When we explicitly request to perform a SIM slot switch operation, we
should ignore the 'slot info status' and 'subscriber ready status'
indications that report an active SIM slot change, or otherwise they'll
end up interfering with the correct reporting of the slot switch
operation.
We could have attempted to complete the slot switch operation early
using the indications, but that solution would have forced us to
assume that theinfo in the indications is the one we had requested,
which is not easy because the 'slot info status' indication is
reported once for each slot, so a single SIM slot switch operation
triggers multiple 'slot info status' indications, as well as
additional 'subscriber ready status' indications reporting the same.
Fixes suggested by Pavan Holla <pholla@chromium.org> and tested by
Jinjian Song <Jinjian.Song@fibocom.com>.
Fixes https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/550
We no longer have separate mm_base_modem_process_sim_event() and
mm_broadband_modem_sim_hot_swap_detected() methods. The only
difference between both of them was that one of them would attempt to
cleanup the ports context associated to the SIM hot swap event logic
as soon as a swap was detected, in order to avoid queueing up multiple
such events.
The previous logic wasn't working well, though, as there could be
mixed AT+QMI or AT+MBIM devices that would also require that same
cleanup and so we didn't always know which one should have been
called.
Now we have a single mm_iface_modem_process_sim_event() method, which
will trigger the reprobe and disabling, but which will also perform
the cleanup of the SIM ports swap setup as specified by the
implementation.
So, if a plugin explicitly initializes the serial ports context for
SIM hot swap handling, it should also explicitly clean it up.
Also, the initialization of the serial ports context for SIM hot swap
handling is no longer done automatically for all modems, it will be
done only for those modems using it; i.e. the modems that explicitly
report support SIM hot swap handling using AT URCs.
This property is used in the MMIfaceModem to flag whether the SIM hot
swap setup has been performed or not. The flag is now moved to the
iface-specific private context.
The property was also used in AT-based modems, so that implementations
supporting the SIM hot swap via AT URCs could flag the upper layers
whether the enabling of the feature was done correctly or not, and if
so, create and keep the AT ports context open. But this feature only
made sense in AT-based modems, i.e. an MBIM modem that detects SIM hot
swaps via MBIM indications exclusively should not require the AT ports
context open for anything. The check in the MMBroadbandModem object
has therefore been removed, and the logic will be updated so that it
only applies to AT-based modems.
The signal and access tech loading operation was using a context that
contained both overall state and state specific to the current
attempt.
The overall modem state is now moved to the iface-specific private
context, and the state specific to the operation is kept in a GTask
that is bound to the ongoing loading operation.
Avoid keeping separate per-operation contexts, it doesn't make
sense. Just have one single Private context for the whole interface
where we can add support for all operations performed by the
interface.
The SIM hot swap operation relies on multiple different types of
indications, but the most important one is the subscriber info
indication. We should make sure that indication is kept enabled for as
long as the SIM hot swap detection should be active.
We also skip using the MM_IFACE_MODEM_SIM_HOT_SWAP_CONFIGURED
property, which should be removed or made MMBroadbandModem-specific,
as it is only really used in the AT-based SIM hot swap detection
logic.
This is not extremely important, because we don't really flag as
disabled those indications that failed to be enabled (as usually we
attempt to enable them in a group, not one by one), but in this case
this will allow us to keep a clean state after a failed SIM hot swap
setup.
