RE: RFC Proposal to enable MSI support in Linux kernel

Nakajima, Jun (jun.nakajima@intel.com)
Wed, 14 May 2003 13:43:15 -0700


ISA devices typically use hard-coded IRQs (< 16), and the drivers for those also assume such IRQs. You can break the driver if you provide the vector number instead of the IRQ for them. Unfortunately, vectors <16 are used for the processor, and thus not available. Good news is that vectors assigned to I/O are equal to or higher than FIRST_DEVICE_VECTOR (> 16), so there are no conflicts between the IRQ and vector space, i.e. you don't need to check if it's vector or IRQ at do_IRQ() time, for example.

Non legacy drivers are all drivers but such legacy drivers. Included are PCI, USB drivers, etc.

Thanks,
Jun

> -----Original Message-----
> From: Zwane Mwaikambo [mailto:zwane@linuxpower.ca]
> Sent: Wednesday, May 14, 2003 12:54 PM
> To: Nakajima, Jun
> Cc: Nguyen, Tom L; linux-kernel@vger.kernel.org; Saxena, Sunil; Mallick,
> Asit K; Carbonari, Steven
> Subject: RE: RFC Proposal to enable MSI support in Linux kernel
>
> On Wed, 14 May 2003, Nakajima, Jun wrote:
>
> > That's a good idea, and that's the way we did this (we added MSI
> > support on top of the vector-based indexing). If people are interested
> in
> > the vector-based indexing (i.e. provide the vector number to device
> > drivers (non-legacy drivers only) instead of IRQ) for some other uses,
> we
> > would like to discuss possible cleaner implementations.
> >
> > Long will post a patch containing the vector-based indexing part only.
>
> Thanks! It'd be a lot easier to test the core changes too. What do you
> mean by non legacy?
>
> Zwane
> --
> function.linuxpower.ca
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/