you should join our discord! we are planning to use PongoOS, our toy kernel, in EL2, and use it to virtualize a GIC to boot UEFI (which will run in EL1). @never_released
-
-
this should simplify things quite a bit..
1 reply 0 retweets 3 likes -
Replying to @qwertyoruiopz @never_released
Sounds interesting, though is it worth the extra latency of adding a virtualized shim for interrupts? I mean, I've written custom interrupt controllers for Linux...
2 replies 0 retweets 1 like -
Replying to @marcan42 @never_released
i think the extra latency is low cost enough to not really matter and the advantages of not having to port OSes to the oddball platform are worth the cost
1 reply 0 retweets 2 likes -
Replying to @qwertyoruiopz @never_released
Sure, if it works and doesn't end up being a user experience compromise down the line, that's fine.
1 reply 0 retweets 0 likes -
Replying to @marcan42 @qwertyoruiopz
What made us seriously think about it is that M1 is new enough to have nested virtualisation, so that the thin hypervisor for this can still expose the full CPU feature set to the guest.
1 reply 0 retweets 5 likes -
Replying to @never_released @qwertyoruiopz
Yeah, that's a big one. Though nested virt usually has a nontrivial perf cost, but it depends on the implementation. E.g. what is the cost of the extra pagetable level (which presumably at least can be using very large pages)?
1 reply 0 retweets 1 like -
But anyway, there is always time to ditch the virtualization if it becomes the limiting factor, so I'm not too worried about that. Crossporting drivers back into the kernel is trivial compared to, like, GPU support (assuming the license permits).
2 replies 0 retweets 2 likes -
Replying to @marcan42 @qwertyoruiopz
pongoOS is wholly developed by the checkra1n team without outside contributions as of today, we currently license it under MIT, but we'll be able to add a dual-license if necessary.
1 reply 0 retweets 1 like -
Replying to @never_released @qwertyoruiopz
Yeah, there are a few approaches to merging this kind of code, but if it was MIT to begin with it's unlikely to be a problem. Mainly, someone needs to put a face (and real name) on it once it gets upstreamed, Linux doesn't take anonymous contributions.
2 replies 0 retweets 1 like
(But it doesn't mean all the authors need to be public, that person just needs to be authorized to upstream under GPL under their name)
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.