Re: [PATCH] iomap: Set all uptodate bits for an Uptodate page

[Matthew Wilcox <willy@infradead.org>]

On Thu, Sep 24, 2020 at 08:27:26PM +0200, Sedat Dilek wrote:
> On Thu, Sep 24, 2020 at 6:36 PM Matthew Wilcox <willy@infradead.org> wrote:
> > On Thu, Sep 24, 2020 at 06:19:03PM +0200, Sedat Dilek wrote:
> > > On Thu, Sep 24, 2020 at 5:27 PM Matthew Wilcox <willy@infradead.org> wrote:
> > > >
> > > > On Thu, Sep 24, 2020 at 05:21:00PM +0200, Sedat Dilek wrote:
> > > > > Great and thanks.
> > > > >
> > > > > Can you send out a seperate patch and label it with "PATCH v5.9"?
> > > > > I run:
> > > > > $ git format-patch -1 --subject-prefix="PATCH v5.9" --signoff
> > > > >
> > > > > Normally, I catch patches from any patchwork URL in mbox format.
> > > >
> > > > Maybe wait a few hours for people to decide if they like the approach
> > > > taken to fix the bug before diving into producing backports?
> > >
> > > That make sense.
> > >
> > > You have a test-case for me?
> > > I have here Linux-Test-Project and FIO available.
> >
> > Qian reported preadv203.c could reproduce it easily on POWER and ARM.
> > They have 64kB pages, so it's easier to hit.  You need to have a
> > filesystem with block size < page size to hit the problem.
> >
> > If you want to check that your test case hits the problem, stick a printk
> > in iomap_page_create().
> 
> I run both linux-kernel on my Debian/unstable AMD64 host (means not in
> a VM) with and without your patch.
> 
> Instructions:
> cd /opt/ltp
> ./runltp -f syscalls -s preadv203
> 
> Unfortunately, the logs in the "results" directory have only the short summary.
> 
> Testcase                                           Result     Exit Value
> --------                                           ------     ----------
> preadv203                                          PASS       0
> preadv203_64                                       PASS       0
> 
> So, I guess I am not hitting the issue?
> Or do I miss some important kernel-config?

Probably.  I don't know how to tell LTP to use a 1kB block size
filesystem.  I tried running that testcase by itself and the logs made
it clear it was using a 4kB block size filesystem.