[RFC PATCH 0/1] enable quirks via module parameter

[Keith Busch]

On Fri, May 16, 2025 at 09:59:10AM -0600, Keith Busch wrote:
> On Fri, May 16, 2025 at 05:50:24PM +0200, Maurizio Lombardi wrote:
> > On some occasions, our customers have complained about misbehaving
> > NVMe devices. We had to compile and provide them
> > with a test kernel just to see if a quirk could resolve the issue.
> > 
> > What do you think about allowing users to enable or disable a
> > quirk via a module parameter, similar to how usbcore does it?
> 
> Can't we already do this for pci using the "new_id" attribute of the
> nvme-pci driver? It takes up to 7 parameters, the last one being the
> "driver_data", which is what we use to setup the quirks, and you can set
> that value to anything you want through that interface.

Oops, maybe not "anything you want", but you can set it to any value
that exists in the driver's table. That may be an unnecessary limitation
in the new_id_store() function, but let's see if that's sufficient
before considering changing that.