On Thu, Apr 28, 2011 at 2:51 PM, Matthew Wilcox <matthew@xxxxxx> wrote: > On Wed, Apr 27, 2011 at 11:25:02PM -0500, Xianghua Xiao wrote: >> I have a few large BARs from various PCI/PCIe devices and they >> exceeded my linear PCI mapping space on the ARM chip. The chip can >> only do 256MB mapping at the same time. I need map all devices at the >> same time so I can only assign a portion of certain large BARs, e.g >> >> DeviceX: BAR0=512MB Â--> only assigned 8M resource to it >> ... >> DeviceY: BAR2=512MB Â--->only assigned 16M resource to it >> >> BAR sizes are provided by hardware and they're read-only. ÂI think the >> above map should not cause trouble, correct? DeviceX/DeviceY will use >> the 8M/16M window to address its whole 512M for inbound transaction >> without problems here. I'm hacking pci_read_base to decrease the size. > > You can't do that. ÂThe low N bits of a BAR are forced to be zero. > So while you might try to assign an 8MB address to a 512MB BAR, the device > is still going to decode the entire 512MB range that it's placed in. That's exactly what I am worrying about. > > What you might be able to do is program the bridge of the upstream PCI > bridge to forward on a smaller range. ÂBut if you've got two devices > with 512MB BARs on the same PCI bus, you won't be able to do this either. > > PCI has a resizable BAR capability ... I presume the devices you're > using don't support this. I don't believe we can support that. > > I think you're trying to do something that your hardware doesn't support. > Might I invite you to try using a processor with a larger address space? > :-) Yes somebody else made that decision, sigh. Thanks a lot! > > -- > Matthew Wilcox             ÂIntel Open Source Technology Centre > "Bill, look, we understand that you're interested in selling us this > operating system, but compare it to ours. ÂWe 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
