fstests.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: "Darrick J. Wong" <darrick.wong@oracle.com>
To: Xiao Yang <yangx.jy@cn.fujitsu.com>
Cc: Ira Weiny <ira.weiny@intel.com>, fstests@vger.kernel.org
Subject: Re: [PATCH v6 1/7] common/rc: Introduce new helpers for DAX mount options and FS_XFLAG_DAX
Date: Wed, 15 Jul 2020 08:56:31 -0700	[thread overview]
Message-ID: <20200715155631.GL7600@magnolia> (raw)
In-Reply-To: <5F0E9A42.5010807@cn.fujitsu.com>

On Wed, Jul 15, 2020 at 01:55:14PM +0800, Xiao Yang wrote:
> On 2020/7/15 12:15, Ira Weiny wrote:
> > On Wed, Jul 15, 2020 at 11:19:58AM +0800, Xiao Yang wrote:
> > > On 2020/7/15 9:59, Ira Weiny wrote:
> > > > On Tue, Jul 14, 2020 at 05:40:03PM +0800, Xiao Yang wrote:
> > > > > 1) _require_scratch_dax_mountopt() checks both old and new DAX mount option
> > > > > 2) _require_dax_iflag() checks FS_XFLAG_DAX
> > > > > 
> > > > > Signed-off-by: Xiao Yang<yangx.jy@cn.fujitsu.com>
> > > > Reviewed-by: Ira Weiny<ira.weiny@intel.com>
> > > Hi Ira,
> > > 
> > > Do you want to check if invalid parameter is passed to
> > > _require_scratch_dax_mountopt? Thought
> > > I think it is not necessary(user should pass correct parament).
> > > Like this:
> > > [ "$mountopt" != "dax" ]&&  [ "$mountopt" != "dax=always" ]&&  _notrun
> > > "$mountopt is invalid"
> > I thought this patch was just checking if the filesystem/device supported DAX
> > such that DAX tests could run using either old _or_ new dax support...
> > But after thinking about it I think we need something more complicated.
> > 
> > Something like:
> > 
> > if ('-o dax=inode')
> > 	// FS is DAX with per file support (ie ext4 || XFS kernel>= 5.8)
> Hi Ira,
> 
> dax=inode/dax=never cannot check if underlying device supports dax, so I
> perfer to use dax=always
> For example:
> XFS:
> # mount -o dax=inode /dev/sda8 /mnt/xfstests/test/
> # mount | grep sda8
> /dev/sda8 on /mnt/xfstests/test type xfs
> (rw,relatime,seclabel,attr2,inode64,logbufs=8,logbsize=32k,noquota)
> ...
> # mount -o dax=always /dev/sda8 /mnt/xfstests/test/
> # mount | grep sda8
> /dev/sda8 on /mnt/xfstests/test type xfs
> (rw,relatime,seclabel,attr2,dax=never,inode64,logbufs=8,logbsize=32k,noquota)
> 
> EXT4:
> # mount -o dax=inode /dev/sda8 /mnt/xfstests/test/
> # mount | grep sda8
> /dev/sda8 on /mnt/xfstests/test type ext4 (rw,relatime,seclabel,dax=inode)
> ...
> # mount -o dax=always /dev/sda8 /mnt/xfstests/test/
> mount: /mnt/xfstests/test: wrong fs type, bad option, bad superblock on
> /dev/sda8, missing codepage or helper program, or other error.
> # dmesg | grep DAX
> [597988.165120] EXT4-fs (sda8): DAX unsupported by block device.
> 
> If we want to test new dax(i.e. dax=inode, dax=always or dax=never), so pass
> "dax=always" to  _require_scratch_dax_mountopt
> If we want to test old dax  or new dax=always(either of them), so pass "dax"
> to _require_scratch_dax_mountopt

That does seem like the only way to figure out if a fs/device
combination supports dax.

The valid arguments to the _require_scratch_dax* helpers are worth
mentioning in a comment above the function, but I suspect that adding
validation code is probably overkill.

