From mboxrd@z Thu Jan  1 00:00:00 1970
From: Hin-Tak Leung <hintak_leung@yahoo.co.uk>
Subject: Re: Kernel 3.1.0-rc4 oops when connecting iPod
Date: Mon, 12 Sep 2011 16:57:30 +0100 (BST)
Message-ID: <1315843050.34041.YahooMailClassic@web29505.mail.ird.yahoo.com>
References: <CAG1a4rsuYVehnSCM3KxeYQ+g94nrivG1vfss=y8275E4D24qTg@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: QUOTED-PRINTABLE
Cc: linux-fsdevel@vger.kernel.org,
	linux-kernel <linux-kernel@vger.kernel.org>,
	Christoph Hellwig <hch@tuxera.com>
To: Christoph Hellwig <hch@infradead.org>,
	Pavel Ivanov <paivanof@gmail.com>
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <CAG1a4rsuYVehnSCM3KxeYQ+g94nrivG1vfss=y8275E4D24qTg@mail.gmail.com>
Sender: linux-kernel-owner@vger.kernel.org
List-Id: linux-fsdevel.vger.kernel.org

--- On Mon, 12/9/11, Pavel Ivanov <paivanof@gmail.com> wrote:

> On Mon, Sep 12, 2011 at 10:34 AM,
> Christoph Hellwig <hch@infradead.org>
> wrote:
> > Does this patch fix your issues with large block
> sizes?
>=20
> I'll be able to try it in the evening but meanwhile I have
> some comments below.
>=20
> >
> >
> > Index: linux-2.6/fs/hfsplus/super.c
> >
> =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=
=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D
> > --- linux-2.6.orig/fs/hfsplus/super.c =A0 2011-09-12
> 09:56:58.619988416 -0400
> > +++ linux-2.6/fs/hfsplus/super.c =A0 =A0 =A0 =A02011-09-12
> 10:07:18.006651395 -0400
> > @@ -344,6 +344,7 @@ static int
> hfsplus_fill_super(struct sup
> > =A0 =A0 =A0 =A0struct inode *root, *inode;
> > =A0 =A0 =A0 =A0struct qstr str;
> > =A0 =A0 =A0 =A0struct nls_table *nls =3D NULL;
> > + =A0 =A0 =A0 u64 last_fs_block, last_fs_page;
> > =A0 =A0 =A0 =A0int err;
> >
> > =A0 =A0 =A0 =A0err =3D -EINVAL;
> > @@ -399,9 +400,13 @@ static int
> hfsplus_fill_super(struct sup
> > =A0 =A0 =A0 =A0if (!sbi->rsrc_clump_blocks)
> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0sbi->rsrc_clump_blocks =3D 1;
> >
> > - =A0 =A0 =A0 err =3D
> generic_check_addressable(sbi->alloc_blksz_shift,
> > - =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0
> =A0 =A0 sbi->total_blocks);
> > - =A0 =A0 =A0 if (err) {
> > + =A0 =A0 =A0 err =3D -EFBIG;
> > + =A0 =A0 =A0 last_fs_block =3D sbi->total_blocks - 1;
> > + =A0 =A0 =A0 last_fs_page =3D (last_fs_block >>
> sbi->alloc_blksz_shift) <<
> > + =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0 PAGE_CACHE_SHIFT;
>=20
> Did you mix left and right shifts here? Expression doesn't
> make sense to me.
>=20
> Also I have a little concern about consistency in using
> PAGE_CACHE_SHIFT and PAGE_SHIFT. hfsplus_read_wrapper()
> limits visible
> block size to PAGE_SIZE, not PAGE_CACHE_SIZE. And although
> now they
> are equal comment in linux/pagemap.h clearly says that
> PAGE_CACHE_SIZE
> can be bigger than PAGE_SIZE. Is it something that should
> be fixed in
> hfsplus_read_wrapper() ?
>=20
> > +
> > + =A0 =A0 =A0 if ((last_fs_block > (sector_t)(~0ULL)
> >> (sbi->alloc_blksz_shift - 9)) ||
>=20
> Maybe this 9 should be extracted from here and
> generic_check_addressable() into some macro?
>=20
> > + =A0 =A0 =A0 =A0 =A0 (last_fs_page > (pgoff_t)(~0ULL)))
> {
> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0printk(KERN_ERR "hfs:
> filesystem size too large.\n");
> > =A0 =A0 =A0 =A0 =A0 =A0 =A0 =A0goto out_free_vhdr;
> > =A0 =A0 =A0 =A0}
> >

I 2nd that this is kind of ugly. The literal "9". How about abstracting=
 this logic out to say, hfs_check_addressable()?