Depending on the transport used on the data connection we want either
the "atmodem" (PPP) driver or the "ubloxmodem". For the "ubloxmodem",
we want to pass the model data so this patch wrangles some parameters to
make sure that right driver and right variant data are passed.
ttyACM0 (USB interface 02) is reportedly unreliable (breaking DHCP setup)
so the recommended approach is to use ttyACM2 (USB interface 06)
exclusively.
Some aspects of a device are detectable at runtime, like the USB profile
detection that was added in a patch preceding this one. This patch
switches the driver over from creating a new "vendor id" for each
profile to just setting a flag. This is more easily extensible as we
detect other features of the modem.
The TOBY L2 series of modems presents a number of different
configurations with different throughtput characteristics. These
configurations are packaged up as USB profiles; moreover, changing the
profile actually changes the USB model ID so this is even more like
selecting a different "device" altogether. Nonetheless, all we need to
know is which profile is selected in order to set things up correctly
and this can be queried directly.
This patch adds a call to UUSBCONF for applicable modems in order to
query the USB configuration to find out which profile is active.
This patch adds a call to CGMM into the modem_enable path in order to
establish the specific device model. From this device model string, a
model-specific capabilities structure can be selected.
Many ublox modems can sit on either the USB bus or talk directly to a
UART. The udev plugin mostly takes care of figuring out what ports to
talk to and the protocol is common for all devices after that.
This patch simplifies the setup a bit:
i) There must always be an aux channel for communication with the modem
ii) The aux channel may be found behind the string Aux for USB modems
or Device for serial modems
iii) If the Modem string is set, use it; if not set, assume it's not
available.
when the voice call driver is probing, I noticed sometimes
a CME Error with the code = 4 (operation not supported)
related to AT commands "AT+CSSN=1,1" and "AT^SLCC=1".
In the current voice call implementation for Gemalto modems
the ofono_voicecall_create() is called on the gemalto_pre_sim().
The sim is not totaly readed by the module, it is not yet
online which is causing a sporadic problem at driver
intialization also when launching some outgoing/incoming calls.
This change moves the ofono_voicecall_create() to the
gemalto_post_online() to be sure that sim is well registred
and online. Several other call related atoms are also moved to
post_online state.
The plugin for Gemalto modems is enriched with all voice-related atoms,
as well as USSD.
All except the voicecall itself are from the atmodem, while the
voicecall is from gemaltomodem.
There are a large number of files in the tree that define _GNU_SOURCE
despite not actually using features hidden behind this flag. This patch
removes all these definitions in one fell swoop...
support of 'none' in mbpi:
the default method remains CHAP, but it is overridden by NONE after
parsing the entire key for the apn and detecting no username/password
The sim atom is now created with the GEMALTO vendor instead of
CINTERION. This is because GEMALTO has superceeded CINTERION and the
gemalto plugin will be updated to handle (legacy) modems from cinterion
as well as current gemalto devices.
In some case linux report 'driver' as valid yet vid and pid as NULL.
Adding NULL check to prevent seg fault.
Log:
ofonod[23829]: plugins/udevng.c:udev_start()
ofonod[23829]: plugins/udevng.c:enumerate_devices()
ofonod[23829]: plugins/udevng.c:check_usb_device() hub [1d6b:0002]
ofonod[23829]: plugins/udevng.c:check_usb_device() usb [1d6b:0002]
ofonod[23829]: plugins/udevng.c:check_usb_device() usbhid [03f0:034a]
ofonod[23829]: plugins/udevng.c:check_usb_device() usbhid [03f0:034a]
ofonod[23829]: plugins/udevng.c:check_usb_device() usb [1d6b:0002]
ofonod[23829]: plugins/udevng.c:check_usb_device() cdc_acm [(null):(null)]
ofonod[23829]: Aborting (signal 11) [./src/ofonod]
Since the merge of udev.c into udevng.c all cleanup function must handle
both usb devices and serial devices. Add this distinction to
check_remove(), so that is doesn't try to iterate the .serial member as
if it were a .devices list.
The return value from ofono_modem_register was not being checked. If this fails
the modem object is not setup and causes a crash. This was specifically seen
when using the mbim driver without having configured with mbim support.
Now the modem object gets destroyed properly if the modem registration fails.
Add support for the SIM7100E modem. We add a new "sim7100" plugin
because there's no other AT-based modem that has support for both GPRS
and voice calls.
This atom needs to be created in post_sim so that the APN can be
written to the default profile before the modem attempts to use the
setting to connect to the network.
In SimManager, the Retries property isn't used for gemalto modems.
The at command AT^SPIC is used to get the remaining retries left
for the current required password type.
This commit enable the use of the driver in the gemalto plugin
The nokia-gpio plugin should not try to create symlinks to relevant gpio
pins under /dev/cmt, since the location it is looking is no longer
correct on newer kernels and it might change again in the future. This
patch removes code from nokia-gpio that tries to create a symlink.
Users will now need to symlink the modem gpios to /dev/cmt themselves.
On the 4.13 kernel, this can be done by, for example, adding a udev rule
to:
# ln -sf /sys/bus/hsi/devices/n900-modem /dev/cmt
The HUP results in errors in gatio which will deref parts
of the AT channel. This makes it impossible to recover and
send further AT commands after the HUP.
This is needed to find out the maximum fragment size that this device
supports. The minimum length specified in the spec is 64, however 512
bytes is recommended.
We also attempt to parse how many simultaneous commands can be
outstanding at a given time. In theory MBIM supports multiple
outstanding commands through the use of transaction-id matching, but it
seems some devices only support a single outstanding command at a time.
This patch fixes up some fallout from the merging of the udev and udevng
modules.
i) The 'serial' modems in question are not necessarily serial modems at
all; for example, the N900 ISI modems appear as network devices and don't
have a devnode. This patch relaxes the requirement that a devnode exist.
ii) The modem driver was being set to 'legacy', which is a remnant of
an earlier approach to merging these modules. The driver needs to be
properly set to the value of the OFONO_DRIVER property in order for the
setup function to be found.
iii) The serial modem's private data was being added to the modem
structure incorrectly...
In particular point iii) above shows that there is essentially nobody
using these devices which makes modifications in this area tricky.
Jonas Bonn
Antara Borwankar
Pičugins Arsenijs
Khaled Romdhani
Denis Kenzior
Nandini Rebello
Clement Viel
Giacinto Cifelli
Anirudh Gargi
Pavel Machek
Martin Hundebøll
Christophe Ronco
Varun Gargi
James Prestwood
Bob Ham
Mark van der Putten
Jonas Bonn
Bassem Boubaker
Mariem Cherif
Gabriel Lucas
Slava Monich
Alexander Couzens
Ankit Navik
Clayton Craft
Richard Röjfors
Joey Hewitt
Marcel Holtmann
Vincent Cesson