android-tools-conf-configfs: handle multiple UDC controllers on IQ-X7181 #1355

shoudil · 2026-01-06T10:33:20Z

On IQ-X7181 EVK board, below command in android-gadget-start script will hit issue as there are multiple UDC controllers. Overwrite the script to ensure adb work normally.

ls /sys/class/udc/ > /sys/kernel/config/usb_gadget/adb/UDC
in https://github.com/openembedded/meta-openembedded/blob/master/meta-oe/dynamic-layers/selinux/recipes-devtool/android-tools/android-tools-conf-configfs/android-gadget-start

Error:
root@iq-x7181-evk:/# ls /sys/class/udc/ > /sys/kernel/config/usb_gadget/adb/UDC
ls: write error: Device or resource busy

lumag · 2026-01-06T15:53:25Z

...bedded-layer/recipes-devtools/android-tools/android-tools-conf-configfs/android-gadget-start

+while true; do
+    count=`ls /sys/class/udc | wc -l`
+    if [ $count -gt 1 ]; then
+        udcname=a600000.usb


How do we pick up this name? Can we pick up the first one instead of hardcoding it?

I checked all qcom targets supported in meta-qcom, for the targets with multiple USB ports, usb@a600000 was configured to the same USB port to download images (in EDL mode). Fixed the udcname here for easy use during development.
Take Hamoa EVK as an example, if we pick up the first one (which is a000000.usb not a600000.usb), developer have to plug out/in USB cable and try out the third type-c usb port for adb connection after images flashed. not friendly for use.
root@iq-x7181-evk:~# ls /sys/class/udc/
a000000.usb a600000.usb a800000.usb

qcom-next/arch/arm64/boot/dts/qcom$ grep usb_1_ss[0-2]_dwc3 x1e80100.dtsi
remote-endpoint = <&usb_1_ss0_dwc3_ss>;
remote-endpoint = <&usb_1_ss1_dwc3_ss>;
remote-endpoint = <&usb_1_ss2_dwc3_ss>;
usb_1_ss2_dwc3: usb@a000000 { => 3rd type-C usb port
usb_1_ss2_dwc3_hs: endpoint {
usb_1_ss2_dwc3_ss: endpoint {
usb_1_ss0_dwc3: usb@a600000 { => 1st type-C usb port, used to flash images in EDL mode as well
usb_1_ss0_dwc3_hs: endpoint {
usb_1_ss0_dwc3_ss: endpoint {
usb_1_ss1_dwc3: usb@a800000 {
usb_1_ss1_dwc3_hs: endpoint {

qcom-next/arch/arm64/boot/dts/qcom$ grep usb@a600000 *.dtsi
lemans.dtsi: usb_0: usb@a600000 {
qcs8300.dtsi: usb_1: usb@a600000 { => first USB port, there is no usb_0
qdu1000.dtsi: usb_1_dwc3: usb@a600000 {
sar2130p.dtsi: usb_1_dwc3: usb@a600000 {
sc7180.dtsi: usb_1_dwc3: usb@a600000 {
sc7280.dtsi: usb_1: usb@a600000 {
sc8180x.dtsi: usb_prim_dwc3: usb@a600000 {
sc8280xp.dtsi: usb_0_dwc3: usb@a600000 {
sdm670.dtsi: usb_1_dwc3: usb@a600000 {
sdm845.dtsi: usb_1_dwc3: usb@a600000 {
sdx75.dtsi: usb_dwc3: usb@a600000 {
sm6350.dtsi: usb_1_dwc3: usb@a600000 {
sm8150.dtsi: usb_1_dwc3: usb@a600000 {
sm8250.dtsi: usb_1_dwc3: usb@a600000 {
sm8350.dtsi: usb_1_dwc3: usb@a600000 {
sm8450.dtsi: usb_1: usb@a600000 {
sm8550.dtsi: usb_1: usb@a600000 {
sm8650.dtsi: usb_1: usb@a600000 {
sm8750.dtsi: usb_1: usb@a600000 {
talos.dtsi: usb_1_dwc3: usb@a600000 {
x1e80100.dtsi: usb_1_ss0_dwc3: usb@a600000 {

IPQ5424 has two USB controllers, usb@1e00000 and usb@8a00000. Likewise IPQ6018, IPQ8074 have two controllers, none of which match this address. MSM8996 has two USB controllers, both useable as OTG devices and their addresses don't match this one.
So, we need to find a way to find a better UDC.

Are these targets enabled in meta-qcom ? I checked the boards under meta-qcom/conf/machine folder and did not find them under this folder.

qcom-armv8a machine covers generic configuration, which might be used for those machines.

At given point, only one write to this node /sys/kernel/config/usb_gadget/adb/UDC is supported. without usb_gadget driver change, can we implement multiple adbd on UDC ports ? and do we really need multiple adb instances during development, once adb supported, what about the support of "adb reboot bootloader" "adb reboot edl", things become more complicated ?

Why does it become more complicated? ADB will still handle those commands.
By running ADB only on one port, you are limiting our customers to be able to use only one USB-C port for debugging. It might be undesirable. If we have multiple OTG or peripheral ports, there should be no difference in their functionality.

Regarding /sys/kernel/config/usb_gadget/adb/UDC. Of course there should be a separate configuration per each port.

Hi @lumag,
Locally I tried to modify the configuration in android-gadget-setup/start scripts for multiple UDC ports, looks not function as expected. currently there are several high priority tasks on-going. would you mind to merge this PR if I update to only pick the first UDC name to unblock test from APT side ?
same as:
https://github.com/openembedded/meta-openembedded/blob/master/meta-oe/recipes-devtools/android-tools/android-tools-conf-configfs/android-gadget-start
ls /sys/class/udc/ | head -n 1 | xargs echo -n > /sys/kernel/config/usb_gadget/adb/UDC

Yes, it's fine with me. Please open a ticket against meta-qcom to support multiple ADB instances.

