Re: USB cdc-acm driver: break and command

[Oliver Neukum]

On 16.07.25 19:49, H. Peter Anvin wrote:
> On 2025-07-16 09:17, Oliver Neukum wrote:

> > Understood. It still seems dirty to me. If you want to send strings to a
> > device
> > the proper way is to use a device node and write().
> >   
>
> There is definitely something to be said for that; or at least a file
> descriptor.  We do have cases in the kernel -- notably opening the pts
> corresponding to a ptmx file descriptor -- that do that sort of
> "auxiliary open" kind of thing.
>
> The big question is how that interacts with the rest of the ACM driver,
> as well as all the lifetime issues you mentioned elsewhere.

It would seem to me that CDC already has something very similar in form
of CDC-WDM. If acm_probe() can call tty_port_register_device(), it can also
register a secondary character device. Or are you worried about how to tell
user space which devices belong together?

	Regards
		Oliver