Consultation about QMI CID Leaks
aleksander at aleksander.es
Fri Dec 24 12:52:16 UTC 2021
> Thanks for your kind reply.
> As you have mentioned, the issue we met has been fixed in commit
> 16383670dcb471d6a5b802a9cca1e2a24eb86de5, but we failed to find this commit in your gitlab.
> Can you please give us a link about this?
Not sure where you looked for it :)
> Also, we have tested thousands times with Modem Manager(1.18.2) and libqmi(1.31.2). When we repeated the
> Syetemctl restart Modem Manager service, there were still CID leaks. And from the hardware signal analysis of modem,
> when restarting Modem Manager, the modem of Quectel did not receive all signal to release, which led to CID leaks.
> The modem failed to reinitialize and stopped to offer service.
> And under same hardware and software conditions, we have tried to modify value of killmode into process on MM and to
> reset MM while not reset qmi-proxy. It seems that there is no abnormality of CID. The QMI clients were implicitly tracked by
> qmi-proxy and the CID was used repeatedly.
> So here is our question. When the service of qmi-proxy stopped, it seems that it's necessary to release the QMI clients which
> are in charge of qmi-proxy.
There are no QMI clients in charge of qmi-proxy, qmi-proxy tracks the
clients managed by other processes. Also, the qmi-proxy must not be
stopped if there are other processes using it. E.g. if MM is still
running, qmi-proxy must never exit, until MM is stopped.
Also, can you confirm you're not running qmicli commands in addition
to MM managing the modem?
> Can you help to explain the root cause of this issue?
Not possible without looking at debug logs of both MM and qmi-proxy
while the issue is happening :) I think it's better if you gather
those logs, find exactly where the issue is happening in the log, and
post them for review to a new Issue in gitlab.
More information about the ModemManager-devel