#asahi-dev · 2023-04-29 08:01
marcanbut since ane1 and ane2 are basically the same, no reason to believe ane1 is different in any way#asahi-dev · 2023-04-29 08:01
marcanso that makes me think ane0 may have some other subtle difference (or maybe not and it's just ARM instruction encoding gave a different instruction count when they changed the base addr)#asahi-dev · 2023-04-29 08:00
marcanane0 and ane2 are more different#asahi-dev · 2023-04-29 08:00
marcanbut actually ane1 and ane2 are almost identical in .text, then .data changes#asahi-dev · 2023-04-29 08:00
marcanthe first difference I see is different MMIO addresses (duh)#asahi-dev · 2023-04-29 08:00
marcane.g. ane0 and ane2 have different firmware even though they are literally the same hardware#asahi-dev · 2023-04-29 07:59
marcanthe firmwares are all different for stupid reasons#asahi-dev · 2023-04-29 07:55
marcanane is actually quite power hungry, it's a major component in the CLPC stuff and has its own power rail, so ane1 is probably not even wired to power on any machine#asahi-dev · 2023-04-29 07:54
marcanI'm sure they had some test hardware somewhere with it enabled, but they never shipped it#asahi-dev · 2023-04-29 07:53
marcanso the double ane for max models was a planned feature that got cut and removed from the next iteration#asahi-dev · 2023-04-29 07:53
marcanm2 max does not have ane1 physically, it is not present in the die#asahi-dev · 2023-04-29 07:52
marcaneiln_: I think they decided not to ever enable it for power delivery/product reasons#asahi-dev · 2023-04-29 07:12
marcandottedmag: ack, let me know if you need anything! there's a bunch of discussion on the linked fdo bug :)#asahi-dev · 2023-04-28 15:37
marcanI mean I guess it could be eDRAM but I don't know if they do that#asahi-dev · 2023-04-28 15:37
marcanChaosPrincess: eh, it's 1MB#asahi-dev · 2023-04-28 15:37
marcan_jannau_: doesn't make much of a difference though#asahi-dev · 2023-04-28 15:36
marcanboth I guess?#asahi-dev · 2023-04-28 15:35
marcan(need to fix the size field though, just that)#asahi-dev · 2023-04-28 15:34
marcansince it never chainloads anything else and will inherit that from u-boot if chainloaded#asahi-dev · 2023-04-28 15:34
marcanthe linux driver doesn't have to do anything :)#asahi-dev · 2023-04-28 15:33
marcanthere's plenty of space at the end of SRAM so it's perfectly fine to steal some space for oslog there IMO#asahi-dev · 2023-04-28 15:33
marcan(not 4k pages)#asahi-dev · 2023-04-28 15:32
marcansize is in bytes#asahi-dev · 2023-04-28 15:32
marcanalso the size field extends down 12 more bits in the message, the DVA isn't as large as we thought#asahi-dev · 2023-04-28 15:32
marcansince that sticks in linux, and forwarding RAM mappings is a major PITA#asahi-dev · 2023-04-28 15:32
marcaninstead of RAM#asahi-dev · 2023-04-28 15:32
marcanstill need to fix oslog but if you're bored: for MTP, we need to allocate oslog out of the end of the sram region in the DT#asahi-dev · 2023-04-28 15:26
marcanI don't expect dcpext to be done by then :p#asahi-dev · 2023-04-28 15:26
marcanwhat I mean is that we'll probably release in O(1-2 weeks) for M2 Pro/Max only#asahi-dev · 2023-04-28 15:26
marcan_jannau_: dcpext would be a very good reason to call it useful#asahi-dev · 2023-04-28 15:25
marcanlina: see https://lore.kernel.org/linux-arm-kernel/20220726145430.bfwidmw6xmeppbfb@bogus/ , looks like it's known broken upstream (turn off CONFIG_PROVE_RAW_LOCK_NESTING to shut it up)#asahi-dev · 2023-04-28 15:23
marcaninitially only the new machines will get 13.3#asahi-dev · 2023-04-28 15:23
marcanonce it's useful, yes#asahi-dev · 2023-04-28 15:14
marcanI'll take a look#asahi-dev · 2023-04-28 15:14
marcanlina: argh :(#asahi-dev · 2023-04-28 15:14
marcancorrect#asahi-dev · 2023-04-28 15:13
marcanand some stuff in the ADT bindings but only m1n1 cares about that#asahi-dev · 2023-04-28 15:13
marcanand GPU#asahi-dev · 2023-04-28 04:29
marcanif I could repro it locally I could outright use a USB analyzer but...#asahi-dev · 2023-04-28 04:28
marcanI don't see any suspicious commits in usb core or xhci#asahi-dev · 2023-04-28 04:25
marcanthe logs just sound like the device itself is resetting, there's no actual errors#asahi-dev · 2023-04-28 04:20
marcanone of the things we changed in 6.3 is cpuidle, so it could even be just a random timing difference#asahi-dev · 2023-04-28 04:20
marcancould even be a firmware bug in the device itself that just coincidentally is getting hit now#asahi-dev · 2023-04-28 04:19
marcanif I had to guess this is a bad interaction between a change in 6.3 and that very specific USB device#asahi-dev · 2023-04-27 15:53
marcanIDs, SRAM offset, OTP offset, plus (hopefully zero but possibly not) firmware API changes#asahi-dev · 2023-04-27 15:52
marcankaazoo: The new chips are not supported yet, there is a known set of things we need to implement for every new chip#asahi-dev · 2023-04-27 15:07
marcanso that one is proobably right#asahi-dev · 2023-04-27 15:07
marcanpovik: IIRC I did see similar behavior for IRQ_NOT_PEND when testing something like mailbox messages, the FIQ one was a guess#asahi-dev · 2023-04-27 12:56
marcanI put an IR on my headphones :D#asahi-dev · 2023-04-27 12:54
marcanbut you can at least correct the speakers