On Sun, 22 Dec 2002, Paul Mackerras wrote:
>
> Linus Torvalds writes:
>
> > Actually, I think it's certainly valid to not allow "printk()" to happen
> > around the BAR probing, at least at bootup when we control all the CPU's
> > tightly anyway.
>
> I'd like us to disable interrupts too. On powermacs, the interrupt
> controller is typically inside a combo I/O ASIC which is on the PCI
> bus.
I certainly wpouldn't mind disabling local interrupts around that code
sequence, but the problem with that is that I don't think it much matters,
since it wouldn't help the SMP case anyway (and in fact even doing an
old-style "cli/sti" if we still supported it would not have helped either,
since the global interrupt disable would still have allowed the other
CPU's to touch the irq controller).
And since the pci bus will be scanned in an "subsys_initcall", the other
CPU's are definitely active at that point.
In short, I don't think we can easily fix this. And again, I think that
the only good approach would be to really make sure the machine is
quescent while the PCI bus is initially scanned (and again, the later
scans behing proper hotplug bus bridges should be fine thanks to bridge
decoding limits).
Sadly, as I already mentioned, there isn't any generic PCI method of
shutting up device interrupts.
I think the only workable setup (which we effectively already do on x86,
not that we should have this problem in the first place) is to just turn
off the interrupts _at_ the interrupt controller. Which simply means that
we really shouldn't be getting any interrupts while probing the root bus.
Then you enable the interrupts as device drivers request them..
Linus
-
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/
This archive was generated by hypermail 2b29 : Mon Dec 23 2002 - 22:00:31 EST