Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <1665725448-31439-1-git-send-email-zhaoyang.huang@unisoc.com>
 <Y0lSChlclGPkwTeA@casper.infradead.org> <CAGWkznG=_A-3A8JCJEoWXVcx+LUNH=gvXjLpZZs0cRX4dhUJfQ@mail.gmail.com>
 <Y017BeC64GDb3Kg7@casper.infradead.org> <CAGWkznEdtGPPZkHrq6Y_+XLL37w12aC8XN8R_Q-vhq48rFhkSA@mail.gmail.com>
 <Y04Y3RNq6D2T9rVw@casper.infradead.org> <20221018223042.GJ2703033@dread.disaster.area>
 <20221019011636.GM2703033@dread.disaster.area> <20221019044734.GN2703033@dread.disaster.area>
 <CAGWkznEGMg293S7jOmZ7G-UhEBg6rQZhTd6ffhjoDgoFGvhFNw@mail.gmail.com> <Y0/2T5KpFurV2MLp@casper.infradead.org>
In-Reply-To: <Y0/2T5KpFurV2MLp@casper.infradead.org>
From:   Zhaoyang Huang <huangzhaoyang@gmail.com>
Date:   Thu, 20 Oct 2022 09:27:39 +0800
Message-ID: <CAGWkznF3y0BFi-9+1GhBv8NS5P7TZKvdbX0qSbaGVR2BmvD3Ew@mail.gmail.com>
Subject: Re: [RFC PATCH] mm: move xa forward when run across zombie page
To:     Matthew Wilcox <willy@infradead.org>
Cc:     Dave Chinner <david@fromorbit.com>,
        "zhaoyang.huang" <zhaoyang.huang@unisoc.com>,
        Andrew Morton <akpm@linux-foundation.org>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, ke.wang@unisoc.com,
        steve.kang@unisoc.com, baocong.liu@unisoc.com,
        linux-fsdevel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Wed, Oct 19, 2022 at 9:06 PM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Wed, Oct 19, 2022 at 01:48:37PM +0800, Zhaoyang Huang wrote:
> > On Wed, Oct 19, 2022 at 12:47 PM Dave Chinner <david@fromorbit.com> wrote:
> > > I removed the mapping_set_large_folios() calls in the XFS inode
> > > instantiation and the test code has now run over 55,000 iterations
> > > without failing.  The most iterations I'd seen with large folios
> > > enabled was about 7,000 - typically it would fail within 2-3,000
> > > iterations.
> > hint from my side. The original problem I raised is under v5.15 where
> > there is no folio yet.
>
> But 5.15 does use 2MB pages in shmem.  You haven't really provided
> any information, so I don't know whether the inode that you're having
> problems with is a shmem inode.
No, it is a f2fs inode using 4KB pages.