From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-qt1-f193.google.com ([209.85.160.193]:46502 "EHLO mail-qt1-f193.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1726163AbeIZAGu (ORCPT ); Tue, 25 Sep 2018 20:06:50 -0400 Received: by mail-qt1-f193.google.com with SMTP id h22-v6so6386155qtr.13 for ; Tue, 25 Sep 2018 10:58:10 -0700 (PDT) Date: Tue, 25 Sep 2018 14:58:05 -0300 From: Ernesto =?utf-8?Q?A=2E_Fern=C3=A1ndez?= Subject: Re: [PATCH v2] fstests: add support for hfsplus Message-ID: <20180925175805.2sjvygawho4d75ft@eaf> References: <20180924214352.7fjpxbcnxpugnoms@eaf> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Disposition: inline In-Reply-To: Sender: fstests-owner@vger.kernel.org Content-Transfer-Encoding: quoted-printable To: Amir Goldstein Cc: fstests , Eryu Guan , Eric Sandeen , Dave Chinner List-ID: On Tue, Sep 25, 2018 at 08:57:58AM +0300, Amir Goldstein wrote: > On Tue, Sep 25, 2018 at 12:44 AM Ernesto A. Fern=C3=A1ndez > wrote: > > > > Running tests on hfsplus requires patched versions of the mkfs and fs= ck > > tools [1] that support filesystems smaller than the device. >=20 > That's not accurate, is it? Wouldn't it be more accurate to say that th= e patched > tools are required to run the _scratch_mkfs_sized tests? I set this up to call _fatal early if the tools are not the patched versions. Using -X on mkfs.hfsplus forces the size to be a whole number of blocks, and using -X on fsck makes it work under that assumption. This is briefl= y explained in the patched man page for the tools. It's a workaround needed because hfsplus only stores the number of allocation blocks in the volume header. If the device size is not a multiple of the block size, the backup header will be past the end, and fsck will take that as corruption. Of course I could let the patched fsck use the -X flag by default. But that could be dangerous if somebody accidentally uses it to check an important filesystem. >=20 > > > > [1] https://github.com/eafer/hfsprogs-linux.git > > > > Signed-off-by: Ernesto A. Fern=C3=A1ndez > > --- > > common/config | 10 ++++++++++ > > common/rc | 15 ++++++++++++++- > > 2 files changed, 24 insertions(+), 1 deletion(-) > > > > diff --git a/common/config b/common/config > > index 1ba8d96c..315f8b1e 100644 > > --- a/common/config > > +++ b/common/config > > @@ -229,6 +229,7 @@ case "$HOSTOS" in > > export MKFS_CIFS_PROG=3D"false" > > export MKFS_OVERLAY_PROG=3D"false" > > export MKFS_REISER4_PROG=3D$(type -P mkfs.reiser4) > > + export MKFS_HFSPLUS_PROG=3D$(type -P mkfs.hfsplus) > > export E2FSCK_PROG=3D$(type -P e2fsck) > > export TUNE2FS_PROG=3D$(type -P tune2fs) > > export FSCK_OVERLAY_PROG=3D$(type -P fsck.overlay) > > @@ -313,6 +314,9 @@ _mount_opts() > > ubifs) > > export MOUNT_OPTIONS=3D$UBIFS_MOUNT_OPTIONS > > ;; > > + hfsplus) > > + export MOUNT_OPTIONS=3D$HFSPLUS_MOUNT_OPTIONS > > + ;; > > *) > > ;; > > esac > > @@ -380,6 +384,9 @@ _mkfs_opts() > > f2fs) > > export MKFS_OPTIONS=3D"$F2FS_MKFS_OPTIONS" > > ;; > > + hfsplus) > > + export MKFS_OPTIONS=3D$HFSPLUS_MKFS_OPTIONS > > + ;; > > *) > > ;; > > esac > > @@ -397,6 +404,9 @@ _fsck_opts() > > f2fs) > > export FSCK_OPTIONS=3D"" > > ;; > > + hfsplus) > > + export FSCK_OPTIONS=3D"-nX" > > + ;; > > *) > > export FSCK_OPTIONS=3D"-n" > > ;; > > diff --git a/common/rc b/common/rc > > index d5bb1fee..a843e9f6 100644 > > --- a/common/rc > > +++ b/common/rc > > @@ -158,6 +158,12 @@ case "$FSTYP" in > > ubifs) > > [ "$UBIUPDATEVOL_PROG" =3D "" ] && _fatal "ubiupdatevol not f= ound" > > ;; > > + hfsplus) > > + [ "$MKFS_HFSPLUS_PROG" =3D "" ] && _fatal "mkfs.hfsplus not f= ound" > > + mkfs.hfsplus -X |& grep "invalid" && > > + _fatal "A patched version of mkfs.hfsplus is required= :" \ > > + "https://github.com/eafer/hfsprogs-linux.git" >=20 > Why fatal and not just notrun the _scratch_mkfs_sized tests if no -X su= pport? > Must all tests run with mkfs -X and fsck -X?? I always use -X on mkfs so the tests will complain if the size of the device is not a whole number of blocks. Otherwise the tests would later fail on fsck. Maybe I should check for -X support when setting the FSCK_OPTIONS, and in _scratch_mkfs()? Then _fatal would not be necessary. >=20 > > + ;; > > esac > > > > if [ ! -z "$REPORT_LIST" ]; then > > @@ -746,6 +752,10 @@ _scratch_mkfs() > > mkfs_cmd=3D"yes | $MKFS_PROG -t $FSTYP --" > > mkfs_filter=3D"grep -v -e ^mkfs\.ocfs2" > > ;; > > + hfsplus) > > + mkfs_cmd=3D"yes | $MKFS_PROG -t $FSTYP -- -X" >=20 > -X here seems like a typo? This is so the tests will complain if the device size is not a multiple of the block size. I think I explained it in the patched man page for mkfs.hfsplus, but it may be better to put a comment here. Thanks, Ernest >=20 > Thanks, > Amir.