Re: [PATCH] ubifs: Fix wrong orphan node deletion in ubifs_jnl_update()

[Zhihao Cheng]

å 2020/7/7 19:52, Richard Weinberger åé:
> On Thu, Jul 2, 2020 at 5:21 PM Zhihao Cheng <chengzhihao1@xxxxxxxxxx> wrote:
> > There a wrong orphan node deleting in error handling path in
> > ubifs_jnl_update(), which may cause following error msg:
> >
> >    UBIFS error (ubi0:0 pid 1522): ubifs_delete_orphan [ubifs]:
> >    missing orphan ino 65
> >
> > Fix this by checking whether the node has been operated for
> > adding to orphan list before being deleted,
> >
> > Signed-off-by: Zhihao Cheng <chengzhihao1@xxxxxxxxxx>
> > ---
> >   fs/ubifs/journal.c | 5 +++--
> >   1 file changed, 3 insertions(+), 2 deletions(-)
> >
> > diff --git a/fs/ubifs/journal.c b/fs/ubifs/journal.c
> > index e5ec1afe1c66..db0a80dd9d52 100644
> > --- a/fs/ubifs/journal.c
> > +++ b/fs/ubifs/journal.c
> > @@ -539,7 +539,7 @@ int ubifs_jnl_update(struct ubifs_info *c, const struct inode *dir,
> >                       const struct fscrypt_name *nm, const struct inode *inode,
> >                       int deletion, int xent)
> >   {
> > -       int err, dlen, ilen, len, lnum, ino_offs, dent_offs;
> > +       int err, dlen, ilen, len, lnum, ino_offs, dent_offs, orphan_added = 0;
> >          int aligned_dlen, aligned_ilen, sync = IS_DIRSYNC(dir);
> >          int last_reference = !!(deletion && inode->i_nlink == 0);
> >          struct ubifs_inode *ui = ubifs_inode(inode);
> > @@ -630,6 +630,7 @@ int ubifs_jnl_update(struct ubifs_info *c, const struct inode *dir,
> >                          goto out_finish;
> >                  }
> >                  ui->del_cmtno = c->cmt_no;
> > +               orphan_added = 1;
> >          }
> >
> >          err = write_head(c, BASEHD, dent, len, &lnum, &dent_offs, sync);
> > @@ -702,7 +703,7 @@ int ubifs_jnl_update(struct ubifs_info *c, const struct inode *dir,
> >          kfree(dent);
> >   out_ro:
> >          ubifs_ro_mode(c, err);
> > -       if (last_reference)
> > +       if (last_reference && orphan_added)
> I think you can just check for orphan_added here.
> Looks good otherwise, thanks for fixing! :-)
Sounds reasonable. I will send v2.