Added OBSW Watchdog #67

muellerr · 2021-07-29T17:29:19+02:00

muellerr commented

2021-07-29 17:29:19 +02:00

This is a software watchdog implementation which traces whether the EIVE OBSW is running by using the Linux FIFO (named pipe) functionality.

The watchdog will create a pipe at /tmp/watchdog-pipe which can be used both by the watchdog and the EIVE OBSW. The watchdog will only read from this pipe while the OBSW will only write to this pipe.

The watchdog checks for basic ASCII commands as a first basic feature set. The most important functionality is that the watchdog can detect if a timeout has happened. This can be beause the OBSW is hanging (or at least the CoreController thread) or there is simply no OBSW running on the system. It does so by checking whether the FIFO is regulary written to, which means the EIVE OBSW is alive.

If the EIVE OBSW is alive, a special file called /tmp/obsw-running will be created. This file can be used by any other software component to query whether the EIVE OBSW is running. The EIVE OBSW itself can be configured to check whether this file exists, which prevents two EIVE OBSW instances running on the Q7S at once.

If a timeout occurs or the OBSW hangs up(e.g. debugging session cancelled), this special file will be deleted by the watchdog as well.
The watchdog and its configuration will be directly integrated into this repostory, which makes adaptions easy. It might be worth thinking about whether the watchdog is allowed to reboot the system.

This is actually an excellent chance to introduce Rust code on the Q7S. Might look into this soon. For now, I will introduce a systemd service as well to start the watchdog at startup.

I also added a feature to make the EIVE OBSW binary name unique, by adding $ENV{USERNAME} to the binary name. This prevents overwriting a currently running *.elf file when debugging.

IMPORTANT: The new obsw binary name is eive-obsw-$ENV{USERNAME} now. Update tcf agent remote debugging configurations accordingly

This is a software watchdog implementation which traces whether the EIVE OBSW is running by using the Linux FIFO (named pipe) functionality. The watchdog will create a pipe at `/tmp/watchdog-pipe` which can be used both by the watchdog and the EIVE OBSW. The watchdog will only read from this pipe while the OBSW will only write to this pipe. The watchdog checks for basic ASCII commands as a first basic feature set. The most important functionality is that the watchdog can detect if a timeout has happened. This can be beause the OBSW is hanging (or at least the `CoreController` thread) or there is simply no OBSW running on the system. It does so by checking whether the FIFO is regulary written to, which means the EIVE OBSW is alive. If the EIVE OBSW is alive, a special file called `/tmp/obsw-running` will be created. This file can be used by any other software component to query whether the EIVE OBSW is running. The EIVE OBSW itself can be configured to check whether this file exists, which prevents two EIVE OBSW instances running on the Q7S at once. If a timeout occurs or the OBSW hangs up(e.g. debugging session cancelled), this special file will be deleted by the watchdog as well. The watchdog and its configuration will be directly integrated into this repostory, which makes adaptions easy. It might be worth thinking about whether the watchdog is allowed to reboot the system. This is actually an excellent chance to introduce Rust code on the Q7S. Might look into this soon. For now, I will introduce a systemd service as well to start the watchdog at startup. I also added a feature to make the EIVE OBSW binary name unique, by adding `$ENV{USERNAME}` to the binary name. This prevents overwriting a currently running `*.elf` file when debugging. IMPORTANT: The new obsw binary name is `eive-obsw-$ENV{USERNAME}` now. Update tcf agent remote debugging configurations accordingly

muellerr added the

feature

label 2021-07-29 17:29:19 +02:00

muellerr added 12 commits 2021-07-29 17:29:20 +02:00

added functions to get current boot copy c0b91a4879

moved sd card init, more functions static now 1a4f907925

README update 1e6ea91438

README update b846a936b4

updated cp script fc769be789

reading chip and copy works 125593be21

new conf option 190590e1b5

added eive-watchdog 59bcf6cec2