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;
From:   "Fabio M. De Francesco" <fmdefrancesco@gmail.com>
To:     Matthew Wilcox <willy@infradead.org>
Cc:     Evgeniy Dushistov <dushistov@mail.ru>,
        Alexander Viro <viro@zeniv.linux.org.uk>,
        Ira Weiny <ira.weiny@intel.com>, linux-fsdevel@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH v3] fs/ufs: Replace kmap() with kmap_local_page()
Date:   Tue, 02 Aug 2022 09:06:26 +0200
Message-ID: <2589292.k3LOHGUjKi@opensuse>
In-Reply-To: <YoJl+lh0QELbv/TL@casper.infradead.org>
References: <20220516101925.15272-1-fmdefrancesco@gmail.com> <YoJl+lh0QELbv/TL@casper.infradead.org>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="utf-8"
Precedence: bulk

On luned=C3=AC 16 maggio 2022 16:55:54 CEST Matthew Wilcox wrote:
> On Mon, May 16, 2022 at 12:19:25PM +0200, Fabio M. De Francesco wrote:
> > The use of kmap() is being deprecated in favor of kmap_local_page().=20
With
> > kmap_local_page(), the mapping is per thread, CPU local and not=20
globally
> > visible.
> >=20
> > The usage of kmap_local_page() in fs/ufs is pre-thread, therefore=20
replace
> > kmap() / kunmap() calls with kmap_local_page() / kunmap_local().
> >=20
> > kunmap_local() requires the mapping address, so return that address=20
from
> > ufs_get_page() to be used in ufs_put_page().
> >=20
> > These changes are essentially ported from fs/ext2 and are largely based=
=20
on
> > commit 782b76d7abdf ("fs/ext2: Replace kmap() with kmap_local_page()").
> >=20
> > Suggested-by: Ira Weiny <ira.weiny@intel.com>
> > Reviewed-by: Ira Weiny <ira.weiny@intel.com>
> > Signed-off-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>
>=20
> Have you done more than compile-tested this?  I'd like to know that it's
> been tested on a machine with HIGHMEM enabled (in a VM, presumably).
> UFS doesn't get a lot of testing, and it'd be annoying to put out a
> patch that breaks the kmap_local() rules.
>=20
As said in another message of this thread, these changes have only been=20
compile-tested. I can't see anything which may break the rules about using=
=20
local mappings properly.

I'm working on converting all kmap() call sites I can do across the whole=20
kernel to kmap_local_page(). Practically all of those conversions have=20
already been reviewed / acked, and many of them have already been taken by=
=20
their respective maintainers. Others are still too recent.

Most of those patches have been properly tested on a QEMU/KVM x86_32 VM,=20
4GB to 6GB RAM, booting kernels with HIGHMEM64GB enabled.

Instead, despite this submission is very old, I haven't yet been able to=20
figure out how to test these changes. I really don't know how I can create=
=20
and test a UFS filesystem.

Can you please help somewhat with hints about how to test this patch or=20
with testing it yourself? I'm thinking of this option because I suppose=20
that you may have access to a Solaris system (if I recall correctly, UFS is=
=20
the default filesystem of that OS. Isn't it?).

I'm sorry to bother you with this issue, however I'd appreciate any help=20
you may provide. I'd hate to see all patches applied but one :-)=20

Thanks,

=46abio