On Fri, Jun 19, 2009 at 09:57:48AM +0200, Thierry Reding wrote: > So the core seems to be able to support multiple functions. On the other > hand, 8 functions is most likely not enough for our purposes. What kind of > limitations does PCIe have with more than eight functions? > > Otherwise I'm probably stuck with either trying to find some other core that > can run in bridge mode or having a monolithic driver to serve all logical > devices. OK. The term you're looking for here is ARI (Alternate Routing-ID Interpretation). This allows a PCIe device to support up to 256 functions instead of 8. It's defined in PCIe 2.1 and is also available as an ECN to PCIe 2.0: http://www.pcisig.com/specifications/pciexpress/specifications/ECN-alt-rid-interpretation-070604.pdf It's not supported by the G33 chipset, but it is supported by the X58 chipset. To find out if your chipset supports it, run lspci -vvv (as root). If you see a line along these lines: DevCap2: Completion Timeout: Range BCD, TimeoutDis+ ARIFwd+ you're in luck (ARIFwd+). If you see a line along these lines: DevCap2: Completion Timeout: Not Supported, TimeoutDis- ARIFwd- you're not (ARIFwd-). DevCtl2 probably says ARIFwd-, but that's just because you don't have an ARI-capable device in the slot. ARI is a somewhat new feature, so you may well encounter bugs in the Linux implementation. For example, I just found a bug in the IRQ swizzling code for ARI devices (and I'll submit a patch for it now). I'm very interested in fixing any bugs you find in the ARI code ;-) -- Matthew Wilcox Intel Open Source Technology Centre "Bill, look, we understand that you're interested in selling us this operating system, but compare it to ours. We can't possibly take such a retrograde step." -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
