[LEDE-DEV,0/3] uqmi: fix more dial-in issues

This series fixes more dial-in issues discovered during debug sessions.

Testing and debugging was performed on following modems:
> Quectel EC25-A
> Sierra Wireless EM7565
> Sierra Wireless MC7455
> Huawei ME909u

3 different SIM cards used covering different operator networks

script changes were verified using shellcheck.net


Koen Vandeputte (3):
  uqmi: use general method for state cleaning
  uqmi: use correct value for connection checking
  uqmi: use built-in command for data-link verification

 .../utils/uqmi/files/lib/netifd/proto/qmi.sh       | 46 +++++++++++++++-------
 1 file changed, 31 insertions(+), 15 deletions(-)

On Wed,  7 Feb 2018 13:23:25 +0100
Koen Vandeputte <koen.vandeputte@ncentric.com> wrote:

> > Quectel EC25-A

Does the EC25 work without any kernel patch? I was testing EC25 with 17.01.1,
and it was working only with a kernel patch. Can you give me a status update
on this? Now I migrated to 17.01.4, but I haven't tested the qmi/ethernet
interface.

For the time being, I'm using it with the PPP interface.

What is your estimate on timing of your patch goes into a release?

Thanks,
Levente

On 07-02-18 22:14, Levente Kovacs wrote:
> On Wed,  7 Feb 2018 13:23:25 +0100
> Koen Vandeputte <koen.vandeputte@ncentric.com> wrote:
>
>>> Quectel EC25-A
> Does the EC25 work without any kernel patch? I was testing EC25 with 17.01.1,
> and it was working only with a kernel patch. Can you give me a status update
> on this? Now I migrated to 17.01.4, but I haven't tested the qmi/ethernet
> interface.
This modem only supports raw-ip mode for QMI, which was only added in 
kernel 4.5.
The 17.01 branch uses 4.4, while the master branch uses 4.9 for my 
target containing this modem (ar71xx - Gl.Mifi)

> For the time being, I'm using it with the PPP interface.
>
> What is your estimate on timing of your patch goes into a release?
afaik, the required patch(es) have not been backported yet.
If it's no too much work, I'm more than happy to provide a backport 
(it's a bit out of my scope as I'm mostly using Master)

I'll try to check it tomorrow and will let you know :)
> Thanks,
> Levente
>
> _______________________________________________
> Lede-dev mailing list
> Lede-dev@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/lede-dev

Koen Vandeputte <koen.vandeputte@ncentric.com> writes:
> On 07-02-18 22:14, Levente Kovacs wrote:
>> On Wed,  7 Feb 2018 13:23:25 +0100
>> Koen Vandeputte <koen.vandeputte@ncentric.com> wrote:
>>
>>>> Quectel EC25-A
>> Does the EC25 work without any kernel patch? I was testing EC25 with 17.01.1,
>> and it was working only with a kernel patch. Can you give me a status update
>> on this? Now I migrated to 17.01.4, but I haven't tested the qmi/ethernet
>> interface.
> This modem only supports raw-ip mode for QMI, which was only added in
> kernel 4.5.
> The 17.01 branch uses 4.4, while the master branch uses 4.9 for my
> target containing this modem (ar71xx - Gl.Mifi)
>
>> For the time being, I'm using it with the PPP interface.
>>
>> What is your estimate on timing of your patch goes into a release?
> afaik, the required patch(es) have not been backported yet.
> If it's no too much work, I'm more than happy to provide a backport
> (it's a bit out of my scope as I'm mostly using Master)

Backporting the raw-ip feature to v4.4 is simple.  Just cherry-pick

 81e0ce79f291 ("usbnet: allow mini-drivers to consume L2 headers")
 32f7adf633b9 ("net: qmi_wwan: support "raw IP" mode")
 6c730080e663 ("net: qmi_wwan: should hold RTNL while changing netdev type")


These still apply cleanly on top of v4.4.115.

The only reason the backport wasn't done upstream is because "raw-ip"
support is considered a new feature and therefore out-of-scope for
stable.

Note that you'll probable have to track down and add a number of device
IDs for the devices with raw-ip support too.  I haven't done this.

