From 9f17295edb5176040f63c0e7eb5c2e053a2192b9 Mon Sep 17 00:00:00 2001
From: mesutoezdil <mesudozdil@gmail.com>
Date: Sun, 11 Oct 2026 14:03:27 +0200
Subject: [PATCH] fix(amdgpu): parse the SMC debugfs line that carries a
 program field

Newer kernels print "SMC feature version: 0, program: 7, firmware version:
0x07551300" in amdgpu_firmware_info. The regex required the firmware field
right after the feature field, so SMC was dropped and
TestFirmwareVersionConsistent compared the ioctl value against zero.

Signed-off-by: mesutoezdil <mesudozdil@gmail.com>
---
 internal/pkg/amdgpu/amdgpu.go                     |  3 ++-
 internal/pkg/amdgpu/amdgpu_test.go                | 15 +++++++++++++++
 .../debugfs-parsing/amdgpu_firmware_info_mi325x   |  5 +++++
 3 files changed, 22 insertions(+), 1 deletion(-)
 create mode 100644 testdata/debugfs-parsing/amdgpu_firmware_info_mi325x

diff --git a/internal/pkg/amdgpu/amdgpu.go b/internal/pkg/amdgpu/amdgpu.go
index 5992623e..e0a46b23 100644
--- a/internal/pkg/amdgpu/amdgpu.go
+++ b/internal/pkg/amdgpu/amdgpu.go
@@ -552,7 +552,8 @@ func ParseTopologyProperties(path string, re *regexp.Regexp) (int64, error) {
 	return 0, errors.New("Topology property not found.  Regex: " + re.String())
 }
 
-var fwVersionRe = regexp.MustCompile(`(\w+) feature version: (\d+), firmware version: (0x[0-9a-fA-F]+)`)
+// Newer kernels print an extra "program: N," field on the SMC line.
+var fwVersionRe = regexp.MustCompile(`(\w+) feature version: (\d+),(?: program: \d+,)? firmware version: (0x[0-9a-fA-F]+)`)
 
 func parseDebugFSFirmwareInfo(path string) (map[string]uint32, map[string]uint32) {
 	feat := make(map[string]uint32)
diff --git a/internal/pkg/amdgpu/amdgpu_test.go b/internal/pkg/amdgpu/amdgpu_test.go
index c07b837b..24e46fa6 100644
--- a/internal/pkg/amdgpu/amdgpu_test.go
+++ b/internal/pkg/amdgpu/amdgpu_test.go
@@ -221,6 +221,21 @@ func TestParseDebugFSFirmwareInfo(t *testing.T) {
 	}
 }
 
+// Kernel 7.0 on an MI325X prints "program: N," between the SMC feature and
+// firmware versions; the SMC entry must still parse.
+func TestParseDebugFSFirmwareInfoSMCProgram(t *testing.T) {
+	feat, fw := parseDebugFSFirmwareInfo("../../../testdata/debugfs-parsing/amdgpu_firmware_info_mi325x")
+	if v, ok := fw["SMC"]; !ok || v != 0x07551300 {
+		t.Errorf("SMC firmware = %#x (present %v), want 0x07551300", v, ok)
+	}
+	if v, ok := feat["SMC"]; !ok || v != 0 {
+		t.Errorf("SMC feature = %d (present %v), want 0", v, ok)
+	}
+	if v := fw["MEC"]; v != 0x000080c2 {
+		t.Errorf("MEC firmware = %#x, want 0x80c2 (a line without program must still parse)", v)
+	}
+}
+
 func TestRenderDevIdsFromTopology(t *testing.T) {
 	renderDevIds := GetDevIdsFromTopology("../../../testdata/topology-parsing-mi308")
 
diff --git a/testdata/debugfs-parsing/amdgpu_firmware_info_mi325x b/testdata/debugfs-parsing/amdgpu_firmware_info_mi325x
new file mode 100644
index 00000000..a9c87aaa
--- /dev/null
+++ b/testdata/debugfs-parsing/amdgpu_firmware_info_mi325x
@@ -0,0 +1,5 @@
+RLC feature version: 1, firmware version: 0x00000046
+MEC feature version: 50, firmware version: 0x000080c2
+SMC feature version: 0, program: 7, firmware version: 0x07551300 (85.19.0)
+SDMA0 feature version: 44, firmware version: 0x0000001b
+VCN feature version: 0, firmware version: 0x09118022