==11025== 40 bytes in 1 blocks are definitely lost in loss record 3,313 of 5,483
==11025== at 0x483E7C5: malloc (vg_replace_malloc.c:380)
==11025== by 0x4A32CA9: g_malloc (in /usr/lib/libglib-2.0.so.0.7000.2)
==11025== by 0x4A4A082: g_slice_alloc (in /usr/lib/libglib-2.0.so.0.7000.2)
==11025== by 0x49F8574: g_array_sized_new (in /usr/lib/libglib-2.0.so.0.7000.2)
==11025== by 0x24DEB1: load_supported_modes_mbim (mm-broadband-modem-mbim.c:958)
==11025== by 0x24E2D5: modem_load_supported_modes (mm-broadband-modem-mbim.c:1042)
==11025== by 0x1D4DC5: interface_initialization_step (mm-iface-modem.c:5718)
==11025== by 0x1D26D6: load_device_identifier_ready (mm-iface-modem.c:4903)
==11025== by 0x4BACB83: ??? (in /usr/lib/libgio-2.0.so.0.7000.2)
==11025== by 0x4BACBB8: ??? (in /usr/lib/libgio-2.0.so.0.7000.2)
==11025== by 0x4A2A434: g_main_context_dispatch (in /usr/lib/libglib-2.0.so.0.7000.2)
==11025== by 0x4A7E7B8: ??? (in /usr/lib/libglib-2.0.so.0.7000.2)
For NR5G NSA, both LTE and NR5G should be set in access technology.
Use NAS "system info" or "serving system", instead of NAS "signal info"
or "signal strength", for deciding access technology based on current
registration state.
We're returning the GError as GTask result, we should not explicitly
free it ourselves.
Fixes the crash sometimes seen in
https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/550
==15994== Invalid free() / delete / delete[] / realloc()
==15994== at 0x484B27F: free (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==15994== by 0x4F9A8A0: g_error_free (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x2251CC: ??? (in /usr/sbin/ModemManager)
==15994== by 0x4DD2B48: ??? (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7102.0)
==15994== by 0x4DD2D4A: ??? (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7102.0)
==15994== by 0x4CA9049: ??? (in /usr/lib/x86_64-linux-gnu/libmbim-glib.so.4.6.0)
==15994== by 0x4CB0367: ??? (in /usr/lib/x86_64-linux-gnu/libmbim-glib.so.4.6.0)
==15994== by 0x4FAF287: ??? (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4FAEC03: g_main_context_dispatch (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x5003497: ??? (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4FAE282: g_main_loop_run (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x16916F: main (in /usr/sbin/ModemManager)
==15994== Address 0x80db490 is 0 bytes inside a block of size 16 free'd
==15994== at 0x484B27F: free (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==15994== by 0x4F9A8A0: g_error_free (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x1984FA: ??? (in /usr/sbin/ModemManager)
==15994== by 0x4DD2B48: ??? (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7102.0)
==15994== by 0x4DD2D4A: ??? (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7102.0)
==15994== by 0x2251A6: ??? (in /usr/sbin/ModemManager)
==15994== by 0x4DD2B48: ??? (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7102.0)
==15994== by 0x4DD2D4A: ??? (in /usr/lib/x86_64-linux-gnu/libgio-2.0.so.0.7102.0)
==15994== by 0x4CA9049: ??? (in /usr/lib/x86_64-linux-gnu/libmbim-glib.so.4.6.0)
==15994== by 0x4CB0367: ??? (in /usr/lib/x86_64-linux-gnu/libmbim-glib.so.4.6.0)
==15994== by 0x4FAF287: ??? (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4FAEC03: g_main_context_dispatch (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== Block was alloc'd at
==15994== at 0x4848899: malloc (in /usr/libexec/valgrind/vgpreload_memcheck-amd64-linux.so)
==15994== by 0x4FB76F8: g_malloc (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4FCF1A4: g_slice_alloc (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4FCF80D: g_slice_alloc0 (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4F9A3A6: ??? (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4F9A5A3: g_error_copy (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4CA903E: ??? (in /usr/lib/x86_64-linux-gnu/libmbim-glib.so.4.6.0)
==15994== by 0x4CB0367: ??? (in /usr/lib/x86_64-linux-gnu/libmbim-glib.so.4.6.0)
==15994== by 0x4FAF287: ??? (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4FAEC03: g_main_context_dispatch (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x5003497: ??? (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
==15994== by 0x4FAE282: g_main_loop_run (in /usr/lib/x86_64-linux-gnu/libglib-2.0.so.0.7102.0)
We cannot call peek_device() when we have already created a GTask for
the async operation, because if the operation failed we would be
completing the async operation with a totally different GTask than the
one we had created ourselves, triggering memory leaks and leaving
modem references around that should not exist (as the GTask we create
holds a modem reference).
