If multiple plugins are selected for probing a modem with AT probes,
earlier plugins may only request some probes but not all. When
those probes complete successfully, subsequent plugins won't re-do
the probes that were already completed, but jump to ones that
haven't yet been done.
For example, plugin A may request only AT probes but plugin B
may request AT, AT_PRODUCT, and AT_VENDOR. Plugin B will start
with AT_PRODUCT probes because plugin A already completed the AT
probe.
The code in probe_step() only opened the serial port for the
AT probe, I guess assuming that all AT-type probes would be
done in sequence. But in the above scenario they are not.
Instead, make sure the port is opened for all AT style probes
that have not yet been executed.
Fixes: 7937a89a37 port-probe: rework and consolidate port probe flow for AT/QCDM/QMI/MBIM
Signed-off-by: Dan Williams <dan@ioncontrol.co>
Add some mmcli D-Bus timeout slack, making them slightly higher than the
ModemManager internal timeout, to ensure the cli doesn't time out the
D-Bus call when MM succeeds.
For 3GPP scan, some modems appear to have internal 5 minute timeouts,
so ensure MM doesn't time out right before the modem returns results:
ModemManager[282578]: <dbg> [1741054422.321400] [ttyUSB2/at] --> 'AT+COPS=?<CR>'
ModemManager[282578]: <wrn> [1741054722.247806] [modem1] failed scanning networks: Serial command timed out
ModemManager[282578]: <dbg> [1741054724.119770] [ttyUSB2/at] <-- '<CR><LF>+COPS: (1,"T-Mobile","T-Mobile","310260",0),,(0,1,2,3,4),(0,1,2)<CR><LF><CR><LF>OK<CR><LF>'
Signed-off-by: Dan Williams <dan@ioncontrol.co>
Instead of relying on getting/setting the UE mode via AT +CEMODE command
which on Quectel RM520N (And I assume other QMI based modems) only allows
getting the value and not setting one it has special AT commands for CS/PS
and voice/data centric values.
So lets instead use the QMI Get/Set System Selection Preference TLV which
offers service domain and usage domain preference info so we can use those
to get/set the 3GPP TS 24.301 UE mode values.
Signed-off-by: Robert Marko <robert.marko@sartura.hr>
Allows setting the renamed `ID_MM_QUALCOMM_FIREHOSE` and
`ID_MM_QUALCOMM_SAHARA` environment variables for all devices.
Signed-off-by: Lukas Voegl <lvoegl@tdt.de>
Add subsystem device ID support, that can be used in pci devices
instead of vendor/device ID customization.
Signed-off-by: Sushrut Shree Trivedi <quic_sushruts@quicinc.com>
The quectel modems that are not using QMI could receive DTMF events but
no dbus notification was sent to the user space. This patch adds the
DTMF notification support to the quectel plugin.
Signed-off-by: Kévin L'hôpital <kevin.lhopital@savoirfairelinux.com>
Since we're also now reading responses we need to check sender
credentials to make sure we're talking to the kernel, and not
some other process spoofing netlink.
Signed-off-by: Dan Williams <dan@ioncontrol.co>
Some modems succeed on enabling "system info" and disabling "serving system" in one
transaction, even though they don't support "system info". This commit separates the
two: first, try to enable "system info", and if it succeeds then disable "serving system".
elogind sends PrepareForSleep(false) when resuming but doesn't necessarily
send PrepareForSleep(true) when sleeping. This means we didn't necessarily
drop our previous inhibitor when it's time to take a new one. So instead of
asserting that we dropped our previous inhibitor, just reuse it.
Seen on a Qualcomm-based Quectel EG915Q on the diag port:
ModemManager[34547]: <dbg> [1732077670.330527] [ttyUSB0/at] <-- ' \0\0\0\8\0\0\0\2\0\0\0\0`%\160$"\6\169\128\200\1\30\197\142\0G\184\225\128\0\1+\127\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254\255\255\255\255\255\255\255\254'
Signed-off-by: Dan Williams <dan@ioncontrol.co>
If by any chance the same ports needs to go through AT and
QMI/MBIM probing, make sure the serial port created during
AT probing is closed before starting the QMI/MBIM probing.
