Messages in this thread | | | Date | Fri, 19 Mar 2004 08:20:52 -0800 | From | "Martin J. Bligh" <> | Subject | Re: [PATCH] Introduce nodemask_t ADT [0/7] |
| |
--Jesse Barnes <jbarnes@sgi.com> wrote (on Thursday, March 18, 2004 15:59:42 -0800):
> On Thursday 18 March 2004 3:43 pm, Martin J. Bligh wrote: >> > It's probably not too late to change this to >> > pcibus_to_nodemask(pci_bus *), or pci_to_nodemask(pci_dev *), there >> > aren't that many callers, are there (my grep is still running)? >> >> It probably shouldn't have anything to do with PCI directly either, >> so .... ;-) My former thought was that you might just want the most >> local memory for DMAing into. > > Right, we want local memory (or potentially remote memory) for DMA, > but what about interrupt redirection? Some chipsets don't support > interrupt round robin, and just target interrupts at one CPU. In that > case (and probably the round robin case too), you want to know which > CPU(s) to send the interrupt at. Can't immediately think of other > in-kernel uses though (administrators will of course want to be able > to locate a given PCI device in a multirack system, but that's another > subject--one that Martin Hicks posted on yesterday).
I think we need both ... maybe a cpumask and a zonelist as the destination types.
M.
- 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/
| |