qmi-proxy crashing?

[Aleksander Morgado]

Hey Paul,

> Have seen this issue a few times now, the qmi-proxy appears to be crashing. Every qmi command for every modem on the system then gets this error:
>
> qmicli --device-open-qmi -p -d /dev/cdc-wdm9 --nas-get-signal-strength
> error: couldn't open the QmiDevice: Couldn't spawn the qmi-proxy
>
>
> Have you seen this before, or can think of a good way to debug it? Can that daemon be restarted?
> So far only rebooting the whole system will bring the modems/qmicli back to operation.
>

If the qmi-proxy crashes, clients get notified via a HUP in the
communication socket, and so e.g. MM would reprobe the modem from
scratch. When using qmicli, if the qmi-proxy has crashed, it should be
able to relaunch it.

The "Couldn't spawn the qmi-proxy" error is really weird; to me it
doesn't look like a proxy crash. Did you check whether the proxy
process still exists or not after this situation? The fact that the
proxy doesn't reply and that further qmicli commands don't reply could
actually mean the proxy is "stuck".

> Here are the logs of it happening in one instance, below. Modem0 reboots, CID allocation fails, I query the modem a few times with  ['--wda-get-data-format'] and it times out (which is normal after it has rebooted - takes a while for it to be able to respond to this)
> and suddenly all qmi commands start to fail. I notice around this time, just beforehand, modem commands start to time out in general. Previously to this I noticed that, for another modem, I had told it to reset via libqmi and it returns "successful", but the modem actually never resets. No matter how many times I run the command to do so.

How are you running the reset command via QMI? Are you putting it
first in "offline" before "reset"? Won't work otherwise.

> Not the first time that has happened also.
>
> modem0: qmi_cmd: resp="error: couldn't create client for the 'wda' service: CID allocation failed in the CTL client: Transaction timed out\n"
> modem0: qmi_cmd: ['--wda-get-data-format']
> modem0: qmi_cmd: resp="error: couldn't open the QmiDevice: Transaction timed out\n"
> modem0: qmi_cmd: ['--wda-get-data-format']
> modem0: qmi_cmd: resp="error: couldn't open the QmiDevice: Transaction timed out\n"
> modem0: qmi_cmd: ['--wda-get-data-format']
> modem3: nas_get_serving_system: get_packet_service_status failed "error: couldn't open the QmiDevice: Transaction timed out\n"
> modem2: nas_get_serving_system: get_packet_service_status failed "error: couldn't open the QmiDevice: Transaction timed out\n"
> modem0: qmi_cmd: resp="error: couldn't open the QmiDevice: Transaction timed out\n"
> modem0: qmi_cmd: ['--wda-get-data-format']
> modem0: qmi_cmd: resp="error: couldn't open the QmiDevice: Couldn't spawn the qmi-proxy\n"
>

If you can somehow reproduce this issue easily, it may be a good idea
to launch the qmi-proxy manually before any qmicli call happens, e..g:
$ sudo /usr/libexec/qmi-proxy --verbose

Or even under gdb, if that's something you can do.

-- 
Aleksander
https://aleksander.es