nvidia-k8s-device-plugin: add two patches

0001: generate missing CDI specifications

The Device Plugin doesn't generate CDI specifications for
the "all" device, which is what would be used when the env
variable NVIDIA_VISIBLE_DEVICES is set to "all".

This breaks compatibility with containers that rely on bypassing
Kubernetes directives to have access to all the GPUs (e.g. DCGM
Exporter).

Follow what nvidia-ctk[1] does to generate the CDI specifications
for the "all" device.

[1]: https://github.com/NVIDIA/nvidia-container-toolkit/blob/6394e9e9e7f8692438f2c921c0bbc95b72d693c6/cmd/nvidia-ctk/cdi/generate/generate.go#L300-L303

0002: fix ldcache parsing for aarch64

k8s-device-plugin carries its own nvidia-container-toolkit and uses
nvidia-ctk to generate the CDI specifications.

The architecture flag for aarch64 is currently missing from the
supported architecture flags list. This omission causes the getEntries
function to exclude all libraries found on aarch64 hosts. As a result
helper programs like nvidia-ctk are unable to generate CDI
specifications for the aarch64 architecture.

This fix adds the missing aarch64 architecture flag, using the same
value as defined in libnvidia-container[1], which maintains a more
comprehensive list of supported architectures.

[1]: https://github.com/NVIDIA/libnvidia-container/blob/a198166e1c1166f4847598438115ea97dacc7a92/src/ldcache.h#L21

Signed-off-by: Jingwei Wang <[email protected]>

Author: sky1122

packages/nvidia-k8s-device-plugin/0001-Add-CDI-specs-for-the-all-device.patch

@@ -0,0 +1,45 @@
+From fcce51f694a9f4410183a93c7b004175cbe46c47 Mon Sep 17 00:00:00 2001
+From: Jingwei Wang <[email protected]>
+Date: Tue, 15 Apr 2025 18:17:21 +0000
+Subject: [PATCH]  Add CDI specs for the "all" device
+
+The Device Plugin cannot generate CDI specifications
+when NVIDIA_VISIBLE_DEVICES is set to "all". This limitation breaks compatibility
+with containers that depend on this setting, e.g. DCGM Exporter.
+
+Fix this by mimicking what nvidia-ctk does[1], and add the "all" device
+to the CDI specifications.
+
+[1]: https://github.com/NVIDIA/nvidia-container-toolkit/blob/6394e9e9e7f8692438f2c921c0bbc95b72d693c6/cmd/nvidia-ctk/cdi/generate/generate.go#L300-L303
+
+Signed-off-by: Jingwei Wang <[email protected]>
+---
+ internal/cdi/cdi.go | 5 +++++
+ 1 file changed, 5 insertions(+)
+
+diff --git a/internal/cdi/cdi.go b/internal/cdi/cdi.go
+index fae421b..c53ee73 100644
+--- a/internal/cdi/cdi.go
++++ b/internal/cdi/cdi.go
+@@ -38,6 +38,7 @@ import (
+ 
+ const (
+ 	cdiRoot = "/var/run/cdi"
++	allDeviceName = "all"
+ )
+ 
+ // cdiHandler creates CDI specs for devices assocatied with the device plugin
+@@ -124,6 +125,10 @@ func New(infolib info.Interface, nvmllib nvml.Interface, devicelib device.Interf
+ 		nvcdi.WithDeviceNamers(deviceNamer),
+ 		nvcdi.WithVendor(c.vendor),
+ 		nvcdi.WithClass("gpu"),
++		nvcdi.WithMergedDeviceOptions(
++			transform.WithName(allDeviceName),
++			transform.WithSkipIfExists(true),
++		),
+ 	)
+ 	if err != nil {
+ 		return nil, fmt.Errorf("failed to create nvcdi library: %v", err)
+-- 
+2.47.0
+

packages/nvidia-k8s-device-plugin/0002-PATCH-ldcache-fix-parsing-for-aarch64.patch

@@ -0,0 +1,49 @@
+From b5a3c4701090ca078ef4f0f0e50de4cb847e8854 Mon Sep 17 00:00:00 2001
+From: Jingwei Wang <[email protected]>
+Date: Wed, 23 Apr 2025 17:17:35 +0000
+Subject: [PATCH] [PATCH] ldcache: fix parsing for aarch64
+
+k8s-device-plugin carries its own nvidia-container-toolkit and uses
+nvidia-ctk to generate the CDI specifications.
+
+The architecture flag for aarch64 is currently missing from the
+supported architecture flags list. This omission causes the getEntries
+function to exclude all libraries found on aarch64 hosts. As a result
+helper programs like nvidia-ctk are unable to generate CDI
+specifications for the aarch64 architecture.
+
+This fix adds the missing aarch64 architecture flag, using the same
+value as defined in libnvidia-container[1], which maintains a more
+comprehensive list of supported architectures.
+
+[1]: https://github.com/NVIDIA/libnvidia-container/blob/a198166e1c1166f4847598438115ea97dacc7a92/src/ldcache.h#L21
+
+Signed-off-by: Jingwei Wang <[email protected]>
+---
+ .../nvidia-container-toolkit/internal/ldcache/ldcache.go       | 3 +++
+ 1 file changed, 3 insertions(+)
+
+diff --git a/vendor/github.com/NVIDIA/nvidia-container-toolkit/internal/ldcache/ldcache.go b/vendor/github.com/NVIDIA/nvidia-container-toolkit/internal/ldcache/ldcache.go
+index 4daf95b..455048c 100644
+--- a/vendor/github.com/NVIDIA/nvidia-container-toolkit/internal/ldcache/ldcache.go
++++ b/vendor/github.com/NVIDIA/nvidia-container-toolkit/internal/ldcache/ldcache.go
+@@ -47,6 +47,7 @@ const (
+ 	flagArchX8664   = 0x0300
+ 	flagArchX32     = 0x0800
+ 	flagArchPpc64le = 0x0500
++	flagArchAarch64 = 0x0a00
+ )
+ 
+ var errInvalidCache = errors.New("invalid ld.so.cache file")
+@@ -195,6 +196,8 @@ func (c *ldcache) getEntries() []entry {
+ 		switch e.Flags & flagArchMask {
+ 		case flagArchX8664:
+ 			fallthrough
++		case flagArchAarch64:
++			fallthrough
+ 		case flagArchPpc64le:
+ 			bits = 64
+ 		case flagArchX32:
+-- 
+2.47.0
+

packages/nvidia-k8s-device-plugin/nvidia-k8s-device-plugin.spec

@@ -18,6 +18,8 @@ Source2: nvidia-k8s-device-plugin-conf
 Source3: nvidia-k8s-device-plugin-exec-start-conf
 Source4: nvidia-k8s-device-plugin-mig-conf
 
+Patch0001: 0001-Add-CDI-specs-for-the-all-device.patch
+Patch0002: 0002-PATCH-ldcache-fix-parsing-for-aarch64.patch
 
 BuildRequires: %{_cross_os}glibc-devel
 Requires: %{name}(binaries)