Yes, it's fine with me. Please open a ticket against meta-qcom to support multiple ADB instances.

Create an issue to track: #1414

On IQ-X7181 EVK board, below command in android-gadget-start script will hit issue as there are multiple UDC controllers. Overwrite the script to ensure adb work normally. ls /sys/class/udc/ > /sys/kernel/config/usb_gadget/adb/UDC in https://github.com/openembedded/meta-openembedded/blob/master/meta-oe/dynamic-layers/selinux/recipes-devtool/android-tools/android-tools-conf-configfs/android-gadget-start Error: root@iq-x7181-evk:/# ls /sys/class/udc/ > /sys/kernel/config/usb_gadget/adb/UDC ls: write error: Device or resource busy Signed-off-by: Shoudi Li <shoudil@qti.qualcomm.com>

github-actions · 2026-01-21T10:05:00Z

Test run workflow

Test jobs for commit `04cbcaf`

Job 134444 on qcs9100-rb8 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134443 on qcs6490 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134447 on qcs9100 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134440 on qcs8300 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134446 on dragonboard-820c - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134442 on qcs9100-rb8 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134441 on qcs8300 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134439 on qcs6490 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134434 on qcs9100 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134445 on qrb2210-rb1 - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134438 on dragonboard-820c - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134435 on dragonboard-410c - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134437 on dragonboard-410c - meta-qcom-qcom-distro-21204792147-1-boot test
Job 134436 on qrb2210-rb1 - meta-qcom-qcom-distro-21204792147-1-boot test

test-reporting-app · 2026-01-21T10:05:28Z

Test Results

14 files ±0 14 suites ±0 1h 57m 39s ⏱️ + 57m 24s
13 tests ±0 9 ✅ - 1 0 💤 ±0 4 ❌ +1
124 runs +2 116 ✅ - 3 0 💤 ±0 8 ❌ +5

For more details on these failures, see this check.

Results for commit 04cbcaf. ± Comparison against base commit 3cfd302.

ricardosalveti · 2026-01-22T21:59:36Z

There are no differences between your file and the one provided via meta-oe, is this really needed?

ricardosalveti · 2026-01-22T22:05:01Z

Ah, I get what you raised before, the issue is that meta-oe/dynamic-layers/selinux/android-tools/android-tools-conf-configfs/android-gadget-start is getting selected instead, which is basically an older version of the same script, which wasn't updated in meta-clang before.

I believe the right fix here would be to send a patch to meta-oe to update meta-oe/dynamic-layers/selinux/android-tools/android-tools-conf-configfs/android-gadget-start based on the latest from meta-oe/recipes-devtools/android-tools/android-tools-conf-configfs/android-gadget-start (or drop and find a way to reuse the file from the other recipe).

shoudil · 2026-01-23T01:00:55Z

Ah, I get what you raised before, the issue is that meta-oe/dynamic-layers/selinux/android-tools/android-tools-conf-configfs/android-gadget-start is getting selected instead, which is basically an older version of the same script, which wasn't updated in meta-clang before.

I believe the right fix here would be to send a patch to meta-oe to update meta-oe/dynamic-layers/selinux/android-tools/android-tools-conf-configfs/android-gadget-start based on the latest from meta-oe/recipes-devtools/android-tools/android-tools-conf-configfs/android-gadget-start (or drop and find a way to reuse the file from the other recipe).

Yes, agree. That's why I tried to understand the background to maintain another version under dynamic-layer/selinux/xx in #1355 (comment).
From project perspective, I would like to merge this to unblock APT and tech teams. once meta-oe updated, we can revert this.
Raised patch to meta-oe: openembedded/meta-openembedded#1022

shoudil requested review from lumag, ndechesne, ricardosalveti, sbanerjee-quic and vkraleti as code owners January 6, 2026 10:33

lumag requested changes Jan 6, 2026

View reviewed changes

shoudil force-pushed the hamoa-iot-evk-adb branch from d6357f6 to 5751ff2 Compare January 8, 2026 03:30

shoudil force-pushed the hamoa-iot-evk-adb branch from 5751ff2 to 04cbcaf Compare January 21, 2026 07:23

shoudil mentioned this pull request Jan 21, 2026

Support multiple ADB instances #1414

Open

lumag approved these changes Jan 21, 2026

View reviewed changes

android-tools-conf-configfs: handle multiple UDC controllers on IQ-X7181 #1355

Are you sure you want to change the base?

android-tools-conf-configfs: handle multiple UDC controllers on IQ-X7181 #1355

Conversation

shoudil commented Jan 6, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

lumag Jan 6, 2026

Choose a reason for hiding this comment

Uh oh!

shoudil Jan 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lumag Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

shoudil Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

lumag Jan 7, 2026

Choose a reason for hiding this comment

Uh oh!

shoudil Jan 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lumag Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

shoudil Jan 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lumag Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

shoudil Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 21, 2026

Test run workflow

Test jobs for commit 04cbcaf

Uh oh!

test-reporting-app bot commented Jan 21, 2026

Test Results

Uh oh!

ricardosalveti commented Jan 22, 2026

Uh oh!

ricardosalveti commented Jan 22, 2026

Uh oh!

shoudil commented Jan 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

shoudil commented Jan 6, 2026 •

edited

Loading

shoudil Jan 7, 2026 •

edited

Loading

shoudil Jan 13, 2026 •

edited

Loading

shoudil Jan 21, 2026 •

edited

Loading

Test jobs for commit `04cbcaf`

shoudil commented Jan 23, 2026 •

edited

Loading