The hotplugged flag was added to simplify modem probe on first time it
is created. Before the re-probe it must be set to FALSE to destroy all
existing netlink interfaces and force full reconfiguration.
The CBMs are read via unsolicted events for which we register
an event handler.
We don't enable unsolicted events explicitly as this already happens
via the `+CNMI=*,*,2,*,*` whens setting up unsolicted events in
`modem_messaging_enable_unsolicited_events` (ETSI TS 127 005).
This adds support for the Cell Broadcast interface allowing to
receive, list, read and delete Cell Broadcast messages via the
org.freedesktop.ModemManager1.Modem.CellBroadcast and
org.freedesktop.ModemManager1.Cbm interfaces.
Signed-off-by: Guido Günther <agx@sigxcpu.org>
This allows to parse the CBMs parameters like channel, part number and
message text. We parse uncompressed GSM7 and UCS2 encodings as
described in 3GPP TS 23.038.
Signed-off-by: Guido Günther <agx@sigxcpu.org>
When setting an operator, AT+COPS overwrites the currently set allowed mode. Equally, the other way around, when setting the current mode, the configured operator is overwritten.
Signed-off-by: Lukas Voegl <lvoegl@tdt.de>
==3851== Conditional jump or move depends on uninitialised value(s)
==3851== at 0x23CF65: network_reject_indication_cb (mm-broadband-modem-qmi.c:5434)
==3851== by 0x4D00865: g_cclosure_marshal_VOID__BOXEDv (in /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0.7200.4)
==3851== by 0x4D196FF: g_signal_emit_valist (in /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0.7200.4)
==3851== by 0x4D19862: g_signal_emit (in /usr/lib/x86_64-linux-gnu/libgobject-2.0.so.0.7200.4)
==3851== by 0x4F7C250: process_indication (qmi-nas.c:70290)
==3851== by 0x4DFA6C4: __qmi_client_process_indication (qmi-client.c:172)
==3851== by 0x4DFEE2D: process_indication_idle (qmi-device.c:1697)
==3851== by 0x4A26C43: g_main_context_dispatch (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7200.4)
==3851== by 0x4A7C2B7: ??? (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7200.4)
==3851== by 0x4A262B2: g_main_loop_run (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7200.4)
==3851== by 0x178042: main (main.c:236)
Fixes https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/931
Referenced commit broke custom init flows because the AT port
was not opened when custom inits were run, which was the case
before the commit landed.
Rework probing flow to follow a step model like much of our
other code, and ensure that the AT port is opened as part
of the custom init flow.
Fixes: 7937a89a37 port-probe: rework and consolidate port probe flow for AT/QCDM/QMI/MBIM
Signed-off-by: Dan Williams <dan@ioncontrol.co>
FM350 modems report an invalid rssi threshold (always 0) in the signal-state response message.
It results in signal quality not being updated. We can use this patch to workaround this bug.
The NL668-EAU 4G, 3G, 2G CAT4 is a modem from fibocom and is using a diffrent
USB_VID for this modem. By adding the new USB_VID, the 'fibocom' plugin is
now responsible and not the 'generic' plugin.
Signed-off-by: Florian Eckert <fe@dev.tdt.de>
Dan Williams
Guido Günther
Robert Marko
Akula Susmitha
Sushrut Shree Trivedi
Michał Sułek
Aleksander Morgado
Lukas Voegl
Kévin L'hôpital
Aleksander Morgado
Daniele Palmas
Valentin Blot
Arnav Singh
Aleksander Morgado
Michal Mazur
Mank Wang
Bolan Wang
Florian Eckert