What's the return of systemd-detect-virt on your device ?

gpatel-fr

Hello

Following the recent post by @mariogrip enumerating among the problems of UT the impossibility to turn off dev mode, I can repro this problem (I think that I use the same device, a Fairphone 5).

I have looked at this problem and seen that this is not a new occurence of the solved issue caused by a code problem in Lomiri settings: the setting is done effectively as returned by

gdbus call --system --dest com.meego.usb_moded --object-path /com/meego/usb_moded --method com.meego.usb_moded.get_config

clearing the switch the mode in settings set the mode to mtp instead of mtp_adb as intended.
However, after rebooting, the mode is again mtp_adb.

There is a systemd service called force-adb that is running at startup and calls the script /usr/libexec/force-adb, in this script there is this line:

# In the emulator, we always want to first have the ADB on
elif systemd-detect-virt --quiet --vm; then
    force_adb_to=1

calling systemd-detect-virt returns on the FP5 returns:

vm-other

In the systemd sources, I have seen that the code is testing /proc/device-tree/hypervisor/compatible and if there is no value at this address it goes on to test other kinds of virtualization and if nothing match return finally VIRTUALIZATION_NONE. Obviously since we are running on a real phone, this is the desired value.
Other code routes, when no other virtualization is detected, are NOT leading to VIRTUALIZATION_NONE but to VIRTUALIZATION_VM_OTHER.

That's the reason why the FP5 can't change the Adb mode.

Now my question: if you have a phone other than a Fairphone 5, can you check on this systemd-detect-virt utility and see if it returns 'vm-other' (bad !) or 'none' (correct) ?

If the FP5 is the only one suffering from this problem this is a device issue.
If all Qualcomm devices (the /proc/device-tree/hypervisor/compatible value is qcom,haven-hypervisor-1.0qcom,haven-hypervisorsimple-bus) have this problem, it's a general problem that must be fixed in the /usr/libexec/force-adb script.

Bonus question: does anyone knows a way to find in which deb package can be found a given file, like this force-adb script ? On normal Ubuntu, apt-file makes it easy, but here we don't have it.

Thanks for helping to advance the Fairphone cause !

Vlad Nirky

@gpatel-fr
root@ubuntu-phablet:/usr/libexec# apt-file search force-adb
usb-moded-ubports-config: /lib/systemd/system/force-adb.path
usb-moded-ubports-config: /lib/systemd/system/force-adb.service
usb-moded-ubports-config: /usr/libexec/force-adb

root@ubuntu-phablet:/home/phablet# dpkg -L usb-moded-ubports-config
/.
/etc
/etc/default
/etc/default/usb-moded.d
/etc/default/usb-moded.d/device-specific-config.conf
/etc/usb-moded
/etc/usb-moded/10-ubports-defaults.ini
/etc/usb-moded/20-ubports-autodetected.ini.in
/etc/usb-moded/90-device-specific-config.ini
/etc/usb-moded/dyn-modes
/etc/usb-moded/dyn-modes/charging_only_adb.ini.in
/etc/usb-moded/dyn-modes/developer_mode.ini.in
/etc/usb-moded/dyn-modes/mtp.ini.in
/etc/usb-moded/dyn-modes/mtp_adb.ini.in
/etc/usb-moded/dyn-modes/rndis.ini.in
/etc/usb-moded/dyn-modes/rndis_adb.ini.in
/etc/usb-moded/run
/etc/usb-moded/run/charging_only_adb-adbd.ini
/etc/usb-moded/run/developer_mode-ssh.ini
/etc/usb-moded/run/developer_mode-udhcpd.ini
/etc/usb-moded/run/mtp_adb-adbd.ini
/etc/usb-moded/run/rndis-tethering.ini
/etc/usb-moded/run/rndis_adb-adbd.ini
/etc/usb-moded/run/rndis_adb-tethering.ini
/lib
détourné par base-files vers : /lib.usr-is-merged
/lib/systemd
/lib/systemd/system
/lib/systemd/system/force-adb.path
/lib/systemd/system/force-adb.service
/lib/systemd/system/ssh.service.d
/lib/systemd/system/ssh.service.d/usb-moded-ssh.conf
/lib/systemd/system/usb-moded-ssh.service
/lib/systemd/system/usb-moded-tethering.service
/lib/systemd/system/usb-moded-udhcpd.service
/lib/systemd/system/usb-moded.service.d
/lib/systemd/system/usb-moded.service.d/ubports-usb-moded-configurator.conf
/usr
/usr/libexec
/usr/libexec/force-adb
/usr/libexec/ubports-usb-moded-configurator
/usr/libexec/usb-moded
/usr/libexec/usb-moded/tethering
/usr/share
/usr/share/doc
/usr/share/doc/usb-moded-ubports-config
/usr/share/doc/usb-moded-ubports-config/changelog.Debian.gz
/usr/share/doc/usb-moded-ubports-config/copyright
/etc/udhcpd.conf
/etc/usb-moded/20-ubports-autodetected.ini
/etc/usb-moded/dyn-modes/charging_only_adb.ini
/etc/usb-moded/dyn-modes/developer_mode.ini
/etc/usb-moded/dyn-modes/mtp.ini
/etc/usb-moded/dyn-modes/mtp_adb.ini
/etc/usb-moded/dyn-modes/rndis.ini
/etc/usb-moded/dyn-modes/rndis_adb.ini

gpatel-fr

@Vlad-Nirky said in What's the return of systemd-detect-virt on your device ?:

root@ubuntu-phablet:/usr/libexec# apt-file search force-adb

thanks - however it don't quite answer my second question that was more general.
On my Fairphone 5, I get:

root@ubuntu-phablet:/home/phablet# apt-file search force-adb
bash: apt-file: command not found

I have tried to install apt-file with 'crackle crack' and got nowhere. I even tried to download it and use it by doing manually what 'crackle crack' is supposed to do and got into a morass of Perl packages.

Vlad Nirky

@gpatel-en
When I first started using Noble, the size of the system partition on my phone did not allow for installation.
So we modified the installer configuration file to exceed this limit.
I kept this initial size (because I haven't repartitioned since and it leaves me a little space to add one or two apt packages (which I don't think will affect the OS).
By the way, in the latest 24.04-2.x, rsync is back.
PS I think this method can not be applied to every device... It depend on the port.

d4f0

@gpatel-fr FP4 here. It returns none.

messayisto

@gpatel-fr Volla Quintus, 24.04.1, output is "none"

gpatel-fr

@Vlad-Nirky said in What's the return of systemd-detect-virt on your device ?:

it leaves me a little space to add one or two apt packages (which I don't think will affect the OS).

I'm afraid I understand what you did

gpatel-fr

@d4f0 @messayisto

thanks for your replies, I'm beginning to think that there is nothing wrong neither in force-adb nor in systemd (!) and the problems comes from the device tree of the FP5, that is, the kernel configuration.
Changing force-adb could be an easy work around but mostly wrong.
I'll try to explore the FP5 kernel code to see if I can find the problem all by myself (enthusiasm level: )

Edit: created this gitlab issue to make the problem more visible.