From:   "Fabio M. De Francesco" <fmdefrancesco@gmail.com>
To:     Jan Kara <jack@suse.cz>, Al Viro <viro@zeniv.linux.org.uk>
Cc:     Jan Kara <jack@suse.com>,
        Linus Torvalds <torvalds@linux-foundation.org>,
        linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org
Subject: Re: [git pull] vfs.git sysv pile
Date:   Wed, 15 Mar 2023 19:08:57 +0100
Message-ID: <3019063.4lk9UinFSI@suse>
In-Reply-To: <Y/9duET0Mt5hPu2L@ZenIV>
References: <Y/gugbqq858QXJBY@ZenIV> <20230301130018.yqds5yvqj7q26f7e@quack3>
 <Y/9duET0Mt5hPu2L@ZenIV>
MIME-Version: 1.0
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain; charset="iso-8859-1"
Precedence: bulk

On mercoled=EC 1 marzo 2023 15:14:16 CET Al Viro wrote:
> On Wed, Mar 01, 2023 at 02:00:18PM +0100, Jan Kara wrote:
> > On Wed 01-03-23 12:20:56, Fabio M. De Francesco wrote:
> > > On venerd=EC 24 febbraio 2023 04:26:57 CET Al Viro wrote:
> > > > 	Fabio's "switch to kmap_local_page()" patchset (originally after t=
he
> > > >=20
> > > > ext2 counterpart, with a lot of cleaning up done to it; as the matt=
er=20
of
> > > > fact, ext2 side is in need of similar cleanups - calling conventions
> > > > there
> > > > are bloody awful).
> > >=20

[snip]

>=20
> I think I've pushed a demo patchset to vfs.git at some point back in
> January... Yep - see #work.ext2 in there; completely untested, though.

The following commits from the VFS tree, #work.ext2 look good to me.

f5b399373756 ("ext2: use offset_in_page() instead of open-coding it as=20
subtraction")
c7248e221fb5 ("ext2_get_page(): saner type")
470e54a09898 ("ext2_put_page(): accept any pointer within the page")
15abcc147cf7 ("ext2_{set_link,delete_entry}(): don't bother with page_addr")
16a5ee2027b7 ("ext2_find_entry()/ext2_dotdot(): callers don't need page_add=
r=20
anymore")

Reviewed-by: Fabio M. De Francesco <fmdefrancesco@gmail.com>

I could only read the code but I could not test it in the same QEMU/KVM x86=
_32=20
VM where I test all my HIGHMEM related work.=20

Btrfs as well as all the other filesystems I converted to kmap_local_page()=
=20
don't make the processes in the VM to crash, whereas the xfstests on ext2 =
=20
trigger the OOM killer at random tests (only sometimes they exit gracefully=
).

=46YI, I tried to run the tests with 6GB of RAM, booting a kernel with=20
HIGHMEM64GB enabled. I cannot add my "Tested-by" tag.

=46abio