Re: [RFC PATCH] mm: move xa forward when run across zombie page

[Matthew Wilcox]

On Tue, Oct 18, 2022 at 10:52:19AM +0800, Zhaoyang Huang wrote:
> On Mon, Oct 17, 2022 at 11:55 PM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
> >
> > On Mon, Oct 17, 2022 at 01:34:13PM +0800, Zhaoyang Huang wrote:
> > > On Fri, Oct 14, 2022 at 8:12 PM Matthew Wilcox <willy@xxxxxxxxxxxxx> wrote:
> > > >
> > > > On Fri, Oct 14, 2022 at 01:30:48PM +0800, zhaoyang.huang wrote:
> > > > > From: Zhaoyang Huang <zhaoyang.huang@xxxxxxxxxx>
> > > > >
> > > > > Bellowing RCU stall is reported where kswapd traps in a live lock when shrink
> > > > > superblock's inode list. The direct reason is zombie page keeps staying on the
> > > > > xarray's slot and make the check and retry loop permanently. The root cause is unknown yet
> > > > > and supposed could be an xa update without synchronize_rcu etc. I would like to
> > > > > suggest skip this page to break the live lock as a workaround.
> > > >
> > > > No, the underlying bug should be fixed.
> >
> >     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> Understand. IMHO, find_get_entry actruely works as an open API dealing
> with different kinds of address_spaces page cache, which requires high
> robustness to deal with any corner cases. Take the current problem as
> example, the inode with fault page(refcount=0) could remain on the
> sb's list without live lock problem.

But it's a corner case that shouldn't happen!  What else is going on
at the time?  Can you reproduce this problem easily?  If so, how?