--D

> Best Regards,
> Xiao Yang
> > else if ('-o dax')
> > 	// FS is DAX (ie ext4 || XFS on kernel<  5.8)
> > else
> > 	// FS has no dax support.
> > 
> > Do we do that with 2 functions?
> > 
> > 	_require_scratch_dax_mountopt()
> > 	_require_scratch_new_dax_mountopt()
> > 
> > If both fail we are in the else clause above?
> > 
> > We can use the dax=inode because it is a new option and is not necessarily the
> > option used while running the tests.  (I think.  as I said my xfstests skills
> > are not great so I'm not always sure of the correct patterns)
> > 
> > Ira
> > 
> > > Best Regards,
> > > Xiao Yang
> > > > > ---
> > > > > 
> > > > > V5->V6:
> > > > > Simplify _require_scratch_dax_mountopt because it is enough to only check
> > > > > dax/dax=always mount option.  See the following reasons:
> > > > > 1) we cannot detect if underlying device supports dax by mounting dax=inode
> > > > >      or dax=never.
> > > > > 2) dax=always, dax=inode, dax=never are always introduced together.
> > > > > 
> > > > >    common/rc | 22 ++++++++++++++++++++++
> > > > >    1 file changed, 22 insertions(+)
> > > > > 
> > > > > diff --git a/common/rc b/common/rc
> > > > > index f17b19f2..aeec1f11 100644
> > > > > --- a/common/rc
> > > > > +++ b/common/rc
> > > > > @@ -3188,6 +3188,28 @@ _require_scratch_dax()
> > > > >    	_scratch_unmount
> > > > >    }
> > > > > 
> > > > > +# Only accept dax/dax=always mount option becasue dax=always, dax=inode
> > > > > +# and dax=never are always introduced together.
> > > > > +_require_scratch_dax_mountopt()
> > > > > +{
> > > > > +	local mountopt=$1
> > > > > +
> > > > > +	_require_scratch
> > > > > +	_scratch_mkfs>   /dev/null 2>&1
> > > > > +	_try_scratch_mount "-o $mountopt">   /dev/null 2>&1 || \
> > > > > +		_notrun "mount $SCRATCH_DEV with $mountopt failed"
> > > > > +
> > > > > +	_fs_options $SCRATCH_DEV | egrep -q "dax(=always|,|$)" || \
> > > > > +		_notrun "$SCRATCH_DEV $FSTYP does not support -o $mountopt"
> > > > > +
> > > > > +	_scratch_unmount
> > > > > +}
> > > > > +
> > > > > +_require_dax_iflag()
> > > > > +{
> > > > > +	_require_xfs_io_command "chattr" "x"
> > > > > +}
> > > > > +
> > > > >    # Does norecovery support by this fs?
> > > > >    _require_norecovery()
> > > > >    {
> > > > > -- 
> > > > > 2.21.0
> > > > > 
> > > > > 
> > > > > 
> > > > .
> > > > 
> > > 
> > > 
> > 
> > .
> > 
> 
> 
> 

  reply	other threads:[~2020-07-15 16:49 UTC|newest]

Thread overview: 29+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-07-14  9:40 [PATCH v6 0/7] Make fstests support new behavior of DAX Xiao Yang
2020-07-14  9:40 ` [PATCH v6 1/7] common/rc: Introduce new helpers for DAX mount options and FS_XFLAG_DAX Xiao Yang
2020-07-15  1:59   ` Ira Weiny
2020-07-15  3:19     ` Xiao Yang
2020-07-15  4:15       ` Ira Weiny
2020-07-15  5:55         ` Xiao Yang
2020-07-15 15:56           ` Darrick J. Wong [this message]
2020-07-15 18:00             ` Ira Weiny
2020-07-14  9:40 ` [PATCH v6 2/7] fstests: Use _require_scratch_dax_mountopt() and _require_dax_iflag() Xiao Yang
2020-07-15 16:08   ` Darrick J. Wong
2020-07-14  9:40 ` [PATCH v6 3/7] generic/223: Don't clear all mkfs options for _scratch_mkfs_geom() roughly Xiao Yang
2020-07-15  2:31   ` Ira Weiny
2020-07-15  3:12     ` Xiao Yang
2020-07-15 16:07       ` Darrick J. Wong
2020-07-16  1:36         ` Xiao Yang
2020-07-14  9:40 ` [PATCH v6 4/7] generic/413, xfs/260: Improve format operation for PMD fault testing Xiao Yang
2020-07-15 16:09   ` Darrick J. Wong
2020-07-14  9:40 ` [PATCH v6 5/7] xfs/260: Move and update xfs/260 Xiao Yang
2020-07-15 16:10   ` Darrick J. Wong
2020-07-14  9:40 ` [PATCH v6 6/7] generic: Verify if statx() can qurey S_DAX flag on regular file correctly Xiao Yang
2020-07-14  9:40 ` [PATCH v6 7/7] generic: Verify the inheritance behavior of FS_XFLAG_DAX flag in various combinations Xiao Yang
2020-07-15  2:48   ` Ira Weiny
2020-07-15  5:39     ` Xiao Yang
2020-07-15  8:10       ` Xiao Yang
2020-07-15 16:43         ` Xiao Yang
2020-07-15  9:44       ` Xiao Yang
2020-07-15 16:19         ` Darrick J. Wong
2020-07-15 16:33           ` Xiao Yang
2020-07-15 18:18             ` Ira Weiny

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200715155631.GL7600@magnolia \
    --to=darrick.wong@oracle.com \
    --cc=fstests@vger.kernel.org \
    --cc=ira.weiny@intel.com \
    --cc=yangx.jy@cn.fujitsu.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).