modem disappears when waking for call

When a call is received during suspend, the modem will disappear immediately after the device wakes and before it rings. This is with 0.1.2, this did not occur with 0.1.1.

I'll try to capture some logging from eg25-manager when this happens over the next day or so when I have more time.

cc @dylanvanassche

As mentioned during the discussion, it could be the resume timeout kicking in too early: on my phone running Mobian it works fine, but for some reason maybe probing the modem takes longer under some circumstances; we could then reach the timeout just as the modem is being brought up by MM and therefore reset it.

Calibrating this timeout is tricky: it must kick in at a point where we're sure MM should have re-probed the modem; but if it's too long, then we could miss a call. In its current state, and based on my testing, it still allows the "incoming call" display to show for a couple seconds most of the time, so at least the user gets a (small) chance to answer the call, and (most importantly) it will be shown in the calls list. Making it longer would prevent that, and the user wouldn't even be aware of the missed call.

Anyway, a full eg25-manager log would help here (might also be an issue with MM vanishing D-bus as seen previously on @dylanvanassche 's device, so eg25-manager wouldn't get notified the modem has been brought back and then hit the timeout)

mentioned in commit querl.dox/linux/alpine_linux/aports@5bf18d92

I'm experiencing the same issue, but even with 0.1.1. After waking up from any sleep (e.g. pressing the power button, or receiving a call), the modem gets reset, resulting in a hangup tone when I'm calling the phone. This is on PostmarketOS 20.05.1

eg25m-suspend.log

Which distro and version of ModemManager are you using? Is it patched?

The ModemManager object '/org/freedesktop/ModemManager1/Modem/10' removed line should be printed before suspend. If MM has a custom patch disabling the suspend/resume management (as I suspect) then you're in "untested and likely to fail" territory.

PostmarketOS 20.05.1, mmcli 1.12.10. I'll go and investigate possible patches.

pmOS is using MM from Alpine, which isn't patching it at all from upstream.

(see source list in the APKBUILD: https://git.alpinelinux.org/aports/tree/community/modemmanager/APKBUILD?h=3.13-stable)

@dusxmt did you upgrade the modem firmware? There was one report when we were testing this a while back of stability getting worse after upgrading the firmware...

@craftyguy Indeed, I'm now running EG25GGBR07A08M2G_01.003.01.003. However, I've tried several different firmware images, as well as a fresh install of pmOS on an SD card, and this behavior remains consistent.

@craftyguy Then I guess that's the problem: IIRC ModemManager needs patches to implement suspend/resume using elogind.

Hopefully 692f9134 should solve your problem (but then it won't detect all modem failures, and might even screw things up), but moving to an MM version with elogind support might be a better idea.

PS: I haven't tested this patch (no time this week-end)

I tried it out, it did actually resume and show a calling dialog screen, but I wasn't able to pick it up (there was a sequence of three high-pitched beeps (with consecutively rising pitch) at the calling phone when I tried, which I think signifies an error), and on second attempt it lost the modem anyways. So yeah, I don't think this is the right approach, modemmanager will probably need to be patched.

mm-suspend-2.log

Okay, the "not being able to pick it up" issue is unrelated, I've noticed that it happens even during normal use (sometimes I can pick up the call, sometimes not), downgrading the modem's firmware to EG25GGBR07A08M2G_01.002.01.002 appears to make it more reliable (I haven't been able to reproduce it here yet).

That's good to know, thanks for the follow-up!

• EG25GGBR07A08M2G_01.003.01.003 has issues with URCs and modem operations in general
• ModemManager on pmOS should be patched for elogind, I did it myself in the edge repo.
• ModemManager should be version 1.14.8 which is more reliable than 1.14.6 and older.

If stable uses ModemManager from pmOS instead of Alpine, it is patched for elogind by me.

I believe it uses the one from Alpine; we should probably ask the maintainers to cherry-pick it into stable.

Ok, an issue was filed in the pmos repo.

But, this is unrelated to the original problem I reported, which is using patched modem manager w/ suspend support on eg25-manager 0.1.2..

I apologize. At the time of first writing a reply, I believed it was related, because of the very similar behavior to my problem, and a lack of any logs - so I provided mine.

No worries, it did seem like a similar problem. I just wanted to make sure it was clear here that this issue should stay open :)

mentioned in issue postmarketOS/pmaports#947 (closed)

EDIT: nevermind, upon further testing this does not appear to be an eg25manager issue. Sorry for the thrash.

@craftyguy @a-wai I finally have time again to work on the PinePhone! I pulled the logs and this happens:

pinephone:~# cat /var/log/eg25-manager 
** Message: 19:41:54.029: Starting modem...
** Message: 19:41:55.030: Executed power-on/off sequence
** Message: 19:41:55.036: ModemManager vanished from D-Bus
** Message: 19:41:55.764: taking systemd sleep inhibitor
** Message: 19:41:55.770: inhibitor fd is 18
** Message: 19:42:05.396: Response: [RDY]
** Message: 19:58:11.368: system is about to suspend
** Message: 19:58:11.368: Sending command: AT+QCFG="urc/cache",1
** Message: 19:58:11.385: Response: [OK]
** Message: 19:58:11.385: Executed suspend sequence
** Message: 19:58:11.386: suspend sequence is over, drop inhibitor
** Message: 19:58:11.386: dropping systemd sleep inhibitor
** Message: 20:00:22.160: system is resuming
** Message: 20:00:22.160: taking systemd sleep inhibitor
** Message: 20:00:22.161: Executed resume sequence
** Message: 20:00:22.185: inhibitor fd is 18
** Message: 20:00:31.193: Modem wasn't probed in time, restart it!

** (eg25manager:2739): WARNING **: 20:00:31.193: Unknown modem USB ID
** Message: 20:00:31.193: Sending command: AT+CFUN=1,1
** Message: 20:00:31.208: Response: [OK]
** Message: 20:00:33.425: Response: [c\u0010]

However, I see the modem coming back on the lockscreen with the signal strength indicator. It stays there for 0.5 - 1 ring tones and disappears.

@dylanvanassche I believe https://gitlab.com/mobian1/devices/eg25-manager/-/commit/3bb6e15de0e14c08135c82dd71d9774ad02ffc1b might solve your problem.

Great! That seems to work!

@a-wai I think we can close this one?

Good idea, yes :)

closed

mentioned in issue #38