But running master with v4.9 or v4.14 sounds like a much better
alternative to me....



Bjørn

Bjørn Mork <bjorn@mork.no> writes:

> Backporting the raw-ip feature to v4.4 is simple.  Just cherry-pick
>
>  81e0ce79f291 ("usbnet: allow mini-drivers to consume L2 headers")
>  32f7adf633b9 ("net: qmi_wwan: support "raw IP" mode")
>  6c730080e663 ("net: qmi_wwan: should hold RTNL while changing netdev type")
>
>
> These still apply cleanly on top of v4.4.115.

Correction. A complete(?) backport should include these two fixes as well:

 a4abd7a80add ("usbnet: fix alignment for frames with no ethernet header")
 0de0add10e58 ("qmi_wwan: Add missing skb_reset_mac_header-call")


Hope I got them all now...




Bjørn

On 2018-02-08 10:53, Bjørn Mork wrote:
> Bjørn Mork <bjorn@mork.no> writes:
>
>> Backporting the raw-ip feature to v4.4 is simple.  Just cherry-pick
>>
>>   81e0ce79f291 ("usbnet: allow mini-drivers to consume L2 headers")
>>   32f7adf633b9 ("net: qmi_wwan: support "raw IP" mode")
>>   6c730080e663 ("net: qmi_wwan: should hold RTNL while changing netdev type")
>>
>>
>> These still apply cleanly on top of v4.4.115.
> Correction. A complete(?) backport should include these two fixes as well:
>
>   a4abd7a80add ("usbnet: fix alignment for frames with no ethernet header")
>   0de0add10e58 ("qmi_wwan: Add missing skb_reset_mac_header-call")
>
>
> Hope I got them all now...
>
Thank you Bjorn!
I'm happy to say I've used exactly these 5 patches this morning for 
backporting it.

For backporting support for the EC25 modem, following patches were used:

--> net: qmi_wwan: MDM9x30 specific power management
--> qmi_wwan: add support for Quectel EC21 and EC25

I'm currently testing the build on a Gl-Mifi device (ar71xx)

Koen
>
>
> Bjørn

hi ,

i have the same modem, with the same ec25 , please pay attention ,

if you put it int gprs mode or if you are in a enviroment where it fall 
back to gprs, the board (gl mifi)  is not designed to supply the correct 
current, then when it is going to attach to the network, the supply 
drops due to high current consumation and reboot

this is clearly stated in the ec25 documentations, they say about 1.5A

you can mitigate this behaviour adding a capacitor on the supply line , 
it doesn't resolve all the reboot , now it reboot only if i use to 
transmit continously


i've reported this problem to glinet about six month ago ....

bye


Il 07/02/2018 22:35, Koen Vandeputte ha scritto:
>
>
> On 07-02-18 22:14, Levente Kovacs wrote:
>> On Wed,  7 Feb 2018 13:23:25 +0100
>> Koen Vandeputte <koen.vandeputte@ncentric.com> wrote:
>>
>>>> Quectel EC25-A
>> Does the EC25 work without any kernel patch? I was testing EC25 with 
>> 17.01.1,
>> and it was working only with a kernel patch. Can you give me a status 
>> update
>> on this? Now I migrated to 17.01.4, but I haven't tested the 
>> qmi/ethernet
>> interface.
> This modem only supports raw-ip mode for QMI, which was only added in 
> kernel 4.5.
> The 17.01 branch uses 4.4, while the master branch uses 4.9 for my 
> target containing this modem (ar71xx - Gl.Mifi)
>
>> For the time being, I'm using it with the PPP interface.
>>
>> What is your estimate on timing of your patch goes into a release?
> afaik, the required patch(es) have not been backported yet.
> If it's no too much work, I'm more than happy to provide a backport 
> (it's a bit out of my scope as I'm mostly using Master)
>
> I'll try to check it tomorrow and will let you know :)
>> Thanks,
>> Levente
>>
>> _______________________________________________
>> Lede-dev mailing list
>> Lede-dev@lists.infradead.org
>> http://lists.infradead.org/mailman/listinfo/lede-dev
>
>
> _______________________________________________
> Lede-dev mailing list
> Lede-dev@lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/lede-dev