#asahi-re · 2022-04-10 05:09
rqou_for bluetooth#asahi-re · 2022-04-10 05:09
rqou_yes, macos definitely uses more#asahi-re · 2022-04-10 05:08
rqou_many parts are similar though#asahi-re · 2022-04-10 05:08
rqou_i have been reading it, but it doesn't look the same?#asahi-re · 2022-04-10 05:02
rqou_yolo#asahi-re · 2022-04-10 05:02
rqou_afaict it's not guaranteed to be page aligned, but happens to be for now#asahi-re · 2022-04-10 04:39
rqou_i added a hack for just the opcodes i need for now, which seems to not crash in my case#asahi-re · 2022-04-10 04:39
rqou_rip ok#asahi-re · 2022-04-10 04:37
rqou_has the code been pushed yet?#asahi-re · 2022-04-10 04:37
rqou_i missed that#asahi-re · 2022-04-10 04:37
rqou_lol ok#asahi-re · 2022-04-10 03:45
rqou_hey marcan how hard would it be for you to fix this?#asahi-re · 2022-04-10 03:44
rqou_... or not, apparently simd regs aren't even available in exc_info#asahi-re · 2022-04-10 03:42
rqou_apparently i need to fix "HV: store not emulated: 0xad008901"#asahi-re · 2022-04-10 03:41
rqou_anyone got an aarch64 opcode map reference guide?#asahi-re · 2022-04-09 22:42
rqou_somebody better start looking at thunderbolt soon or i will be sad#asahi-re · 2022-04-09 22:36
rqou_bit31 of register 0 is also set, but i don't know what that means#asahi-re · 2022-04-09 22:34
rqou_sven: i also tested the 36-bit iova on the prores DART and it seems to work#asahi-re · 2022-04-09 22:28
rqou_from the build params: version is still 1.0, supports full bypass (unlike the prores/scaler/avd ones which do not), 42-bit PA, 32-bit VA (not 36-bit), 64 streams (not 16)#asahi-re · 2022-04-09 22:27
rqou_sven: ok i bought a thunderbolt widget just to answer the DART question you had asked about#asahi-re · 2022-04-09 09:35
rqou_tfw i kept wondering where a giant pile of config settings were coming from, but they're in the info.plist#asahi-re · 2022-04-09 09:24
rqou_need to do more tracing to figure out how descriptor rings work#asahi-re · 2022-04-09 09:23
rqou_(the only evidence is that there is a register that changed from "0" to "2")#asahi-re · 2022-04-09 09:23
rqou_afaict i've successfully booted the bluetooth firmware now#asahi-re · 2022-04-09 08:19
rqou_no more crashing now#asahi-re · 2022-04-09 08:19
rqou_yup that was it#asahi-re · 2022-04-09 07:59
rqou_there's definitely a lot of parts of this that are similar to broadcom wifi chips, but also a lot of it is different#asahi-re · 2022-04-09 07:29
rqou_it writes some "BAR window mapping" magic numbers to the config space#asahi-re · 2022-04-09 07:29
rqou_lmao ok i chased some vtable hell and this is probably it#asahi-re · 2022-04-09 07:26
rqou_and then prints "window setup done"#asahi-re · 2022-04-09 07:26
rqou_it writes something to a bunch of pci config space registers at around 0x80#asahi-re · 2022-04-09 07:25
rqou_hmm, AppleConvergedPCI::setupVendorSpecificConfigurationGated might be important#asahi-re · 2022-04-09 07:03
rqou_but if i read BAR0+0x5908 (which macos definitely does!) it hangs the entire system#asahi-re · 2022-04-09 07:02
rqou_the BAR is accessible because i can read BAR0+0x302C#asahi-re · 2022-04-09 06:44
rqou_how would i hv trace the config space?#asahi-re · 2022-04-09 06:43
rqou_do i have to do something with the config space? i don't really know much about how pci works#asahi-re · 2022-04-09 06:42
rqou_literally the first access macos does is to read BAR0+0x5908, according to the hv trace#asahi-re · 2022-04-09 06:41
rqou_so macos is doing something i don't understand to make the device not crash#asahi-re · 2022-04-09 06:41
rqou_hmm ok, there's definitely something i don't understand happening, because reading one of the "bad" registers (that macos is blatantly reading) from a "proper" linux pci driver crashes the system still#asahi-re · 2022-04-09 03:46
rqou_the macos drivers are really really overengineered#asahi-re · 2022-04-09 03:36
rqou_i think but haven't confirmed that hci/sco/acl are tunneled on top#asahi-re · 2022-04-09 03:36
rqou_the pcie interface doesn't look _that_ complicated#asahi-re · 2022-04-09 03:35
rqou_i _think_ i understand how to upload bluetooth firmware now, and have a vague idea that there exist descriptor rings and doorbells for 7(?) streams of data, but i'm blocked on not having any good way to prototype a driver#asahi-re · 2022-04-09 01:42
rqou_ah, apparently i can just dump mmio ranges from ioregistryexplorer#asahi-re · 2022-04-09 01:07
rqou_i still don't have tracing working, but i REd the macos driver a bunch#asahi-re · 2022-04-09 01:06
rqou_bluetooth seems to have a broadcom chipcommon at BAR0+0x3000#asahi-re · 2022-04-08 06:54
rqou_lmao no they're different sizes#asahi-re · 2022-04-08 06:52
rqou_fwiw bluetooth has 2 BARs so it's not "simple"#asahi-re · 2022-04-08 06:50
rqou_i'm guessing wifi was made working without involving much tracing?#asahi-re · 2022-04-08 06:49
rqou_i'm still not super familiar with all of the trace tooling