Ports flagged as being QCDM will not be probed for AT. QCDM probing
will be performed normally, the udev tags only give hints, we still
need to know whether the TTYs are QCDM-capable or not.
We keep the pflags input in mm_base_modem_grab_port() so that plugins
can use other methods to gather port type hints (e.g. querying with AT
commands as in Huawei/Telit or looking at sysfs properties as in HSO).
For standard udev tag port type hints, it will be the base modem
looking them up.
Note that there is no longer any need to ignore non-flagged ports for
those modems that require primary/secondary flags. They will be
implicitly ignored when mm_base_modem_organize_ports() decides which
ports to use, as the flagged ones are preferred over the non-flagged
ones.
Ports flagged with one of the common AT port type hints will not be
probed for QCDM.
AT probing will be performed normally, the udev tags only give hints,
we still need to know whether the TTYs are AT-capable or not.
We define 3 common udev tag ids to be used by all plugins:
* ID_MM_PORT_TYPE_AT_PRIMARY: the primary modem port. It will be used
for AT control and also as PPP if there is no other port flagged
explicitly to do PPP.
* ID_MM_PORT_TYPE_AT_SECONDARY: the secondary modem port. It will be
used when/if the primary port gets connected to do PPP.
* ID_MM_PORT_TYPE_PPP: the port to be used to do PPP only. This tag
makes sense only when the primary port shouldn't be used for PPP,
i.e. when there is a port dedicated to do PPP and one port
dedicated for control.
Unfortunately, G_MINDOUBLE is basically 0.0, so MM_SIGNAL_UNKNOWN ends
up giving us a value that may fall in the range of expected values for
the signal component.
Update the MM_SIGNAL_UNKNOWN symbol to match a value which is
definitely out any other possible valid range, so that we can easily
detect which values are set and which aren't.
While API is maintained, this fix is introducing an ABI break. Not a
big deal anyway, as the purpose of the value is just to detect unset
fields.
The "location area code" field is given in GSM/UMTS networks
exclusively. LTE networks use the concept of "tracking area code"
instead.
This patch updates the Location interface to Provide separate fields
for LAC and TAC, instead of giving TAC values in the LAC field.
Setting bands is a very device-specific operation. Sometimes the
device requires specific band combinations, or sometimes the 'any'
specific logic doesn't apply to all supported bands (e.g. may apply
only to the currently selected modes, as in XMM based devices).
So, don't assume that if the set command succeeds we have set all
expected bands. Instead, do an explicit loading of the current bands
after the set operation, same thing as we do when setting modes.
Also bump required libqmi version, as this change uses flags from the
latest available API.
UMTS B19 is NTT DoCoMo's 800MHz band.
UMTS B6 is NTT DoCoMo's 850MHz band
When we detect that the modem is QMI-capable or MBIM-capable, we still
want to be able to use TTYs, for features unsupported by the main
protocols.
So, don't flag all the TTYs as non-AT non-QCDM, let them probe as
usual instead.
The first step in the Enable() processing is to wait for a final
state, so we can definitely wait for the ongoing disabling or
initializing states to finish before we go on with the enable
operation, there is no need to early fail if the disabling or
initializing intermediate states are detected.
Being a part of a series that adds PCO support for MBIM modems that
implement the MBIM_CID_PCO extension, this patch issues a
MBIM_CID_DEVICE_SERVICES query during the modem initialization to check
if a modem implements MBIM_CID_PCO. If the modem does, ModemManager will
subscribe to MBIM_CID_PCO notifications to get PCO updates.
E.g. if the modem switches from 4G to 3G while the extended signal
information is enabled, we should no longer expose LTE specific signal
quality values, only the UMTS specific ones.
E.g. to avoid this:
$ mmcli -m 1 --signal-get
-------------------------
UMTS | RSSI: '0,00' dBm
| RSCP: '-92,00' dBm
| EcIo: '-13,00' dB
-------------------------
LTE | RSSI: '0,00' dBm
| RSRQ: '-6,50' dB
| RSRP: '-96,00' dBm
| SNR: '0,00' dB
The sort_band() method used in the tester was totally wrong, it was
comparing the addresses of the variables instead of the MMModemBand
values.
Use the common mm_common_bands_garray_sort() instead, which works as
expected.
The default AT^SCFG="Radio/Band" value for Cinterion PLS8-J devices is
"16819472". Add UMTS band 19 and LTE band 19 entries based on the
information given in the PLS8 datasheet.
To have proper ordering in the D-Bus signals, the skeleton's property
changes must be flushed before the Call{Add,Delet}ed signals are
emitted. Without this flush, the emission of the PropertiesChanged
signal is delayed until the main loop is idle. This causes problems
on the client side, for example the CallAdded signal being received
before the Calls property contains the call.
Closes: #81
This info comes from PLS8-X/E/J/V/US, HC25 & PHS8 references, the
last two can be found publicly via Google search.