Issue found in the context of debugging this, may actually be fully
related:
https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/550
When initializing the Modem interface, we set the signal strength
to the default value (0%) and we are certain this is the recent value
until the modem reports the actual signal strength if available.
Therefore, mark the signal strength as recent instead of cached.
During mm logging, some of the information like simIccId, Telephone
numbers need to be hidden from displaying in the logs to protect
some of the user information.
Implemented for MBIM requiring libmbim 1.27.6, which is the
development version that includes the needed API.
g_date_time_new, and g_date_time_new_utc return NULL if inputs are out
of range, and currently mm_new_iso8601_time passes the GDateTime created
by those two functions to date_time_format_iso8601 without checking for
NULL values, causing a g_date_time_format_iso8601 crash if PDU data is
corrupted with wrong date.
To prevent this, mm_new_iso8601_time now can return NULL and set a new
GError if GDateTime created by g_date_time_new is NULL.
Fixes#546
When the modem sends an error which is outside the range
defined in MBIM_NW_ERROR list then MM should use a default
error(MBIM_NW_ERROR_NONE) instead of crash.
The user or the packages can install custom dispatcher scripts that
will be run by the ModemManaer daemon itself upon bearer connection
status updates.
Scripts installed by users should go into the
/etc/ModemManager/connection.d directory.
Scripts installed by packages should go into the
/usr/lib/ModemManager/connection.d directory.
The name of the script will specify the order how it's run, so it's
suggested to name them with a numeric prefix, e.g. "99-something".
When the user provides the profile parameter, as per the MM documentation
profile-id is an optional parameter. So with the fix the profile-id
check in input is removed. apn-type is also optional when the user
provides the settings. This check is also removed as part of the fix.
The BUFFER_FULL signal handler is effectively working in the same way
as the response buffer processor, in both cases we may have scheduled
the completion of the serial command, and that in turn may end up
fully disposing the port object. We must make sure the port object is
valid for as long as we need it in this function, so we take a
reference while processing the response buffer.
Fixes https://gitlab.freedesktop.org/mobile-broadband/ModemManager/-/issues/535
If no DMS capabilities are reported, we fall back to the capabilities
reported by NAS SSP or NAS TP.
Signed-off-by: Sven Schwermer <sven.schwermer@disruptive-technologies.com>
Alignment padding is not accounted for in the netlink attribute
length. Fix to avoid warnings later if we add shorter fixed
length attributes, even though it currently makes no difference
since we only add a 4 byte attribute
Signed-off-by: Bjørn Mork <bjorn@mork.no>
This fixes the missing NOARP flag on the second and following VLAN
sub-interfaces when multiplexing over the cdc_mbim driver.
ifi_change specifies all the flags we want to change. Any listed
flag which isn't set in ifi_flags will be cleared by the kernel:
if (ifm->ifi_change)
flags = (flags & ifm->ifi_change) |
(rtnl_dev_get_flags(dev) & ~ifm->ifi_change);
Drivers like cdc_mbim set some netdev flags by default, indicating
supported or unsupported features. The driver emulates an ethernet
netdev but it doesn't implement ARP. Clearing NOARP on cdc_mbim
netdevs will block IP transmissions, waiting for an ARP reply we
never will receive.
Signed-off-by: Bjørn Mork <bjorn@mork.no>
foxconn modules get firmware version:
Try with the new FOX service first. If FOX service failed, try with DMS service.
