Re: [PATCH v4 0/3] driver core: add TAINT_FORCED_BIND for when userspace manually messes with devices and drivers

[Jason Gunthorpe <jgg@ziepe.ca>]

On Sun, Oct 25, 2026 at 01:48:04PM +0100, Uwe Kleine-König wrote:

> The more I think about this taint, the more I believe the commit should
> be reverted and replaced by a some stop gaps for driver_override. (Be it
> a taint or better yet
> https://lore.kernel.org/all/0f7446324f6a0c8f0153d6532d92a6eeecd6a308.1790612298.git.u.kleine-koenig@baylibre.com/
> .)

I also think this patch is way too broad and should be reverted until
something better is designed.

To me this is very similar to the filesystem issue with syzkaller
fuzzing the block device and crashing filesystems. They said
filesystems touching untrusted block devices is out of scope.

Same here, kernel drivers touching untrusted HW are out of scope, we
have never supported that.

I think a taint from force binding outside a match entry and using
new_id are probably about as far as this idea should go.

Even this would still regress most common VFIO instructions though, so
some kind of disable for VFIO would be needed. I don't agree with the
assertion 'servers are already always tainted' that doesn't match my
experience at all.

unbind should not taint, we support hotplug, and manually binding an
authentic match'd driver should not taint.

Jason