Skip to content

Commit 8058d3c

Browse files
mythimythi
authored
Merge pull request intel#802 from eero-t/nfd-gen-label
Adapt GPU label support to debugfs DRM entry changes
2 parents 434b7d6 + bcc737b commit 8058d3c

File tree

3 files changed

+119
-18
lines changed

3 files changed

+119
-18
lines changed

‎cmd/gpu_nfdhook/README.md‎

Lines changed: 4 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -43,9 +43,11 @@ Depending on your kernel driver, running the NFD hook as root may introduce foll
4343

4444
name | type | description|
4545
-----|------|------|
46-
|`gpu.intel.com/platform_gen`| string | GPU platform generation name, typically a number.
46+
|`gpu.intel.com/platform_gen`| string | GPU platform generation name, typically an integer. Deprecated.
47+
|`gpu.intel.com/media_version`| string | GPU platform Media pipeline generation name, typically a number.
48+
|`gpu.intel.com/graphics_version`| string | GPU platform graphics/compute pipeline generation name, typically a number.
4749
|`gpu.intel.com/platform_<PLATFORM_NAME>_.count`| number | GPU count for the named platform.
4850
|`gpu.intel.com/platform_<PLATFORM_NAME>_.tiles`| number | GPU tile count in the GPUs of the named platform.
4951
|`gpu.intel.com/platform_<PLATFORM_NAME>_.present`| string | "true" for indicating the presense of the GPU platform.
5052

51-
For the above to work as intended, installed GPUs must be identical in their capabilities.
53+
For the above to work as intended, GPUs on the same node must be identical in their capabilities.

‎cmd/gpu_nfdhook/labeler.go‎

Lines changed: 52 additions & 15 deletions
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
// Copyright 2020 Intel Corporation. All Rights Reserved.
1+
// Copyright 2020-2021 Intel Corporation. All Rights Reserved.
22
//
33
// Licensed under the Apache License, Version 2.0 (the "License");
44
// you may not use this file except in compliance with the License.
@@ -171,33 +171,70 @@ func (l *labeler) createCapabilityLabels(cardNum string, numTiles uint64) {
171171
}
172172
defer file.Close()
173173

174-
// define strings to search from the file, and the actions to take in order to create labels from those strings (as funcs)
174+
gen := ""
175+
media := ""
176+
graphics := ""
177+
// define string prefixes to search from the file, and the actions to take in order to create labels from those strings (as funcs)
175178
searchStringActionMap := map[string]func(string){
176-
"platform: ": func(platformName string) {
179+
"platform:": func(platformName string) {
177180
l.labels.addNumericLabel(labelNamespace+"platform_"+platformName+".count", 1)
178181
l.labels[labelNamespace+"platform_"+platformName+".tiles"] = strconv.FormatInt(int64(numTiles), 10)
179182
l.labels[labelNamespace+"platform_"+platformName+".present"] = "true"
180183
},
181-
"gen: ": func(genName string) {
182-
l.labels[labelNamespace+"platform_gen"] = genName
184+
// there's also display block version, but that's not relevant
185+
"media version:": func(version string) {
186+
l.labels[labelNamespace+"media_version"] = version
187+
media = version
188+
},
189+
"graphics version:": func(version string) {
190+
l.labels[labelNamespace+"graphics_version"] = version
191+
graphics = version
192+
},
193+
"gen:": func(version string) {
194+
l.labels[labelNamespace+"platform_gen"] = version
195+
gen = version
183196
},
184197
}
185198

186199
// Finally, read the file, and try to find the matches. Perform actions and reduce the search map size as we proceed. Return at 0 size.
187200
scanner := bufio.NewScanner(file)
201+
scanning:
188202
for scanner.Scan() {
189-
for searchString, action := range searchStringActionMap {
190-
var stringValue string
191-
n, _ := fmt.Sscanf(scanner.Text(), searchString+"%s", &stringValue)
192-
if n > 0 {
193-
action(stringValue)
194-
delete(searchStringActionMap, searchString)
195-
if len(searchStringActionMap) == 0 {
196-
return
197-
}
198-
break
203+
line := scanner.Text()
204+
for prefix, action := range searchStringActionMap {
205+
if !strings.HasPrefix(line, prefix) {
206+
continue
207+
}
208+
fields := strings.Split(line, ": ")
209+
if len(fields) == 2 {
210+
action(fields[1])
211+
} else {
212+
klog.Warningf("invalid '%s' line format: '%s'", file.Name(), line)
213+
}
214+
delete(searchStringActionMap, prefix)
215+
if len(searchStringActionMap) == 0 {
216+
break scanning
199217
}
218+
break
219+
}
220+
}
221+
if gen == "" {
222+
// TODO: drop gen label before engine types
223+
// start to have diverging major gen values
224+
if graphics != "" {
225+
gen = graphics
226+
} else if media != "" {
227+
gen = media
228+
}
229+
if gen != "" {
230+
// truncate to major value
231+
gen = strings.SplitN(gen, ".", 2)[0]
232+
l.labels[labelNamespace+"platform_gen"] = gen
200233
}
234+
} else if media == "" && graphics == "" {
235+
// 5.14 or older kernels need this
236+
l.labels[labelNamespace+"media_version"] = gen
237+
l.labels[labelNamespace+"graphics_version"] = gen
201238
}
202239
}
203240

