Faster resume probing

Sat Jan 23 23:10:57 UTC 2021

Hi,

> So with these your modem always stays connected over USB? How does ModemManager react to this when resuming/suspending? ModemManager still reprobes the device right?
Yes, the modem remains connected over USB, even over many
suspend/resume cycles. MM doesn't reprobe upon resume. My
understanding is that MM is completely unaware that a suspend/resume
has occurred - hence it doesn't feel the need to reprobe upon resume.
On your PinePhone, when it comes out of suspend does the USB system
re-enumerate the modem? (e.g. do you get the USB device connected &
qmi/option driver kernel messages, or a message like 'usb1-1: reset
full speed device...'). If so, then I'd look into that first.

If possible, I'd start by getting a USB keyboard and make sure it can
wake up the host (and not re-initialize or reset the driver each
time). This will depend on whether the PinePhone powers it's OTG port
during suspend, but if it does then I think that is a much better way
to dig into a USB suspend issue.

> PinePhone uses the RING indicator of the EG25-G modem which gets triggered when an URC or something is available.
Makes sense, though I naively suspect it's somewhat of a crutch (at
least when the USB connection is up) as it feels like these events
could simply come via USB - assuming USB is properly set up for
suspend/resume. However I see no reason why both interrupt sources
can't coexist.

> We use indeed systemctl suspend of elogind to suspend the system.
Looking at your other replies, it looks like you're aware of the
systemd suspend/resume hooks that MM has by default. I'd experiment
with using /sys/power/state too, to see if there are other suspend
hooks being a bit trigger-happy.

Regards,
Tom

On Sun, 24 Jan 2021 at 05:12, Dylan Van Assche <me at dylanvanassche.be> wrote:
>
> Hi,
>
>
> > I've also been working with USB QMI modems (Quectel and Simcom) and
> > getting suspend-to-ram/resume working smoothly, so I figured perhaps
> > I
> > can share some of the things I've learnt along the way. But do note
> > that I'm not by any means an expert here - this is my first foray
> > into
> > usb & modems, so do take what I say with a grain of salt.
> >
> > Firstly, I agree with Aleksander's opinion above - I think the
> > underlying problem is the fact that the USB connection is being reset
> > upon resume. I'd try to solve that first - disable MM & NM altogether
> > for now, and focus on the qmi & usb layers. Use qmicli to set up a
> > connection or at least check basic info with commands like
> > qmicli --device=/dev/cdc-wdm0 --device-open-proxy --dms-get-model
> > Then start looking at the USB side of things. The kernel docs
> >
> > https://www.kernel.org/doc/html/v4.14/driver-api/usb/power-management.html
> > provide some good background info on this area.
> >
> > If I had a pinephone or similar hardware to test with I might be able
> > to provide some more guidance, however I can perhaps shed some light
> > by showing what was required to get my particular device working.
> > My device is an IMX6ULL, and the modem is connected directly to the
> > USB host pins on the chip itself. No USB hub ICs involved. Power to
> > the modem is controlled externally (in my case it's always on).
>
> Sounds like the PinePhone :)
> The EG25-G modem is also continously powered on and directly connected
> to the A64 USB. Audio is routed over I2S directly in the hardware.
> I will have a look at your suggestion on the PinePhone :)
>
> > USB info:
> >     root at imx6ul-var-dart:~# lsusb -tv
> >     /:  Bus 02.Port 1: Dev 1, Class=root_hub, Driver=ci_hdrc/1p, 480M
> >         ID 1d6b:0002 Linux Foundation 2.0 root hub
> >     /:  Bus 01.Port 1: Dev 1, Class=root_hub, Driver=ci_hdrc/1p, 480M
> >         ID 1d6b:0002 Linux Foundation 2.0 root hub
> >         |__ Port 1: Dev 2, If 0, Class=Vendor Specific Class,
> > Driver=option, 480M
> >             ID 2c7c:0125 Quectel Wireless Solutions Co., Ltd. EC25
> > LTE modem
> >         |__ Port 1: Dev 2, If 1, Class=Vendor Specific Class,
> > Driver=option, 480M
> >             ID 2c7c:0125 Quectel Wireless Solutions Co., Ltd. EC25
> > LTE modem
> >         |__ Port 1: Dev 2, If 2, Class=Vendor Specific Class,
> > Driver=option, 480M
> >             ID 2c7c:0125 Quectel Wireless Solutions Co., Ltd. EC25
> > LTE modem
> >         |__ Port 1: Dev 2, If 3, Class=Vendor Specific Class,
> > Driver=option, 480M
> >             ID 2c7c:0125 Quectel Wireless Solutions Co., Ltd. EC25
> > LTE modem
> >         |__ Port 1: Dev 2, If 4, Class=Vendor Specific Class,
> > Driver=qmi_wwan, 480M
> >             ID 2c7c:0125 Quectel Wireless Solutions Co., Ltd. EC25
> > LTE modem
> >
> > The following 3 tweaks are needed to get the modem persisting across
> > a
> > suspend/resume:
> >    echo auto > /sys/bus/usb/devices/1-1/power/control
> >    echo enabled > /sys/bus/platform/devices/2184000.usb/power/wakeup
> >    echo enabled > /sys/bus/platform/devices/ci_hdrc.0/power/wakeup
> >
>
> So with these your modem always stays connected over USB?
> How does ModemManager react to this when resuming/suspending?
> ModemManager still reprobes the device right?
>
> > Obviously your USB setup will be different so the above won't work,
> > but perhaps it might provide some direction. The first one - setting
> > control to auto - makes sense to me, but I cannot see why the other
> > two are needed. Having a good poke around the /sys/ tree is a good
> > start anyway - set control=auto and wakeup=enabled on all the usb
> > devices on/above the modem. Finding the equivalent of my 2nd line
> > might be trickier and need some datasheet or device-tree reading to
> > correctly identify it.
> > If you have USB hub devices in the chain then things will be more
> > complex, but I'd be surprised if they've set up the hardware that
> > way.
> >
> > Finally, to allow the modem to wakeup the host, the following two
> > tweaks were needed:
> >     echo enabled > /sys/bus/usb/devices/usb1/power/wakeup
> >     echo enabled > /sys/bus/usb/devices/1-1/power/wakeup
> >
>
> PinePhone uses the RING indicator of the EG25-G modem which gets
> triggered when an URC or something is available. This RING indicator is
> connected to a interrupt capable GPIO pin of the A64 host CPU.
> As soon as the RING indicator is triggered, the host CPU will wake up,
> even without any USB connection.
>
> > Also, note I'm suspending via the very raw
> >    echo mem > /sys/power/state
> > so I can be sure that no setup/teardown is occuring. But obviously
> > pm-suspend or systemctl suspend is the right way to do things for
> > production units.
> >
>
> We use indeed systemctl suspend of elogind to suspend the system.
>
> Kind regards,
> Dylan Van Assche
>
>