On Thu, Aug 27, 2020 at 10:42 AM Dave Chinner <david@xxxxxxxxxxxxx> wrote:
>
> On Thu, Aug 27, 2020 at 10:13:15AM +0800, Yafang Shao wrote:
> > On Thu, Aug 27, 2020 at 9:58 AM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
> > >
> > > On Thu, Aug 27, 2020 at 09:34:44AM +0800, Yafang Shao wrote:
> > > > @@ -1500,9 +1500,9 @@ iomap_do_writepage(struct page *page, struct writeback_control *wbc, void *data)
> > > >
> > > > /*
> > > > * Given that we do not allow direct reclaim to call us, we should
> > > > - * never be called in a recursive filesystem reclaim context.
> > > > + * never be called while in a filesystem transaction.
> > > > */
> > > > - if (WARN_ON_ONCE(current->flags & PF_MEMALLOC_NOFS))
> > > > + if (WARN_ON_ONCE(wbc->fstrans_recursion))
> > > > goto redirty;
> > >
> > > Erm, Dave said:
> > >
> > > > I think we should just remove
> > > > the check completely from iomap_writepage() and move it up into
> > > > xfs_vm_writepage() and xfs_vm_writepages().
> > >
> > > ie everywhere you set this new bit, just check current->journal_info.
> >
> >
> > I can't get you. Would you pls. be more specific ?
> >
> > I move the check of current->journal into xfs_vm_writepage() and
> > xfs_vm_writepages(), and I think that is the easiest way to implement
> > it.
> >
> > /* we abort the update if there was an IO error */
> > @@ -564,6 +565,9 @@ xfs_vm_writepage(
> > {
> > struct xfs_writepage_ctx wpc = { };
> >
> > + if (xfs_trans_context_active())
> > + wbc->fstrans_recursion = 1; <<< set for XFS only.
> > +
> > return iomap_writepage(page, wbc, &wpc.ctx, &xfs_writeback_ops);
>
> Get rid of wbc->fstrans_recursion. Just do
>
> if (WARN_ON_ONCE(current->journal_info))
> .....
>
> right here in the XFS code.
>
Understood.
But we have to implement the 'redirty' as well in the XFS code, that
may make the implementation more complicated.
--
Thanks
Yafang
