On 03/16/2015 05:28 PM, Michael S. Tsirkin wrote:
On Mon, Mar 16, 2015 at 02:16:40PM +0200, Marcel Apfelbaum wrote:
On 03/10/2015 07:42 PM, Michael S. Tsirkin wrote:
On Tue, Mar 10, 2015 at 06:21:14PM +0200, Marcel Apfelbaum wrote:
On 03/10/2015 05:47 PM, Michael S. Tsirkin wrote:
On Tue, Mar 10, 2015 at 05:32:14PM +0200, Marcel Apfelbaum wrote:
Signed-off-by: Marcel Apfelbaum marcel@redhat.com
docs/pci_expander_bridge.txt | 52 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 52 insertions(+) create mode 100644 docs/pci_expander_bridge.txt
diff --git a/docs/pci_expander_bridge.txt b/docs/pci_expander_bridge.txt new file mode 100644 index 0000000..58bf7a8 --- /dev/null +++ b/docs/pci_expander_bridge.txt @@ -0,0 +1,52 @@ +PCI EXPANDER BRIDGE (PXB) +=========================
+Description +=========== +PXB is a "light-weight" host bridge in the same PCI domain +as the main host bridge whose purpose is to enable +the main host bridge to support multiple PCI root buses. +It is implemented only for i440fx.
BTW what makes it i440fx specific? Also, what happens if you try to use it with a different machine type?
Is is i440fx specific, please look at patch 22/28. Also we have a specific check for i440fx, so CRS will not be emitted for other machine types.
Thanks, Marcel
In fact it won't work at all. Need to think about it, maybe we can make it work more generally. For CRS, should be possible to emit for q35 too?
We can make it work, but not on the scope of this series. However, I'll add a IHostBridgeSnoop interface that will make the device work only with associated bus and this will make it less general.
Thanks, Marcel
OK, this works too. Do you mean PCIHostBridgeSnoop?
Sure.
Thanks, Marcel
+As opposed to PCI-2-PCI bridge's secondary bus, PXB's bus +is a primary bus and can be associated with a NUMA node +(different from the main host bridge) allowing the guest OS +to recognize the proximity of a pass-through device to +other resources as RAM and CPUs.
+Usage +===== +A detailed command line would be:
+[qemu-bin + storage options] +-bios [seabios-dir]/out/bios.bin -L [seabios-dir]/out/ +-m 2G +-object memory-backend-ram,size=1024M,policy=bind,host-nodes=0,id=ram-node0 -numa node,nodeid=0,cpus=0,memdev=ram-node0 +-object memory-backend-ram,size=1024M,policy=interleave,host-nodes=0,id=ram-node1 -numa node,nodeid=1,cpus=1,memdev=ram-node1 +-device pxb-device,id=bridge1,bus=pci.0,numa_node=1,bus_nr=4 -netdev user,id=nd-device e1000,bus=bridge1,addr=0x4,netdev=nd +-device pxb-device,id=bridge2,bus=pci.0,numa_node=0,bus_nr=8 -device e1000,bus=bridge2,addr=0x3 +-device pxb-device,id=bridge3,bus=pci.0,bus_nr=40 -drive if=none,id=drive0,file=[img] -device virtio-blk-pci,drive=drive0,scsi=off,bus=bridge3,addr=1
+Here you have:
- 2 NUMA nodes for the guest, 0 and 1. (both mapped to the same NUMA node in host, but you can and should put it in different host NUMA nodes)
- a pxb host bridge attached to NUMA 1 with an e1000 behind it
- a pxb host bridge attached to NUMA 0 with an e1000 behind it
- a pxb host bridge not attached to any NUMA with a hard drive behind it.
+Implementation +============== +The PXB is composed by: +- HostBridge (TYPE_PXB_HOST)
- The host bridge allows to register and query the PXB's rPCI root bus in QEMU.
+- PXBDev(TYPE_PXB_DEVICE)
- It is a regular PCI Device that resides on the piix host-bridge bus and its bus uses the same PCI domain.
- However, the bus behind is exposed through ACPI as a primary PCI bus and starts a new PCI hierarchy.
- The interrupts from devices behind the PXB are routed through this device the same as if it were a
- PCI-2-PCI bridge. The _PRT follows the i440fx model.
+- PCIBridgeDev(TYPE_PCI_BRIDGE_DEV)
- Created automatically as part of init sequence.
- When adding a device to PXB it is attached to the bridge for two reasons:
- Using the bridge will enable hotplug support
- All the devices behind the bridge will use bridge's IO/MEM windows compacting
- the PCI address space.
-- 2.1.0