T99W175(vid:105b) ; T77W968(0489:e0b4 ; 0489:e0b5):
qmi service: dms
T99W265(0489:e0da ; 0489:e0db):
qmi service: fox
Recent ModemManager versions no longer work on the Samsung A5 (Qualcomm
MSM8916) smartphone: MM ends up in an infinite loop trying to probe the
modem over and over again.
Looking at the debug log it seems to be related to a "Slot Status"
indication that comes in during probe time and causes MM to restart
the probing process:
[/dev/wwan0qmi0] Sent generic request (translated)...
<< QMUX: service = "uim"
<< QMI: message = "Register Events" (0x002E)
<< TLV: translated = physical-slot-status
[/dev/wwan0qmi0] Received generic indication (translated)...
<< QMUX: service = "uim"
<< QMI: message = "Slot Status" (0x0048)
<< TLV: translated = { [0] = '[ physical_card_status = 'present', ...
[/dev/wwan0qmi0] Received generic response (translated)...
<< QMUX: service = "uim"
<< QMI: message = "Register Events" (0x002E)
<< TLV: translated = SUCCESS
[modem42] registered for slot status indications
[/dev/wwan0qmi0] Sent generic request (translated)...
<< QMUX: service = "uim"
<< QMI: message = "Get Slot Status" (0x0047)
[modem42] received slot status indication
[modem42] An active slot had a status change, will reprobe the modem
<infinite loop>
The "Slot Status" indication (uim_slot_status_indication_cb) arrives
immediately after starting to listen for the event, which is before
ModemManager has actually loaded the initial slot status in
uim_get_slot_status_ready. In this situation !slot_array_status_equal()
will always cause the modem to reprobe, even though it is not necessary
during the probe phase.
To fix this add a simple check to uim_slot_status_indication_cb() that
ignores the indication if the initial slot status was not loaded yet.
The numeric fields in the +COPS=? response were relying on a very weak
parsing logic, assuming that they were single-digit numeric values and
not using the common string to integer conversion utilities.
This commit improves the conversion from the 3GPP/ETSI defined network
availability and access technology values to the MM defined ones,
providing enum-based matches even if the numeric values are the same.
The commit also fixes the parsing of access technology values > 10,
required to report 5G related values.
+CGDCONT? may list profiles with IDs that are illegal to write, i.e.
+CGDCONT=? returns a minimum ID larger than some of the existing
profiles. E.g. for Fibocom L610-EU, +CGDCONT=? returns
+CGDCONT: (1-7),"IP",,,(0-3),(0-4)
+CGDCONT: (1-7),"IPV6",,,(0-3),(0-4)
+CGDCONT: (1-7),"IPV4V6",,,(0-3),(0-4)
+CGDCONT: (1-7),"PPP",,,(0-3),(0-4)
+CGDCONT: (1-7),"Non-IP",,,(0-3),(0-4)
while the default EPS bearer is established at profile 0:
+CGDCONT: 0,"IP","xxx","xxx.xxx.xxx.xxx",0,0
[...]
Signed-off-by: Sven Schwermer <sven.schwermer@disruptive-technologies.com>
meson generates the `mm-daemon-enums-types` source and header files.
These are used when building `ModemManager` executable and different
plugins. However, these enums are only stated as dependencies on the
`ModemManager` executable build.
This has been fixed by also adding the generated files targets as
dependencies in the required plugins.
The `broadmobi`, `dlink`, `telit` and `tplink` plugins include the
`mm-port-enums-types.h` header. However, they do not use any symbol
defined there.
The `huawei` plugin as includes the `mm-port-enums-types.h` header
but it does not include the build targets as dependencies.
These issues have been fixed by removing the unnecessary includes
from `broadmobi`, `dlink`, `telit` and `tplink` plugins and by
including the enums build target in the `huawei` target.
Jack Song
Aleksander Morgado
Prakash Pabba
Dylan Van Assche
Maxim Anisimov
som
Carlo Lobrano
Sven Schwermer
Andrew Lassalle
Bjørn Mork
Freedom Liu
Stephan Gerhold
Iñigo Martínez