From: Ankit Agrawal <ank...@nvidia.com> There are upcoming devices which allow CPU to cache coherently access their memory. It is sensible to expose such memory as NUMA nodes separate from the sysmem node to the OS. The ACPI spec provides a scheme in SRAT called Generic Initiator Affinity Structure [1] to allow an association between a Proximity Domain (PXM) and a Generic Initiator (GI) (e.g. heterogeneous processors and accelerators, GPUs, and I/O devices with integrated compute or DMA engines).
Implement the mechanism to build the GI affinity structures as Qemu currently does not. Introduce a new acpi-generic-initiator object that links a node to a device BDF. During SRAT creation, all such objected are identified and used to add the GI Affinity Structures. A single node per BDF is insufficient for a full utilization of the NVIDIA GPUs MIG (Mult-Instance GPUs) [2] feature. The feature allows partitioning of the GPU device resources (including device memory) into several (upto 8) isolated instances. Each of the partitioned memory requires a dedicated NUMA node to operate. The partitions are not fixed and they can be created/deleted at runtime. Linux OS does not provide a means to dynamically create/destroy NUMA nodes and such feature implementation is expected to be non-trivial. The nodes that OS discovers at the boot time while parsing SRAT remains fixed. So we utilize the GI Affinity structures that allows association between nodes and devices. Multiple GI structures per BDF is possible, allowing creation of multiple nodes in the VM by exposing unique PXM in each of these structures. Implement a new nvidia-acpi-generic-initiator object to associate a range of nodes with a device. The admin will create a range of 8 nodes and associate that with the device using the nvidia-acpi-generic-initiator object. While a configuration of less than 8 nodes per device is allowed, such configuration will prevent utilization of the feature to the fullest. This setting is applicable to all the Grace+Hopper systems. The following is an example of the Qemu command line arguments to create 8 nodes and link them to the device 'dev0': -numa node,nodeid=2 \ -numa node,nodeid=3 \ -numa node,nodeid=4 \ -numa node,nodeid=5 \ -numa node,nodeid=6 \ -numa node,nodeid=7 \ -numa node,nodeid=8 \ -numa node,nodeid=9 \ -device vfio-pci-nohotplug,host=0009:01:00.0,bus=pcie.0,addr=04.0,rombar=0,id=dev0 \ -object nvidia-acpi-generic-initiator,id=gi0,device=dev0,numa-node-start=2,numa-node-count=8 \ The performance benefits can be realized by providing the NUMA node distances appropriately (through libvirt tags or Qemu params). The admin can get the distance among nodes in hardware using `numactl -H`. This series goes along with the vfio-pci variant driver [3] under review. It is expected for a vfio-pci driver to expose this feature through sysfs. Presence of the feature is checked to enable these code changes. Applied over v8.1.0-rc4. Signed-off-by: Ankit Agrawal <ank...@nvidia.com> --- [1] ACPI Spec 6.5, Section 5.2.16.6 [2] https://www.nvidia.com/en-in/technologies/multi-instance-gpu [3] https://lore.kernel.org/all/20230912153032.19935-1-ank...@nvidia.com/ Link for v1: https://lore.kernel.org/all/20230915024559.6565-1-ank...@nvidia.com/ v1 -> v2 - Removed dependency on sysfs to communicate the feature with variant module. - Use GI Affinity SRAT structure instead of Memory Affinity. - No DSDT entries needed to communicate the PXM for the device. SRAT GI structure is used instead. - New objects introduced to establish link between device and nodes. Ankit Agrawal (3): qom: new object to associate device to numa node hw/acpi: Implement the SRAT GI affinity structure qom: Link multiple numa nodes to device using a new object hw/acpi/acpi-generic-initiator.c | 213 +++++++++++++++++++++++ hw/acpi/meson.build | 1 + hw/arm/virt-acpi-build.c | 3 + hw/vfio/pci.c | 2 - hw/vfio/pci.h | 2 + include/hw/acpi/acpi-generic-initiator.h | 64 +++++++ qapi/qom.json | 40 ++++- 7 files changed, 321 insertions(+), 4 deletions(-) create mode 100644 hw/acpi/acpi-generic-initiator.c create mode 100644 include/hw/acpi/acpi-generic-initiator.h -- 2.17.1
