Re: [PATCH 00/22] Explicitly deny IRQ0 in the USB host drivers

[Alan Stern <stern@rowland.harvard.edu>]

On Tue, Oct 19, 2021 at 09:28:08PM +0300, Sergey Shtylyov wrote:
> On 10/19/21 10:31 AM, Greg Kroah-Hartman wrote:
> 
> [...]
> >>> Here are 22 patches against the 'usb-next' branch of Greg KH's 'usb.git' repo.
> >>> The affected drivers use platform_get_irq() which can return IRQ0 (considered
> >>> invalid, according to Linus) that means broken HCD when passed to usb_add_hcd()
> >>> called at the end of the probe methods. I think that the solution to this issue
> >>> is either explicitly deny or accept IRQ0 in usb_add_hcd()... /but/ here's this
> >>> patch set to get the things going...
> >>
> >> Why not fix the root of the problem for your platform that is failing to
> >> assign a valid irq for the device?
> >>
> >> Are you going to make this change to all callers of this function in the
> >> kernel tree?
> > 
> > Also, you should have gotten a huge WARNING in your kernel log if this
> > happens to let you know that something bad is going on.
> 
>    That's the relatively recent addition, yet it doesn't override IRQ0 to s/th
> like -EINVAL.
> 
> > Is this patch
> > series going to really change any of that?
> 
>    How? It doesn't touch drivers/base/platform.c...
> 
> > 
> > What is the root problem here that you are trying to paper over with
> > this patchset?
> 
>    As I said, it would be preferrable to either deny IRQ0 in usb_add_hcd() or
> just don't try to filter it out. The real problem is that usb_add_hcd() does
> add a non-functioning HCD without the necessary IRQ handling (it only hooks
> an IRQ when it's non-zero).

This is because some HCDs don't use interrupts (e.g., dummy-hcd).

Alan Stern