#asahi · 2021-08-06 10:36
sventhat would at least allow to use the official distro installer + then add a custom package for the 16k kernel after everything already works#asahi · 2021-08-06 10:35
sveni'm sure some of them will come around eventually, but if supporting 4k kernels with some limitations isn't too painful i'd like to do that#asahi · 2021-08-06 10:16
svenmy plan is to send a RFC soonish anyway just to see if the iommu maintainers will even consider accepting such changes with their disadvantages#asahi · 2021-08-06 09:44
svenpossibly because they moved the AES stuff to their weird iommu-like thing#asahi · 2021-08-06 09:43
svenoh.. there's also NVME_QUIRK_128_BYTES_SQES, but that's not required anymore on the m1#asahi · 2021-08-06 09:38
sveni thought previous macs were just that the tags are shared between queues, that there's only a single interrupt and that the queue size is limited#asahi · 2021-08-06 09:37
svenapple calls it "nvme-linear-sq" in the ADT#asahi · 2021-08-06 09:37
svenif yes, than this queue thing is new#asahi · 2021-08-06 09:37
svennvme works on the T2 mac, doesn't it?#asahi · 2021-08-06 09:36
sveni think so#asahi · 2021-08-06 09:00
svenwhat other drivers need to use the SMC in which ways? smc -> usb phy could e.g. just be an interrupt#asahi · 2021-08-06 08:29
sveni think that could actually work. make the usb phy listen for that SMC notification and then just flip the reset bits. dwc3 will actually get a "new device" event then#asahi · 2021-08-06 08:29
svenhrm. or maybe we'll just need to hook it up to the usb phy driver#asahi · 2021-08-06 08:28
svenpossibly. but it's still annoying because we'll have to hook up that notification to the dwc3 driver :/#asahi · 2021-08-06 08:27
svenfwiw, mac os will lock down *a lot* while initializing the hardware#asahi · 2021-08-06 08:26
sveni'm still sad that they didn't hook up the "new usb connection" line to the dwc3 controllers :(#asahi · 2021-08-06 06:16
svenno idea if they ever started working on that though#asahi · 2021-08-06 06:16
sveniirc the pongoOs/checkrain people (never_released, but i haven't seen him here in a while) wanted to write a small hv to allow running windows#asahi · 2021-08-05 17:48
svenbut well...#asahi · 2021-08-05 17:48
svenit's technically an optimization but apparently there's broken code somewhere that relies on it#asahi · 2021-08-05 17:48
svenhttps://lore.kernel.org/linux-iommu/0d20bd6b-d0a1-019c-6398-b12f83f4fdf7@arm.com/#asahi · 2021-08-05 17:47
svenmy additions? they're skipping an optimization that would break#asahi · 2021-08-05 17:46
svenwhich bit?#asahi · 2021-08-05 17:40
svenyes! the bad thing is that i tried hacks first which were _more_ work than doing it correctly :D#asahi · 2021-08-05 17:38
sveni'll be a little bit sad if this works. it's less code than either of the identity bypass hacks i tried :<#asahi · 2021-08-05 17:36
svenFind First Set bit :-)#asahi · 2021-08-05 17:35
svenand i kinda expect anything that breaks to be subtle#asahi · 2021-08-05 17:35
svenbut i could boot from my usb device connected to pcie and use the pcie ethernet adapter just fine#asahi · 2021-08-05 17:35
svennot very well tested and possibly broken#asahi · 2021-08-05 17:35
svenj_ey: https://github.com/AsahiLinux/linux/tree/dart/dev#asahi · 2021-08-05 16:45
svendinner first, then cleaning it up a little bit and then git push :-)#asahi · 2021-08-05 16:43
sven 1 file changed, 76 insertions(+), 10 deletions(-) <-- maybe, just maybe, i should've directly taught dma-iommu how to deal with 16k pages on a 4k kernel instead of trying to hack this....#asahi · 2021-08-03 18:16
sventhe only thing that's annoying is that compiling the kernel on mac os is not much fun#asahi · 2021-08-03 18:16
sveni really enjoy my macbook air + mini setup :-)#asahi · 2021-08-03 18:14
svenalyssa> mps: If you're a linux hat, the M1 isn't a computer, it's an expensive toy for reverse-engineering addicts like marcan and me <-- that's... very very accurate.#asahi · 2021-08-03 18:02
svenand then the dcp or whatever drivers need to be made away of the right memory locations as well i guess#asahi · 2021-08-03 18:01
svenpretty sure it'll work#asahi · 2021-08-03 18:00
svenor, well, doing that in a sane way is going to be tricky#asahi · 2021-08-03 18:00
svenyeah, that's gonna be more tricky#asahi · 2021-08-03 17:59
svenpretty much. and force a DMA domain. but that's all easily possible#asahi · 2021-08-03 17:59
svenand ack errors#asahi · 2021-08-03 17:59
svenyou can still flush the tlb but not much more#asahi · 2021-08-03 17:59
svenmost registers are ro, yes#asahi · 2021-08-03 17:58
svenbut first the current driver needs to be accepted anyway :)#asahi · 2021-08-03 17:58
sven+dart#asahi · 2021-08-03 17:58
svenfor the locked we'll need to somehow hack the io-pgtable code to not allocate the first level pagetable but just take some pointer instead#asahi · 2021-08-03 17:56
svenj_ey: unmanaged domain just means "translation enabled, but not managed by the iommu-dma layer"#asahi · 2021-08-03 17:56
svenj_ey: no#asahi · 2021-08-03 17:55
svenbut i think it only maps the framebuffer#asahi · 2021-08-03 17:55
svenso disp0 dart should just work if you can do anything useful with that