Re: [PATCH v5 5/8] virtio: Treat alloc_dax() -EOPNOTSUPP failure as non-fatal

From: Greg Kroah-Hartman
Date: Wed Feb 14 2024 - 02:00:02 EST


On Tue, Feb 13, 2024 at 02:46:05PM -0500, Mathieu Desnoyers wrote:
> On 2024-02-13 01:25, Lukas Wunner wrote:
> > On Mon, Feb 12, 2024 at 11:30:58AM -0500, Mathieu Desnoyers wrote:
> > > In preparation for checking whether the architecture has data cache
> > > aliasing within alloc_dax(), modify the error handling of virtio
> > > virtio_fs_setup_dax() to treat alloc_dax() -EOPNOTSUPP failure as
> > > non-fatal.
> > >
> > > Co-developed-by: Dan Williams <dan.j.williams@xxxxxxxxx>
> > > Signed-off-by: Dan Williams <dan.j.williams@xxxxxxxxx>
> > > Fixes: d92576f1167c ("dax: does not work correctly with virtual aliasing caches")
> >
> > That's a v4.0 commit, yet this patch uses DEFINE_FREE() which is
> > only available in v6.6 but not any earlier stable kernels.
>
> I asked this question to Greg KH before creating this patch, and his
> answer was to implement my fix for master, and stable kernels would take
> care of backporting all the required dependencies.

That is correct.

> Now if I look at latest 6.1, 5.15, 5.10, 5.4, 4.19 stable kernels,
> none seem to have include/linux/cleanup.h today. But I suspect that
> sooner or later relevant master branch fixes will require stable
> kernels to backport cleanup.h, so why not do it now ?

Yes, eventually we will need to backport cleanup.h to the older kernel
trees, I know of many patches "in flight" that are using it, so it's not
unique to this one at all, so this is fine to have.

Remember, make changes for Linus's tree, don't go through any gyrations
to make things special for stable releases, that's something to only
consider later on, if you want to. stable kernels should have no affect
on developers OTHER than a simple cc: stable tag on stuff that they know
should be backported, unless they really want to do more than that,
that's up to them.

thanks,

greg k-h