Will modem inhibition truly release MBIM port ?

[Aleksander Morgado]

Hey Bjørn & Edward

> >> I've opened a new issue in gitlab for your problem:
> >> https://gitlab.freedesktop.org/mobile-broadband/libmbim/-/issues/9
> >> Will take a look at the problem sometime this week.
> >
> > Something is incrementing the transaction id between each fragment:
> >
> > [21  ä¹   2020, 09:55:23] [Debug] [/dev/cdc-wdm_pcie] Sent message (translated)...
> > <<<<<< Header:
> > <<<<<<   length      = 4096
> > <<<<<<   type        = command (0x00000003)
> > <<<<<<   transaction = 4
> > <<<<<< Fragment header:
> > <<<<<<   total   = 65
> > <<<<<<   current = 0
> > <<<<<< Contents:
> > <<<<<<   service = 'qdu' (6427015f-579d-48f5-8c54-f43ed1e76f83)
> > <<<<<<   cid     = 'file-write' (0x00000003)
> > <<<<<<   type    = 'set' (0x00000001)
> > <<<<<< Fields:
> > <<<<<<   DataBuffer =
> > ..
> >
> > <<<<<< Header:
> > <<<<<<   length      = 4096
> > <<<<<<   type        = command (0x00000003)
> > <<<<<<   transaction = 5
> > <<<<<< Fragment header:
> > <<<<<<   total   = 65
> > <<<<<<   current = 1
> > <<<<<< Contents:
> > <<<<<<   service = 'invalid' (00000000-0000-0000-0000-000000000000)
> > <<<<<<   cid     = '(null)' (0x00000000)
> > <<<<<<   type    = 'query' (0x00000000)
> >
> >
> >
> > That's not good.  The function is supposed to err out both transactions
> > when it sees this second fragment with a new transaction id:
> >
> >  If the function receives a fragment that is not the first fragment of a
> >  new command with a new TransactionId, both commands shall be
> >  discarded. One MBIM_FUNCTION_ERROR_MSG message per TransactionId shall
> >  be sent.
>
> A proposed quickfix here:
> https://gitlab.freedesktop.org/mobile-broadband/libmbim/-/merge_requests/32
>

Nice! this makes sense, nice catch :)

Edward, would you be able to test with Bjørn's patch?

> But as noted: A real fix for proxied fragments will have to prevent
> other clients from interfering.  Access to a device should be blocked
> once a client starts a fragmented transaction, until either all
> fragments have been submitted or some reasonable timeout. I guess
> MAX_TIME_BETWEEN_FRAGMENTS_MS could be re-used.
>
> This means that fragmented transactions must be tracked by the proxy.
>

I've opened a new Issue to track this:
https://gitlab.freedesktop.org/mobile-broadband/libmbim/-/issues/10

-- 
Aleksander
https://aleksander.es