Messages in this thread | | | Date | Mon, 28 Jul 2008 11:54:12 +0200 | From | Michal Schmidt <> | Subject | Re: PCI: MSI interrupts masked using prohibited method |
| |
On Fri, 25 Jul 2008 09:51:46 -0600 Matthew Wilcox <matthew@wil.cx> wrote:
> On Fri, Jul 25, 2008 at 03:53:29PM +0200, Michal Schmidt wrote: > > Maybe we should just not use MSI for devices without maskbits. > > That seems excessive. I wouldn't object to forbidding CPU affinity > changes for devices: > > - Without maskbits > - With the intx quirk > > I'd want to look into it in a bit more detail .... perhaps we could > allow irq affinity if we don't have to change the address, only the > data.
At least on x86, the destination processor is encoded in the message address (see Intel Architecture Software Developer's Manual Vol. 3A, 9.11.1), so it won't help.
Michal
| |