Yes, instead it’s possible for the userland processes that provide the equivalent functionality (drivers, etc.) to crash. Some of these are probably recoverable, but some of them probably aren’t (especially if the drivers in question were ported from kernel environments that don’t expect crashes to be recoverable.)
This is absolutely not the case at all. I have repeatedly followed up with the TuneTracker folks about KDLs and other issues they’ve reported, and more than once (if memory serves) when emailing them, I found out that there were others they hadn’t reported. Yes, many of them I couldn’t reproduce, but the story hardly ended there; I would try to guide them through collecting more information to make the problem easier to track down on our end. But oftentimes these email threads would simply go dormant without me having received enough information to make an “educated guess” at solving the problem.
Two of the issues they had, last I knew, was that they were trying to run on AMD “Bulldozer” hardware, which of course is infamously unstable under Haiku when running with SMP on; and they were also using “Disable AHCI” in the BIOSes, which seemed to break disk I/O after some period of time, or perhaps randomly. I recommended turning AHCI mode back on, and I think they said this improved things, but I don’t remember hearing back as to whether they had any better success with SMP disabled, or with non-Bulldozer hardware.
That was years ago. I emailed them more recently to try and find out if anything had changed on this front, but that thread didn’t seem to be much more productive than past ones.
This may be related to using “Disable AHCI mode” on newer motherboards, if that’s what they were doing. But it also could have been BFS bugs that have since been solved. I don’t know of any BFS tickets about data corruption which happen with any frequency (if at all, anymore), besides ones caused by the underlying hardware disconnecting or something like that.
Genode is surely not immune to filesystem bugs more than any other OS. Of course if they use better-tested code, then that’s something, but microkernels do basically nothing to prevent data corruption on the filesystem level, if it’s bugs in the filesystem driver and/or caching code that are causing the corruption.
This is the first I’ve heard of this (and again, I’ve reached out to the TuneTracker people multiple times over the years asking what issues they tended to encounter. Noticing a pattern?)
Since beta1, there have been more experiments by various developers and users running Haiku systems “in the cloud” with public IPs, and some have gotten lots of random SSH login attempts, etc. I don’t know how much I trust our IP stack to be exposed to the public either, but at least I haven’t heard of external traffic causing KDLs.
But more to the point: why would you ever expose a radio station’s broadcast hardware to the public internet? Surely it should be gated behind a firewall at the very least.
