Peach-Oled Segmentation Fault

[mhfowler]

peach issues 12.21.20

so the wifi is working again on my re-flashed pi now,
but alas some of the microservices are not running again.
peach-oled and peach-stats both throwing segmentation faults,
although peach-network is working!

1. peach-oled

• running peach-oled via apt-get install:
  installs successfully, but then the service fails to start and throws an error

• running peach-oled from binary crosscompiled on the vps:
  segmentation fault

• running peach oled via cross-compiling binary on my laptop in debian, and then running the binary on the pi, also has errors:

if I run peach-oled without sudo, this is the output:
[2020-12-21T18:23:04Z ERROR peach_oled] Application error: Failed to create interface for I2C device: Permission denied (os error 13)

if I run peach-oled with sudo, this is the output:
thread 'jsonrpc-eventloop-1' panicked at 'attempted to leave type `std::mem::ManuallyDrop<internal::SealedBag>` uninitialized, which is invalid', /rustc/7eac88abb2e57e752f3302f02be5f3ce3d7adfb4/library/core/src/mem/mod.rs:658:9 note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace Segmentation fault

as a sanity check, I tried building peach-oled on the pi, and then running the pi-built version,
and got the same errors

if I run cat /etc/modules, this is the output:

# This file contains the names of kernel modules that should be loaded
# at boot time, one per line. Lines beginning with "#" are ignored.
i2c-dev
i2c-bcm2835

if I run i2cdetect -y 1, this is the output:

00:          -- -- -- -- -- -- -- -- -- -- -- -- -- 
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
30: -- -- -- -- -- -- -- -- -- -- -- -- 3c -- -- -- 
40: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- 
70: -- -- -- -- -- -- -- --    

I looked in signal and saw that you said that when I run i2cdetect -y 1 I should see UU somewhere in the output, which I don't see. so maybe there is something wrong with the I2C settings.

any idea?

2. peach-stats

running via apt-get install:

• successfully installs, but fails to start

running binary compiled on vps:

thread 'http.worker00' panicked at 'attempted to leave type `std::mem::ManuallyDrop<internal::SealedBag>` uninitialized, which is invalid', /rustc/7eac88abb2e57e752f3302f02be5f3ce3d7adfb4/library/core/src/mem/mod.rs:658:9
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace
Segmentation fault

compiling on local debian laptop:

• can't compile
  mostly this error:

error[E0277]: `?` couldn't convert the error to `jsonrpc_core::Error`
  --> src/lib.rs:60:37
   |
60 |         let uptime = stats::uptime()?;
   |                                     ^ the trait `From<error::StatError>` is not implemented for `jsonrpc_core::Error`

3. peach-network

this is the service which originally wasn't working, and led to the whole re-flashing. now its working, so thats good news (via apt-get install or binary) :)

[mycognosist]

@mhfowler

Thanks for the detailed issue. Let's start from the core and work our way outward from there.

First things first, the I²C initialization. As you noticed, the presence of an alphanumeric (3c in your case) and lack of UU in the output of i2cdetect -y 1 means that an I²C device has been detected but not initialized in the kernel.

Thanks for sharing the contents of /etc/modules. That all looks good.

Can you please look at the contents of /boot/firmware/config.txt on your Pi and see if the following lines are present:

# Activate I2C
dtparam=i2c_arm=on
# Apply device tree overlay to enable pull-up resistors for buttons
device_tree_overlay=overlays/mygpio.dtbo

Please also check the output of this command groups peach-oled (shows the groups that the peach-oled user belongs to). Output should look like this:

peach-oled : peach i2c gpio-user

As a meta-thought, it might be a good idea to have setup_dev_env.py write to a simple output file on the Pi with the settings it was run with. That would allow us to quickly check if the script was run with the --i2c flag or if it was left-out etc.

---

Second troubleshooting inquiry: the seg fault from the VPS-compiled peach-oled binary:

I can confirm that the latest peach-oled binary on the VPS throws a segmentation fault error on my Pi.

When cross-compiled on my laptop and run with the peach-oled.service file managing the service, as is the case when installing via .deb, there are no errors and peach-oled runs as intended. So something is going wrong with the cross-compilation on the VPS and the error is not in the code / peach-oled crate itself.

I don't know why cross-compiling on the VPS is proving to be such a headache. It almost seems better to have me cross-compile on my laptop and push releases to the VPS for packaging. I'll spend some time on it today and see what I come up with.

[mhfowler]

thanks for all this!

/boot/firmware/config.txt looks right on this end

peach-oled was not a member of the gpio-user group
from my look at peach-config, it doesn't look like they are added (https://github.com/peachcloud/peach-config/blob/main/scripts/setup_dev_env.py#L92)

however, after adding peach-oled to the gpio-user group, I still get the same segfault when I run peach-oled

I'm going to open an ngrok tunnel so you can poke around on the pi directly