#asahi · 2021-06-07 14:41
marcanand then it's up to the distros whether to eventually cave or not#asahi · 2021-06-07 14:41
marcanpeople will use them#asahi · 2021-06-07 14:41
marcanand then if the distros don't ship 16k kernels, someone else will#asahi · 2021-06-07 14:41
marcanif we can't make the kernel somehow deal with a 16K IOMMU when built for 4k, there is no other good option#asahi · 2021-06-07 14:41
marcanI mean, look, if there is no better solution, that's that#asahi · 2021-06-07 14:41
marcan*shrug*#asahi · 2021-06-07 14:40
marcane.g. ubuntu has 64k and 4k options#asahi · 2021-06-07 14:40
marcansome already do#asahi · 2021-06-07 14:40
marcanI expect distros to ship multiple kernels#asahi · 2021-06-07 14:39
marcando we even know if the GPU MMU does 4K pages? because if it doesn't, lol no 3D for 4K kernels#asahi · 2021-06-07 14:38
marcanand besides, 16K has advantages on other platforms too#asahi · 2021-06-07 14:38
marcanenough people will want M1 not to suck#asahi · 2021-06-07 14:38
marcanwe're going to change that ;)#asahi · 2021-06-07 14:36
marcanbut the wifi attack surface is a couple orders of magnitude worse (and it's had a bunch of CVEs already)#asahi · 2021-06-07 14:36
marcanyeah#asahi · 2021-06-07 14:36
marcanthough really nobody should be using these things with 4K pages, given all this, except for like installers#asahi · 2021-06-07 14:36
marcanyeah, if we find a sane way to disable wifi in this scenario I'd be happier#asahi · 2021-06-07 14:35
marcanso if we do that at least I want a scary kmsg warning#asahi · 2021-06-07 14:34
marcanthat's a *significant* downgrade#asahi · 2021-06-07 14:34
marcanBroadcom *wifi* firmware#asahi · 2021-06-07 14:34
marcanyou are extending the system attack surface to include Broadcom firmware#asahi · 2021-06-07 14:34
marcanwell...#asahi · 2021-06-07 14:33
marcanthe tbt ones in particular#asahi · 2021-06-07 14:33
marcanI still think we outright shouldn't support putting those in bypass mode, or at most do it only with a nondefault, scary parameter#asahi · 2021-06-07 11:01
marcanfrom 1TR it's safer#asahi · 2021-06-07 11:01
marcanpitust[m]: are you doing this from 1TR? it should work live too, but much more likely to hang#asahi · 2021-06-07 09:14
marcanWe do know that OSX configures some "interesting" nonstandard memory behavior, but I don"t *think* our chicken bits do. Also, that would probably break VMs unless they reset them in Hypervisor.framework#asahi · 2021-06-06 14:02
marcantop of RAM, outside the OS-exposed RAM region#asahi · 2021-06-06 14:02
marcanthe shared-region stuff is just a physical address afaik#asahi · 2021-06-06 14:01
marcanbloom: I'm pretty sure GPUs designed to share the CPU pagetables exist#asahi · 2021-06-06 05:51
marcanseems like it would save a ton of management/overhead for unified memory systems if the GPU can just outright use the userspace page tables#asahi · 2021-06-06 05:50
marcanbloom: do you know if Linux supports this kind of thing? IIRC the PS4 does it like this with their OS, but I don't know if the linux amdgpu driver could ever do that#asahi · 2021-06-06 05:50
marcanso it sounds like we can probably expect the GPU to share the CPU page tables, and have its own SPRR reg to do remapping#asahi · 2021-06-06 05:44
marcanapple system coprocessor? (wild guess)#asahi · 2021-06-06 05:44
marcansven: ascwrap-v4 is their general "sub cpu" thing#asahi · 2021-06-06 05:43
marcan32 bits is enough for access perms for just userspace, without GL#asahi · 2021-06-06 05:42
marcansven: that multiple table stuff sounds like the extra registers I found...#asahi · 2021-06-05 00:34
marcanplus I want to also use it to control m1n1 (turn mmiotrace on/off)#asahi · 2021-06-05 00:34
marcanso the msr trap thing is probably a better idea long term#asahi · 2021-06-05 00:34
marcanthough that shows up on the wrong ACM port, it's not serialized with mmiotrace#asahi · 2021-06-05 00:34
marcan*but#asahi · 2021-06-05 00:34
marcan:p#asahi · 2021-06-05 00:34
marcancut also yeah debug-console works I guess#asahi · 2021-06-05 00:32
marcanbloom: I thought of using the m1racles register for that :D (userspace hypercalls)#asahi · 2021-06-04 20:45
marcanpipcet[m]: ^#asahi · 2021-06-04 20:45
marcanjust woke up#asahi · 2021-06-04 20:45
marcaner, yeah, sorry#asahi · 2021-06-04 20:44
marcanthe correct way to support dual-booting without the "hold down the power button" issue is to have whatever boot picker you want to have, but actually clean reboot into the new OS (via nvram var) the same way 1TR does#asahi · 2021-06-04 20:38
marcanbut hey, whatever floats your boat :p#asahi · 2021-06-04 20:38
marcantwo OSes using the same SEP storage/context is not a good idea