10:17 a.m.
Hi,
This patch is to enable Huawei EM770 modem, which is a full feature modem that supports
voicecall, phonebook, call forwarding, call barring, etc.
There are two known parts of code that may have problem and need your comments.
1. Once connecting this modem to PC via USB, we can get 5 tty ports: ttyUSB0, ttyUSB1,
ttyUSB2, ttyUSB3 and ttyUSB4. The correct way to talk to this modem is using ttyUSB2.
I'm not sure if I handled this correctly in file udev.c.
2. When creating the modem, a global modem id will be appended to the name of it while
constructing the path. This modem has name "huawei_em770", thus the path becomes
things like "/huawei_em7700" or "/huawei_em7701". This looks strange.
Need I change the name of this modem, or can we use some smarter way to append the modem
id?
There is also one known issue for this modem: The return of "CSMS=?" from this
modem is not compatible with spec. This is deemed as a modem issue, so the sms
wouldn't work before this hardware issue get solved.
Comments are welcome!
Regards,
-Yang
5:03 p.m.
Hi Yang,
On 10 February 2010 11:17, Gu, Yang <yang.gu(a)intel.com> wrote:
1. Once connecting this modem to PC via USB, we can get 5 tty ports:
ttyUSB0, ttyUSB1, ttyUSB2, ttyUSB3 and ttyUSB4. The correct way to talk to this modem is
using ttyUSB2. I'm not sure if I handled this correctly in file udev.c.
If there are more devices on the system that have a USB serial
interface the ttyUSBn numbers will be different so it would be more
reliable to decide based on some other characteristic of ttyUSB2 than
the name (don't know which, maybe you can retrieve USB
config/interface number). If there's nothing better, then maybe take
n = 2 + the lowest number of the five devices.
Regards,
10:37 p.m.
Hi Yang,
There are two known parts of code that may have problem and need
your
comments. 1. Once connecting this modem to PC via USB, we can get 5 tty
ports: ttyUSB0, ttyUSB1, ttyUSB2, ttyUSB3 and ttyUSB4. The correct way to
talk to this modem is using ttyUSB2. I'm not sure if I handled this
correctly in file udev.c. 2. When creating the modem, a global modem id
So Andrew had it exactly correct, you can't just use ttyUSB2 here. Most
devices have other udev attributes you can examine for what type of port it
is, to help you pick the right one. Perhaps examination of 'udevadm info --
attribute-walk --path=<device>' might be in order?
will be appended to the name of it while constructing the path. This
modem
has name "huawei_em770", thus the path becomes things like
"/huawei_em7700" or "/huawei_em7701". This looks strange. Need I
change
the name of this modem, or can we use some smarter way to append the modem
id?
We should be using the udev short serial from the device. For some reason
this seems to be failing on the EM770. Can you check whether get_serial
returns NULL for this device?
Regards,
-Denis
5:54 a.m.
Hi, Andrew and Denis,
-----Original Message-----
From: ofono-bounces(a)ofono.org [mailto:ofono-bounces@ofono.org] On Behalf Of
Denis Kenzior
Sent: Thursday, February 11, 2010 6:38 AM
To: ofono(a)ofono.org
Subject: Re: [PATCH] Add support of Huawei EM770 modem
Hi Yang,
> There are two known parts of code that may have problem and need your
> comments. 1. Once connecting this modem to PC via USB, we can get 5 tty
> ports: ttyUSB0, ttyUSB1, ttyUSB2, ttyUSB3 and ttyUSB4. The correct way to
> talk to this modem is using ttyUSB2. I'm not sure if I handled this
> correctly in file udev.c. 2. When creating the modem, a global modem id
So Andrew had it exactly correct, you can't just use ttyUSB2 here. Most
devices have other udev attributes you can examine for what type of port it
is, to help you pick the right one. Perhaps examination of 'udevadm info --
attribute-walk --path=<device>' might be in order?
The check of "ttyUSB2" in my code is not the same as you see in /dev directory.