Swapped bit-mask locations for G850 & PCS bands as they may have
changed with FW or where accidently put in the wrong place.
Updated many 3G & 4G bit-mask fields.
Not all modems support AT command concatenation (e.g. u-blox TOBY-L4
doesn't, according to a specific note about it in the AT command
reference). So just skip concatenation in the few places where it's
used.
The TOBY-L2 allowed to skup the user/pass string fields when no
authentication was requested, but according to the AT command
reference, all the remaining u-blox modules do require these two
fields given always (e.g. just as empty strings).
As per this sequence in a TOBY-L4:
(ttyACM2): --> 'AT+UAUTHREQ=?<CR>'
(ttyACM2): <-- '<CR><LF>+UAUTHREQ: (1-4),(0-2),,<CR><LF><CR><LF>OK<CR><LF>'
(ttyACM2): --> 'AT+UAUTHREQ=1,0<CR>'
(ttyACM2): <-- '<CR><LF>+CME ERROR: 4<CR><LF>'
It should have been: AT+UAUTHREQ=1,0,"",""
If we're going to overwrite the timezone context, make sure everything
is cleaned up, including the signal handler. Otherwise, we may end up
running the signal handler without a context attached:
(gdb) bt
#0 network_timezone_state_changed (self=0x6aa1e8) at mm-iface-modem-time.c:266
#1 0x76a8d418 in g_closure_invoke (closure=0x6c9810, return_value=0x0, return_value@entry=0x1, n_param_values=2, param_values=0x7edb58c0, param_values@entry=0x6cdab8,
invocation_hint=invocation_hint@entry=0x7edb585c) at gclosure.c:804
#2 0x76a9e674 in signal_emit_unlocked_R (node=node@entry=0x6794d0, detail=0, detail@entry=391, instance=instance@entry=0x6aa1e8, emission_return=0x6794a8, emission_return@entry=0x0,
instance_and_params=instance_and_params@entry=0x7edb58c0) at gsignal.c:3629
#3 0x76aa31f4 in g_signal_emit_valist (instance=instance@entry=0x6aa1e8, signal_id=signal_id@entry=1, detail=detail@entry=6767564, var_args=..., var_args@entry=...) at gsignal.c:3385
#4 0x76aa34ec in g_signal_emit (instance=instance@entry=0x6aa1e8, signal_id=signal_id@entry=1, detail=391) at gsignal.c:3441
#5 0x76a91698 in g_object_dispatch_properties_changed (object=0x6aa1e8, n_pspecs=1, pspecs=<optimized out>) at gobject.c:1062
#6 0x76a90e7c in g_object_notify_queue_thaw (object=object@entry=0x6aa1e8, nqueue=nqueue@entry=0x75409418) at gobject.c:296
#7 0x76a9470c in g_object_set_valist (object=object@entry=0x6aa1e8, first_property_name=first_property_name@entry=0xad566 "iface-modem-state", var_args=..., var_args@entry=...) at gobject.c:2171
#8 0x76a94b18 in g_object_set (_object=_object@entry=0x6aa1e8, first_property_name=0xad566 "iface-modem-state") at gobject.c:2275
#9 0x0004b8a4 in __iface_modem_update_state_internal (self=0x6aa1e8, new_state=MM_MODEM_STATE_DISABLED, reason=MM_MODEM_STATE_CHANGE_REASON_USER_REQUESTED, failed_reason=MM_MODEM_STATE_FAILED_REASON_NONE)
at mm-iface-modem.c:1487
#10 0x0007047c in disabling_context_free (ctx=0x6ba850) at mm-broadband-modem.c:9053
#11 0x76b4e538 in g_task_finalize (object=0x682488) at gtask.c:636
#12 0x76a92800 in g_object_unref (_object=0x682488) at gobject.c:3183
#13 0x00072154 in iface_modem_disable_ready (self=0x6aa1e8, result=0x7011b0, task=0x682488) at mm-broadband-modem.c:9101
#14 0x76b4e6ec in g_task_return_now (task=0x7011b0) at gtask.c:1107
#15 0x76b4e720 in complete_in_idle_cb (task=0x7011b0) at gtask.c:1121
#16 0x7694d144 in g_main_dispatch (context=0x6765a8) at gmain.c:3154
#17 g_main_context_dispatch (context=context@entry=0x6765a8) at gmain.c:3769
#18 0x7694d40c in g_main_context_iterate (context=0x6765a8, block=block@entry=1, dispatch=dispatch@entry=1, self=<optimized out>) at gmain.c:3840
#19 0x7694d838 in g_main_loop_run (loop=0x6750d0) at gmain.c:4034
#20 0x0002bec0 in main (argc=<optimized out>, argv=<optimized out>) at main.c:181
The problem of full vendor blacklist are hubs.
Because ATTRS{} matches all devices in the tree,
a modem connected to a OpenMoko hub will be blacklisted as well.
