On Thu, Nov 09, 2023 at 12:25:59PM -0800, Wesley Cheng wrote: > > > > Since the DeviceTree passed to the OS, should describe the hardware to > > > > the OS, and should represent the hardware from the point-of-view of the > > > > OS, adding one interrupt (ctrl_irq[0]) might be sufficient as Linux > > > > would not use the other interrupts. > > > > > > I've only skimmed the virtualisation bits in xHCI spec, but it seems > > > Linux as VMM would still be involved in assigning these interrupts to > > > VMs. > IMO it might be a bit premature to add definitions for how to utilize > secondary interrupters since design wise, there's nothing really too well > defined yet. At least for the XHCI path, we will have a slew of potential > use cases for secondary interrupters, such as USB audio offloading, or for > VMMs, etc... I've only heard mentions about some of them after pushing the > usb audio offloading series, but I don't have much details on it. I tend to agree. > > > This may possibly be something that we can ignore for now, but perhaps > > > someone more familiar with the hardware, like Thinh, can chime in. > > > You need to get into the same mindset when it comes to devicetree. Even > > > if Linux currently does not use an interrupt, like the pwr_event_irq, > > > you should still add it so that when/if someone implements support for > > > it, an older platform using the original dt may also take advantage of > > > it. > > Yeah, I totally agree with this point, but I'm not sure if adding it into > the "interrupts" array is the way to go. It would probably have to change > as support is added. Yes, that in itself would probably not be sufficient and possibly not even correct. > Sorry for jumping in, but just giving my two cents since I'm the one trying > to do the initial push for the support for secondary interrupters :). Appreciate your input. Johan
