cameras: unexplained shutdowns

I'm not sure how many issues this is for the price of one, but in general, the input pattern is:

0. Have a Birch unit
1. Start Megapixels
2. Play around
3. Stop Megapixels
4. Leave the phone for half an hour or longer (or maybe 10 minutes is enough?) while connected to a USB3 port

and the result is either a hang or a shutdown. The phone is on the warm side when it's a hangup. Recovery is limited, as in I was trained to take the battery out. Most times the battery is empty, although I'm not sure if this is the result of previous attempts.

I haven't limited this to either camera yet.

added camera label

I have seen some kernel oopses when using the cameras which then sometimes triggered some unusual behavior, like non-working touchscreen until a blank-unblank cycle. Could it be something similar?

changed the description

I have not seen anything in dmesg while this was happening. It's possible it got cut off (viewing over ssh). I haven't checked USB or network connectivity.

I opened Megapixels, got a few frames from the selfie cam and turned it off. The battery was low, but it was on the cable.

Last console lines:

[  274.817867] S5K3L6-CIS 3-0010: OFF
[  295.661247] max17042 3-0036: SOC threshold INTR
[  484.134456] max17042 3-0036: SOC threshold INTR
[  515.586680] max17042 3-0036: SOC threshold INTR
[  529.866329] rfkill: input handler enabled
Connection to 10.42.0.69 closed by remote host.
Connection to 10.42.0.69 closed.

Then it got into a reboot loop.

Found a different failure while a photo was getting saved:

[ 1006.319798] S5K3L6-CIS 3-0010: set_dtream sd ffff00002f708220
[ 1006.320135] s5k3l6xx: i2c_write to 0x0100 : 0x01
[ 1006.321200] s5k3l6xx: i2c_read: 0x0100 : 0x01
[ 1015.230392] buck3: failed to disable: -ETIMEDOUT
[ 1015.235062] imx-pgc imx-pgc-domain.5: failed to disable regulator: -110

Phone needed a hard reset.

that's the pmic failing to turn off the GPU regulator (which then prevents the GPU PD from suspend) which basically can only mean that the i2c call to do so timed out. PMIC is on a different i2c bus than the camera though. Did you try monitoring power usage when doing this?

No, how do I do this?

Also noticed some hangup when the camera was switched off.

This is now about 90 minutes after having made some pictures and everything still seems to be running fine (on Evergreen with Byzantium).

it think @dorota.czaplejewicz is using birch and i'm seeing 'hangs' there too reason being that thermal-cpufreq kicks in early and grinds some things to halt (not using camera at all here).

New data point: I just started a new kernel (camera modules blacklisted), and within seconds of ssh-ing into it, ssh response turned sluggish, and now I'm stuck with the display showing "14:50".

My Birch is quickly getting unuseable.

I think the regression is somewhere between 5.9.* and now, because it used not to be this way (assuming it's not the loose component inside shorting something).

PS. What was the command to check power usage again?

PPS. The phone isn't very warm.

I changed my mind. This might be temperature/load related. I haven't used continuous camera streaming extensively before, best I could do was:

• compiling Squeekboard (no hangup, maybe due to slow swap)
• restoring via Jumpdrive (rather reliable hangups)

Today it's colder (more moisture?) and the phone behaves much better.

Power usage: cat /sys/class/power_supply/max170xx_battery/uevent

Temperature and thermal throttling status: gthcli

FYI in the past when L5 was still a portable hand heater I have used to take the back cover off and put a USB fan near the phone when compiling or doing other CPU intensive stuff, which was helping my Birch a lot. Just having the air slightly circulate around the phone was making a huge difference.

FWIW restoring via Jumpdrive has a property of keeping the DRAM at maximum frequency (since you can't turn the screen off), which eats quite a bit of power.

Most recent hangup: battery @ 42°C (high), rest at 54°C (normal-high).