5 ms·
If I thought there was a good chance of having the work upstreamed so that a mainline Linux kernel would eventually work on these computers, maybe. In my exper
by heavyset_go 6y ago
If I thought there was a good chance of having the work upstreamed so that a mainline Linux kernel would eventually work on these computers, maybe.
In my experience, there are not many ARM SoCs that have mainline support, even some of the Raspberry Pi boards don't, and they're some of the most popular ARM SoCs used for Linux.
A lot of these projects get to the stage where a forked kernel will run on the SoC, but then support and attention peters out when it comes upgrading the fork or getting the work into mainline Linux because a new and better ARM SoC will be released, and then that requires the same amount of work on a new Linux fork. Since each ARM SoC released is a unique hardware configuration, a lot of this work will be for naught when Apple releases new SoCs that require the same amount of work, or even more work, to get Linux running on them.
- sly010 6y agoWouldn't device tree solve (or at least mitigate) this problem somewhat? e.g. most broadcom chips have the same I2C hardware block, just mapped to a different memory address.
- marcan_42 6y agoThis would absolutely all be done with upstreamability in mind, and upstreamed as soon as is practical. All that horrible forked kernel stuff is the realm of companies with closed development teams who don't care and just want a product out the door. The PS4 Linux stuff I did was not upstreamed for lack of time/interest, but I absolutely considered it being decent enough to upstream when writing it. PS3 Linux running under GameOS mode patches were indeed upstreamed (though there was a bit of a lawsuit along the way so all of my code ended up attributed to a friend instead since I was keeping a lower profile at the time...).
- lmm 6y ago> All that horrible forked kernel stuff is the realm of companies with closed development teams who don't care and just want a product out the door. Or people who gave upstreaming an honest try but got fed up. It's not the easiest project to contribute to.
- marcan_42 6y agoYou do need to know what you're doing, meet quality standards, and do things in a way that fits existing architecture (unless you have a really good reason to revamp something). It's not beginner-friendly for sure, but I can tell you the vast majority of non-upstreamed vendor code is not upstreamed because it's terrible (which is also bad for users: things like blatant security problems are much more common in forks than upstream).
- krageon 6y agoIf you can't be bothered to learn how to write linux kernel patches properly you shouldn't be making them at all, in a forked kernel or otherwise.
- chippiewill 6y agoI think this is a legitimate concern, but it seems less likely to me, specifically in the case of Apple hardware, for two reasons: 1. There will be far more interest in being able to run Linux on that hardware, and in being able to keep it up to date (since it'll be a daily driver). 2. Apple are probably less likely to regularly make drastic changes between hardware revisions because they will need to support multiple hardware revisions themselves (Apple are typically pretty good with supporting old hardware). Their only strong motivator would be to actively prevent people running other OSes but I don't think Apple care that much about screwing over that niche demographic - especially when they're already paying for the hardware.
- oblio 6y agoI'm skeptical. Linux is reasonably popular for "Windows" hardware, most statistics put it at about 0.5% market share. I've never seen anyone running Linux on Intel Macs. Why would ARM Macs, which seem quite a bit more locked down, ever become a better supported and more popular Linux hardware platform? I wish they would become one, but I don't see how we can make this leap.
- marcan_42 6y agoHi from Linux on an Intel Mac. (I've got fixing up amdgpu to work on this 2015 iMac on my TODO list too... might even do that as a bonus warm-up if this takes off, while I get my hands on an M1 mac)
- oblio 6y agoI admire the work but you have to admit that even after all this time it's easier to install Ubuntu, say, on a random PC :-)
- marcan_42 6y agoTo be honest, the only problems on this iMac are audio (fair to say Apple's fault), the screen only running at 4K max (Apple custom stuff involved for 5K, but hey, not complaining about 4K, go find me an AiO PC with a 5K screem), the amdgpu issue (not really an Apple problem, it's just a rarer chip and probably a dumb bugfix and the older radeon driver works fine), and the Ethernet and SD card reader being problematic (that's a Broadcom chip and their silicon is universally buggy; not Apple's fault, these chips have issues on PCs too) So really just audio is the one thing that jumps out as broken and specific to this being a Mac. I personally happen not to care as I use an external audio interface anyway :-) Actual installation is trivial, it's just standard UEFI pretty much. As long as the right GPU driver loads you're fine. Thunderbolt and all that works out of the box.
- ndesaulniers 6y agoI'd be curious if you had a rough outline of what are the steps involved in porting Linux to a new target? In particular, I see folks working on a downstream port the the 3DS: https://github.com/xerpi/linux_3ds https://github.com/xerpi/linux_3ds. I'm kind of curious where do people even start?
- heavyset_go 6y agoHere's an outline for getting Linux running on ARM SoCs: https://elinux.org/images/a/ad/Arm-soc-checklist.pdf https://elinux.org/images/a/ad/Arm-soc-checklist.pdf
- zozbot234 6y ago> In my experience, there are not many ARM SoCs that have mainline support You have a point, but the pmOS project among others is working on upstreaming mainline support for these devices. This involves a lot of effort, to be sure but meaningful progress is being made.