WIP: Added timestamping of log entries written to the log server

[perlun]

(Title of this PR changed halfway. The original title was "//ramdisk/config/servers/boot/startup not found")

---

This is an attempt to work in a systematic way with trying to nail down this issue...

When booting up chaos now, I get this error:

[boot] beginning of boot
[boot] Mounted the first available block service as //ramdisk.
[boot] Reading startup script...
[virtual_file_system] Mounting 23 at //ramdisk.
[boot] //ramdisk/config/servers/boot/startup not found.

My conclusions thus far:

• The boot server is responsible both for mounting the file system (from the initial ramdisk) and reading its file. The former seems to work, the latter seems to fail.
• The FAT server (which should be serving this filesystem) never seems to get any messages, which seem to indicate that the connection between VFS and individual file systems are broken somehow. To me right now, it feels unclear how this work: what is it that determines what file system server should be used for a given mount point?
• When trying to debug the vfs server, I've placed breakpoints in vfs_get_file_info to try and step through the code. However, it's been an extremely unpleasant situation trying to get my hands dirty on this one. Because of chaos multiprocess/multi-threaded nature, stepping in the debugger will make things "jump around" very much, making it really really hard to understand what's going on.

So I think:

1. Either we have to fix gdb to properly understand our process/thread model. Probably doable, amount of work unknown OR
2. Resort to printf style debugging, i.e. log_print_formatted. This is the approach I'm leaning towards and which I've tried to a certain points.

Comments/suggestions/ideas from all of you are highly appreciated. This bug is really bothering me, and it's frustrating to not understand the code that we wrote 15 years ago... 😄

[perlun]

Alright. After having spent a bunch of unfruitful attempts at getting this working, I'm leaning towards the notion that this stuff has never worked. Which is strange, since I seem to remember it working. But it could very well be that I am mistaken, and the average quality (or lack thereof) of the code seems to indicate that...

In chaos4ever/chaos-old@1aadeee, I noted that we seem to support starting cluido from grub instead. That's of course much easier. I don't like that approach, but maybe we'll live with that for now.

[perlun]

The only real value that his PR adds is the timestamping of log entries. However, for whatever reason it doesn't work as intended; it always logs the same time... I don't fully know what the problem is here, but I feel like I want to resolve that before we merge this.

[perlun]

Debugged this a bit further. The value that the log server receives always seem to have the same value in the timestamp field, so I'm suspecting the system call doesn't work as expected (or the output constraints messes things up).

[perlun]

Weird. The timers_update is indeed updating the timeslice in the kernel, but all system_call_timer_read calls seemed to return the same value...

[perlun]

This PR is stale and should be rebased and fixed.