NVIDIA/open-gpu-kernel-modules
1
0
Fork 0
mirror of https://github.com/NVIDIA/open-gpu-kernel-modules.git synced 2026-02-24 00:43:57 +00:00
Code Issues Packages Projects Releases Wiki Activity

535.43.02

Browse Source
This commit is contained in:
Andy Ritger
2023-05-30 10:11:36 -07:00
parent 6dd092ddb7
commit eb5c7665a1
1403 changed files with 295367 additions and 86235 deletions
Download Patch File Download Diff File Expand all files Collapse all files

17
kernel-open/nvidia/nv-procfs.c
View File

@@ -1016,6 +1016,23 @@ numa_status_read(
rm_status = rm_get_gpu_numa_info(sp, nv,
nid, numa_mem_addr, numa_mem_size,
list->addresses, &list->numEntries);
if (rm_status == NV_OK && *nid == NUMA_NO_NODE)
{
//
// RM returns NUMA_NO_NODE when running MIG instances because
// this rmClient is not subscribed to any MIG partition since
// it was subscribed to whole GPU only during RMInit and is not
// updated when MIG partitions are created.
// Returning error here so that numa_status results in EIO
// because of missing support in numa_status to use it for multiple
// numa nodes.
//
// TODO: add support for multiple numa nodes in numa_status interface
// and remove this check, bug 4006012
//
rm_status = NV_ERR_NOT_SUPPORTED;
}
*status = nv_get_numa_status(nvl);
done: