancel:
> >         xfs_trans_cancel(tp);
> > +out_unlock:
> > +       /* Unlock inodes in reverse order */
> > +       for (i = num_inodes - 1; i >= 0; i--) {
> > +               if (inodes[i])
> > +                       xfs_iunlock(inodes[i], XFS_ILOCK_EXCL);
> > +
> > +               /* Skip duplicate inodes if src and target dps are
> > the same */
> > +               if (i && (inodes[i] == inodes[i - 1]))
> > +                       i--;
> > +       }
> 
> Could you hoist this to a static inline xfs_iunlock_after_rename
> function that is adjacent to xfs_sort_for_rename, please?  It's
> easier
> to verify that it does the right thing w.r.t. multiple array
> references
> pointing to the same incore inode when the two array management
> functions are right next to each other.
> 
> static inline void
> xfs_iunlock_after_rename(
>         struct xfs_inode        **i_tab,
>         int                     num_inodes)
> {
>         for (i = num_inodes - 1; i >= 0; i--) {
>                 /* Skip duplicate inodes if src and target dps are
> the same */
>                 if (!i_tab[i] || (i > 0 && i_tab[i] == i_tab[i - 1]))
>                         continue;
>                 xfs_iunlock(i_tab[i], XFS_ILOCK_EXCL);
>         }
> }
> 
Sure, that looks fine.  Will do.  Thanks!
Allison

> With that cleaned up,
> Reviewed-by: Darrick J. Wong <djwong@kernel.org>
> 
> --D
> 
> >  out_release_wip:
> >         if (wip)
> >                 xfs_irele(wip);
> > -- 
> > 2.25.1
> >