#asahi-dev · 2023-04-15 10:06
svenNice!#asahi-dev · 2023-04-13 19:54
svencould just be a dart debug interface that’s not synthesized for each dart#asahi-dev · 2023-04-13 19:54
svenSo it’s somehow related to the dart itself but no idea what exactly it does or why it’s separate #asahi-dev · 2023-04-13 19:54
svenI never figured out exactly what it is but it seems to expose the most recently accessed pagetable entries inside it’s MMIO #asahi-dev · 2023-04-13 19:53
svenjust look at that instance property in the ADT. It’d a bit weird but if that thing contains the SMMU fourcc and the two regs of the dart are separated by 0x4000 the second thing isn’t a DART#asahi-dev · 2023-04-13 19:52
sventhat also seems to be dart plus smmu #asahi-dev · 2023-04-13 19:50
svenjust ignore it#asahi-dev · 2023-04-13 19:50
svenso that second thing is that apple SMMU (not to be confused with arm‘s architectural SMMU)#asahi-dev · 2023-04-13 19:50
svenfor that dispdfr thingy it claims dart and smmu #asahi-dev · 2023-04-13 19:49
svenLook at the dart node, there’s an instance property#asahi-dev · 2023-04-13 19:47
svenare you sure they are darts and not dart plus what apple calls SMMU or something like that?#asahi-dev · 2023-04-13 19:47
sventhose two darts are very suspicious because they’re just next to each other #asahi-dev · 2023-04-13 19:46
svendo you recognize the fault address?#asahi-dev · 2023-04-13 19:46
sventhat usually only happens when that dart is actually used #asahi-dev · 2023-04-13 19:46
svenif you get translation faults on the second dart something tried to dma through it. #asahi-dev · 2023-04-12 09:37
sventhe biggest issues are that reverse engineering the firmware is usually much more challenging and time consuming than just reverse engineering its interface and that the firmware interface probably helps us when supporting new chip revisions#asahi-dev · 2023-04-12 09:35
sventhere might be power consumption penalties if we bypassed co-processors but I doubt we'd see a measurable performance penalty#asahi-dev · 2023-04-12 07:28
svenit doesn’t matter for the pcie darts so far because appclk is already enabled inside m1n1#asahi-dev · 2023-04-12 07:28
svenfwiw, technically pcie should be a clock controller and dart should have that APPCLK as a dependency #asahi-dev · 2023-04-11 15:30
svenI’ll take a look this weekend#asahi-dev · 2023-04-11 15:19
sveni thought they got dropped but maybe I messed up when I checked that ages ago #asahi-dev · 2023-04-11 15:19
svenwould be nice if the fuses could also just be applied from m1n1#asahi-dev · 2023-04-11 15:19
svenI also wonder if the older atcphys already keep the fuses on pd power off and/or reset or if apple changed their kext to skip that reset #asahi-dev · 2023-04-11 10:26
svenshould get back to staring at thunderbolt #asahi-dev · 2023-04-09 16:22
svennice :D#asahi-dev · 2023-04-08 16:33
sven… interrupts are fired the thunderbolt coprocessor crashes sometimes#asahi-dev · 2023-04-08 16:33
svenif I just do pcie hotplug when the link#asahi-dev · 2023-04-08 16:32
svenit’s pretty dependent on thunderbolt now though #asahi-dev · 2023-04-08 16:32
svenuhh.. kinda #asahi-dev · 2023-03-28 09:41
svenI think I did a quick hack and the WiFi chip didn’t like just exposing very small (8 bytes or so) regions #asahi-dev · 2023-03-28 09:39
svenI wanted to do that but never got around because that also requires changes to the iommu api #asahi-dev · 2023-03-28 09:13
svenI should also get back to TB and clean that up now that $work has calmed down a bit #asahi-dev · 2023-03-28 08:12
svenbut it shouldn’t break anything on 16k builds #asahi-dev · 2023-03-28 08:12
svenill have to check again #asahi-dev · 2023-03-28 08:11
sventhe patches may expose adjacent pages on 4K for untrusted fwiw#asahi-dev · 2023-03-28 08:02
svenother than that im rather sure the patches work, they’re just a bit awkward due to the way the iommu api works #asahi-dev · 2023-03-28 08:01
svenI don’t remember if I disallow dma for those devices entirely #asahi-dev · 2023-03-28 08:01
svenim not sure what happens with thunderbolt/untrusted plus my old 4K patches #asahi-dev · 2023-03-23 21:17
svenbut generally drivers are inside the Linux kernel. m1n1 usually only contains code to e.g. forward tunables to the kernel #asahi-dev · 2023-03-23 21:16
svenuh, what’s a thunderbolt usb 2 driver?#asahi-dev · 2023-03-21 12:09
svenno need for bypassing anything #asahi-dev · 2023-03-21 12:09
svenshe can just rebase my old 4K series #asahi-dev · 2023-03-19 10:33
svendon’t think so, seems to work for me #asahi-dev · 2023-03-15 10:37
sventhat also all sounds off topic #asahi-dev · 2023-03-15 09:50
svenah, that’s only vaguely related #asahi-dev · 2023-03-15 09:38
svenuhh… what’s yadro and how is that related to tbt?#asahi-dev · 2023-03-14 12:43
svenit would be nice to improve that api ofc but the last tries weren't very successful *shrug*#asahi-dev · 2023-03-14 12:42
svenand the API was really built for very simple and dumb mailboxes#asahi-dev · 2023-03-14 12:41
sventhe DT abstractions are nice and in theory the idea of mailbox also makes sense. it's just that there are >10 different hw implementations that make different assumptions about the API#asahi-dev · 2023-03-14 12:41
svenI just forgot to remove it from mailbox