==8663== 1 bytes in 1 blocks are definitely lost in loss record 5 of 4,864
==8663== at 0x4C2CEDF: malloc (vg_replace_malloc.c:299)
==8663== by 0x669FAC9: g_malloc (in /usr/lib/libglib-2.0.so.0.5600.1)
==8663== by 0x66B9AEA: g_strndup (in /usr/lib/libglib-2.0.so.0.5600.1)
==8663== by 0x66ACE52: g_match_info_fetch (in /usr/lib/libglib-2.0.so.0.5600.1)
==8663== by 0x4E96DD2: mm_location_gps_raw_add_trace (mm-location-gps-raw.c:208)
==8663== by 0x18721E: mm_iface_modem_location_gps_update (mm-iface-modem-location.c:264)
==8663== by 0x1B9C7F: loc_location_nmea_indication_cb (mm-shared-qmi.c:712)
==8663== by 0x640C9E7: g_cclosure_marshal_VOID__BOXEDv (in /usr/lib/libgobject-2.0.so.0.5600.1)
==8663== by 0x6409C95: ??? (in /usr/lib/libgobject-2.0.so.0.5600.1)
==8663== by 0x64259E8: g_signal_emit_valist (in /usr/lib/libgobject-2.0.so.0.5600.1)
==8663== by 0x642612F: g_signal_emit (in /usr/lib/libgobject-2.0.so.0.5600.1)
==8663== by 0x54AAB2D: process_indication (qmi-loc.c:7454)
Reported by: Thomas Weißschuh <thomas@weissschuh.net>
ModemManager[18223]: <debug> [1530023432.028473] Modem has 3GPP capabilities, enabling the Modem 3GPP interface...
Thread 1 "ModemManager" received signal SIGSEGV, Segmentation fault.
0x0000555555618baa in common_enable_disable_unsolicited_events_signal_strength (task=0x5555558e4440) at mm-broadband-modem-qmi.c:6545
6545 if (ctx->enable)
(gdb) bt
#0 0x0000555555618baa in common_enable_disable_unsolicited_events_signal_strength (task=0x5555558e4440) at mm-broadband-modem-qmi.c:6545
#1 0x0000555555618d70 in common_enable_disable_unsolicited_events (self=0x55555597a3f0, enable=1, callback=0x5555555c8c20 <enable_unsolicited_events_ready>, user_data=0x5555559779b0)
at mm-broadband-modem-qmi.c:6719
#2 0x0000555555618e5b in modem_3gpp_enable_unsolicited_events (self=0x55555597a3f0, callback=0x5555555c8c20 <enable_unsolicited_events_ready>, user_data=0x5555559779b0) at mm-broadband-modem-qmi.c:6749
#3 0x00005555555c9095 in interface_enabling_step (task=0x5555559779b0) at mm-iface-modem-3gpp.c:1907
#4 0x00005555555c8c0a in setup_unsolicited_events_ready (self=0x55555597a3f0, res=0x5555558efad0, task=0x5555559779b0) at mm-iface-modem-3gpp.c:1795
#5 0x00007ffff681e204 in () at /usr/lib/libgio-2.0.so.0
#6 0x00007ffff681e239 in () at /usr/lib/libgio-2.0.so.0
#7 0x00007ffff62731d6 in g_main_context_dispatch () at /usr/lib/libglib-2.0.so.0
#8 0x00007ffff62735b1 in () at /usr/lib/libglib-2.0.so.0
#9 0x00007ffff62738e2 in g_main_loop_run () at /usr/lib/libglib-2.0.so.0
#10 0x000055555559182c in main (argc=2, argv=0x7fffffffe4a8) at main.c:181
Fixes: baefe53ab9
After the tty is opened, the fd is passed off to a GIOChannel which
manages it. When the serial port is closed with _close_internal(),
g_io_channel_shutdown() is called which then calls g_io_unix_close() which
in turn close()s the fd.
_close_internal() performs a second explicit close() after the
g_io_channel_shutdown() which *in the best case scenario* will fail with
EBADF.
This is a race condition that can result in other file descriptors being
closed.
This change avoids double closing the serial port fd -- the shutdown() call
is removed in favour of the explicit close().
GVfs' libgvfsdbus.so GIo module, when automatically loaded, attempts to
discover if there's a session D-Bus instance to use. It tries real hard
to get the socket name it would use -- in absence of XDG_RUNTIME_DIR it
decides to make it up with user's home directory. When HOME is unset
too, it just tries to figure it out by looking into /etc/passwd. Which
upsets SELinux that would better not see us looking into it.
We trigger the load of the GIo modules, by using the GFile API to access
the the ports in /dev. They're utterly uesless to us, but there doesn't
seem to be a way to disable their load. Oh well.
For now, let's just ensure HOME is set and the problematic path in glib
is not taken.