Below is the output of this modem along its path (not a full path and I removed some
unimportant attributes in each path):
looking at device
'/devices/pci0000:00/0000:00:1a.7/usb1/1-3/1-3:1.2/ttyUSB2/tty/ttyUSB2':
(device1)
KERNEL=="ttyUSB2"
SUBSYSTEM=="tty"
DRIVER==""
looking at parent device
'/devices/pci0000:00/0000:00:1a.7/usb1/1-3/1-3:1.2/ttyUSB2': (device2)
KERNELS=="ttyUSB2"
SUBSYSTEMS=="usb-serial"
DRIVERS=="option1"
ATTRS{port_number}=="0"
looking at parent device '/devices/pci0000:00/0000:00:1a.7/usb1/1-3/1-3:1.2':
(device3)
KERNELS=="1-3:1.2"
SUBSYSTEMS=="usb"
DRIVERS=="option"
ATTRS{bInterfaceNumber}=="02"
ATTRS{modalias}=="usb:v12D1p1404d0000dc00dsc00dp00icFFiscFFipFF"
ATTRS{supports_autosuspend}=="0"
looking at parent device '/devices/pci0000:00/0000:00:1a.7/usb1/1-3':(device4)
KERNELS=="1-3"
ATTRS{configuration}=="Huawei Configuration"
ATTRS{bNumInterfaces}==" 5"
ATTRS{idVendor}=="12d1"
ATTRS{idProduct}=="1404"
ATTRS{manufacturer}=="HUAWEI Technology"
ATTRS{product}=="HUAWEI Mobile"
In oFono code, "device4" represents the modem, while "device1" is the
udev_device when calling function add_huawei_em770(modem, udev_device). When I check if
it's ttyUSB2 or not in my code, I think I'm checking the device1 and its sysname
(Showed as KERNEL in device1, but not the one showed in /dev directory), and in the
context of this modem (device4). I do believe there may be some other ttyUSB2 in the
system, for example, we have another em770 modem pluged in. But these two ttyUSB2 are in
different context. I ever tried to change the name showed in /dev directory by writing
some udev rules, but this sysname (KERNEL) seems stable (It's still ttyUSB2). Is it
possible this sysname gets changed under some circumstance expect the kernel or driver
changes? If not, I still think it's safe and convenient to check this value here.
Maybe we can use another attribute "bInterfaceNumber". But this attribute is
udev_device's parent's parent (device3), so the code could become more
complicated. What's your opinion?
> will be appended to the name of it while constructing the path. This modem
> has name "huawei_em770", thus the path becomes things like
> "/huawei_em7700" or "/huawei_em7701". This looks strange. Need I
change
> the name of this modem, or can we use some smarter way to append the modem
> id?
We should be using the udev short serial from the device. For some reason
this seems to be failing on the EM770. Can you check whether get_serial
returns NULL for this device?
Yes. get_serial() will return NULL, and we just use its type as name.
Regards,
-Denis
_______________________________________________
ofono mailing list
ofono(a)ofono.org
http://lists.ofono.org/listinfo/ofono
Regards,
-Yang
1:53 a.m.
On 11 February 2010 06:54, Gu, Yang <yang.gu(a)intel.com> wrote:
In oFono code, "device4" represents the modem, while
"device1" is the udev_device when calling function add_huawei_em770(modem,
udev_device). When I check if it's ttyUSB2 or not in my code, I think I'm checking
the device1 and its sysname (Showed as KERNEL in device1, but not the one showed in /dev
directory), and in the context of this modem (device4). I do believe there may be some
other ttyUSB2 in the system, for example, we have another em770 modem pluged in. But these
two ttyUSB2 are in different context. I ever tried to change the name showed in /dev
directory by writing some udev rules, but this sysname (KERNEL) seems stable (It's
still ttyUSB2). Is it possible this sysname gets changed under some circumstance expect
the kernel or driver changes? If not, I still think it's safe and convenient to check
this value here.
Honestly I don't know. But the whole /dev path may change between
kernel versions, for example I believe "ttyUSB" was "ttyU" a long
time
ago (I'm not 100% sure). So in my opinion the safe way is to compare
the "bInetrfaceNumber" and "configuration" of the parent devices to
"02" and "Huawei Configuration".
Best regards
3:23 a.m.
Hi,
This is the updated patch to guarantee the correct port is used by using
"bInterfaceNumber" attribute. Please review!
Regards,
-Yang
4356
days inactive
4370
days old
5 comments
3 participants
participants (3)
-
Andrzej Zaborowski -
Denis Kenzior -
Gu, Yang