‎cmd/gpu_nfdhook/labeler_test.go‎

Lines changed: 63 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,4 +1,4 @@
1-
// Copyright 2020 Intel Corporation. All Rights Reserved.
1+
// Copyright 2020-2021 Intel Corporation. All Rights Reserved.
22
//
33
// Licensed under the Apache License, Version 2.0 (the "License");
44
// you may not use this file except in compliance with the License.
@@ -58,6 +58,8 @@ func getTestCases() []testcase {
5858
"gpu.intel.com/platform_new.count": "1",
5959
"gpu.intel.com/platform_new.present": "true",
6060
"gpu.intel.com/platform_new.tiles": "1",
61+
"gpu.intel.com/graphics_version": "9",
62+
"gpu.intel.com/media_version": "9",
6163
"gpu.intel.com/platform_gen": "9",
6264
"gpu.intel.com/cards": "card0",
6365
},
@@ -104,6 +106,8 @@ func getTestCases() []testcase {
104106
"gpu.intel.com/platform_new.count": "1",
105107
"gpu.intel.com/platform_new.present": "true",
106108
"gpu.intel.com/platform_new.tiles": "2",
109+
"gpu.intel.com/graphics_version": "9",
110+
"gpu.intel.com/media_version": "9",
107111
"gpu.intel.com/platform_gen": "9",
108112
"gpu.intel.com/cards": "card0",
109113
},
@@ -132,6 +136,8 @@ func getTestCases() []testcase {
132136
"gpu.intel.com/platform_new.count": "1",
133137
"gpu.intel.com/platform_new.present": "true",
134138
"gpu.intel.com/platform_new.tiles": "1",
139+
"gpu.intel.com/graphics_version": "9",
140+
"gpu.intel.com/media_version": "9",
135141
"gpu.intel.com/platform_gen": "9",
136142
"gpu.intel.com/cards": "card0",
137143
},
@@ -157,6 +163,8 @@ func getTestCases() []testcase {
157163
"gpu.intel.com/platform_new.count": "1",
158164
"gpu.intel.com/platform_new.present": "true",
159165
"gpu.intel.com/platform_new.tiles": "1",
166+
"gpu.intel.com/graphics_version": "9",
167+
"gpu.intel.com/media_version": "9",
160168
"gpu.intel.com/platform_gen": "9",
161169
"gpu.intel.com/cards": "card0",
162170
},
@@ -184,6 +192,60 @@ func getTestCases() []testcase {
184192
"gpu.intel.com/cards": "card0",
185193
},
186194
},
195+
{
196+
sysfsdirs: []string{
197+
"card0/device/drm/card0",
198+
},
199+
sysfsfiles: map[string][]byte{
200+
"card0/device/vendor": []byte("0x8086"),
201+
},
202+
name: "gen version missing, but media & graphics versions present",
203+
memoryOverride: 16000000000,
204+
capabilityFile: map[string][]byte{
205+
"0/i915_capabilities": []byte(
206+
"platform: new\n" +
207+
"media version: 12.5\n" +
208+
"graphics version: 12.2"),
209+
},
210+
expectedRetval: nil,
211+
expectedLabels: labelMap{
212+
"gpu.intel.com/millicores": "1000",
213+
"gpu.intel.com/memory.max": "16000000000",
214+
"gpu.intel.com/platform_new.count": "1",
215+
"gpu.intel.com/platform_new.present": "true",
216+
"gpu.intel.com/platform_new.tiles": "1",
217+
"gpu.intel.com/graphics_version": "12.2",
218+
"gpu.intel.com/media_version": "12.5",
219+
"gpu.intel.com/platform_gen": "12",
220+
"gpu.intel.com/cards": "card0",
221+
},
222+
},
223+
{
224+
sysfsdirs: []string{
225+
"card0/device/drm/card0",
226+
},
227+
sysfsfiles: map[string][]byte{
228+
"card0/device/vendor": []byte("0x8086"),
229+
},
230+
name: "only media version present",
231+
memoryOverride: 16000000000,
232+
capabilityFile: map[string][]byte{
233+
"0/i915_capabilities": []byte(
234+
"platform: new\n" +
235+
"media version: 12.5"),
236+
},
237+
expectedRetval: nil,
238+
expectedLabels: labelMap{
239+
"gpu.intel.com/millicores": "1000",
240+
"gpu.intel.com/memory.max": "16000000000",
241+
"gpu.intel.com/platform_new.count": "1",
242+
"gpu.intel.com/platform_new.present": "true",
243+
"gpu.intel.com/platform_new.tiles": "1",
244+
"gpu.intel.com/media_version": "12.5",
245+
"gpu.intel.com/platform_gen": "12",
246+
"gpu.intel.com/cards": "card0",
247+
},
248+
},
187249
{
188250
sysfsdirs: []string{
189251
"card0/device/drm/card0",

0 commit comments

Comments
 (0)