Re: [PATCH 0/5] Enable per-file/directory DAX operations

From: Boaz Harrosh <boaz@plexistor.com>
To: Dave Chinner <david@fromorbit.com>, Boaz Harrosh <boaz@plexistor.com>
Cc: ira.weiny@intel.com, linux-kernel@vger.kernel.org,
	Alexander Viro <viro@zeniv.linux.org.uk>,
	"Darrick J. Wong" <darrick.wong@oracle.com>,
	Dan Williams <dan.j.williams@intel.com>,
	Christoph Hellwig <hch@lst.de>,
	"Theodore Y. Ts'o" <tytso@mit.edu>, Jan Kara <jack@suse.cz>,
	linux-ext4@vger.kernel.org, linux-xfs@vger.kernel.org,
	linux-fsdevel@vger.kernel.org
Subject: Re: [PATCH 0/5] Enable per-file/directory DAX operations
Date: Fri, 25 Oct 2019 04:15:44 +0300	[thread overview]
Message-ID: <9ffbc2a5-c85b-3633-1ad5-a9a3fe33cd2e@plexistor.com> (raw)
In-Reply-To: <20191025003603.GE4614@dread.disaster.area>

On 25/10/2019 03:36, Dave Chinner wrote:
> On Fri, Oct 25, 2019 at 02:29:04AM +0300, Boaz Harrosh wrote:
<>

>> Perhaps we always go by the directory. And then do an mv dir_DAX/foo dir_NODAX/foo
> 
> The inode is instatiated before the rename is run, so it's set up
> with it's old dir config, not the new one. So this ends up with the
> same problem of haivng to change the S_DAX flag and aops vector
> dynamically on rename. Same problem, not a solution.
> 

Yes Admin needs a inode-drop_caches after the mv if she/he wants an effective
change.

>> to have an effective change. In hard links the first one at iget time before populating
>> the inode cache takes affect.
> 
> If something like a find or backup program brings the inode into
> cache, the app may not even get the behaviour it wants, and it can't
> change it until the inode is evicted from cache, which may be never.

inode-drop-caches. (echo 2 > /proc/sys/vm/drop_caches)

> Nobody wants implicit/random/uncontrollable/unchangeable behaviour
> like this.
> 

You mean in the case of hard links between different mode directories?
I agree it is not so good. I do not like it too.

<>
> We went over all this ground when we disabled the flag in the first
> place. We disabled the flag because we couldn't come up with a sane
> way to flip the ops vector short of tracking the number of aops
> calls in progress at any given time. i.e. reference counting the
> aops structure, but that's hard to do with a const ops structure,
> and so it got disabled rather than allowing users to crash
> kernels....
> 

Do you mean dropping this patchset all together? I missed that.

Current patchset with the i_size == 0 thing is really bad I think.
Its the same has dropping the direct change all together and only
supporting inheritance from parent.
[Which again for me is really not interesting]

> Cheers,
> -Dave.

Lets sleep on it. Please remind me if xfs supports clone + DAX

Thanks Dave
Boaz