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

[Ira Weiny]

On Fri, Oct 25, 2019 at 11:36:03AM +1100, Dave Chinner wrote:
> On Fri, Oct 25, 2019 at 02:29:04AM +0300, Boaz Harrosh wrote:
> > On 25/10/2019 00:35, Dave Chinner wrote:
> 
> 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.

Why would this be never?

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

I'm thinking this could work with a bit of effort on the users part.  While the
behavior does have a bit of uncertainty, I feel like there has to be a way to
get the inode to drop from the cache when a final iput() happens on the inode.

Admin programs should not leave files open forever, without the users knowing
about it.  So I don't understand why the inode could not be evicted from the
cache if the FS knew that this change had been made and the inode needs to be
"re-loaded".  See below...

> 
> > (And never change the flag on the fly)
> > (Just brain storming here)
> 
> 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....

Agreed.  We can't change the a_ops without some guarantee that no one is using
the file.  Which means we need all fds to close and a final iput().  I thought
that would mean an eviction of the inode and a subsequent reload.

Yesterday I coded up the following (applies on top of this series) but I can't
seem to get it to work because I believe xfs is keeping a reference on the
inode.  What am I missing?  I think if I could get xfs to recognize that the
inode needs to be cleared from it's cache this would work, with some caveats.

Currently this works if I remount the fs or if I use <procfs>/drop_caches like
Boaz mentioned.

Isn't there a way to get xfs to do that on it's own?

Ira