#asahi-dev · 2026-10-27 13:51
jannauthe dcp crash on t8122 is odd. either something unmaps holes into dcp's main shmem or dma_map_resource() maps display register into it#asahi-dev · 2026-10-27 13:21
jannauthe redesign is expected for tomorrow only for m5 pro and max#asahi-dev · 2026-10-27 13:20
jannaudcp is not resetted#asahi-dev · 2026-10-27 13:18
jannaubase m6 macbook pro/iMac, nothing new except 2.5 Gbps ethernet for the imac (and apple N1 wifi/bt) if I'm not mistaken#asahi-dev · 2026-10-27 11:34
jannaut8122 dcp works until it crashes during initial modeset#asahi-dev · 2026-10-27 11:29
jannauhuh, why did I use the top level comaptible in kboot.c:dt_set_display()?#asahi-dev · 2026-10-27 11:06
jannaus/but /#asahi-dev · 2026-10-27 11:05
jannaubut from dva "values"#asahi-dev · 2026-10-27 11:05
jannauyou might still need to trim the 0x100_00000000 off#asahi-dev · 2026-10-27 11:03
jannaumaybe not, In my stash I see "dva & 0xFFFFFFFFF" in proxyclient/m1n1/trace/asc.py#asahi-dev · 2026-10-27 11:01
jannauyuka: I think the m1n1 experiment has somewhere a "dva & ((1 << 36) - 1)" or so#asahi-dev · 2026-10-27 11:00
jannauI think the remap encoding isn't the same for all generations#asahi-dev · 2026-10-27 10:57
jannaut6031 has the same though#asahi-dev · 2026-10-27 10:55
jannauand wrong ADT, that was t6030#asahi-dev · 2026-10-27 10:55
jannauerr dart-dcp#asahi-dev · 2026-10-27 10:55
jannaufrom dcp-dart#asahi-dev · 2026-10-27 10:47
jannauyuka: ^^^#asahi-dev · 2026-10-27 10:47
jannauhow did we decode "remap = 0x50700000506" again? dcp is probably missing iommus on m3 max. it probably should be "iommus = <&dcp_dart 5>, <&dcp_dart 6>, <&dcp_dart 7>;"#asahi-dev · 2026-10-27 10:33
jannaum1n1 doesn't touch dcp on devices with internal panel (ignoring the broken m1/m2 dcp firmware workaround)#asahi-dev · 2026-10-27 10:24
jannauif the mailbox were the cause of the issue I'd expect it to either not work at all or fail earlier#asahi-dev · 2026-10-27 10:23
jannauI'd expect that all the mailboxes ASCWrap v6 beghave in the same way. it's probably not necessary for DCP because iboot already did that#asahi-dev · 2026-10-27 10:10
jannausorry#asahi-dev · 2026-10-27 10:10
jannaushould be device 5#asahi-dev · 2026-10-27 10:06
jannauyuka: please verify that linux mapped something at 0xfffe00000 in dart-dcp stream 0. something like `dart_dcp_tracer.dart.dump_device(0)` in the HV shell#asahi-dev · 2026-10-27 10:03
jannauthe difference shouldn't matter that early#asahi-dev · 2026-10-27 09:59
jannauthey stopped with the ipad nonsense for t8112 and t6030 uses a different firmware than max/ultra: t6030dcp.im4p vs t603xdcp.im4p#asahi-dev · 2026-10-27 09:51
jannauthat would mean it breaks earlier than expected#asahi-dev · 2026-10-27 09:49
jannauand then `python3 -m m1n1.fw.dcp.parse_log dcp_13.5/t6001_macos_dcpext0_hdmi_2.trace`#asahi-dev · 2026-10-27 09:48
jannau uncomment "dcp_tracer.ep.dcpep.state.dumpfile = ..." in trace_dcp.py#asahi-dev · 2026-10-27 09:47
jannauMichaelReeves[m]: any luck with rebasing the downstream wifi branch on your upstream submission?#asahi-dev · 2026-10-27 09:45
jannauany significant differences in linux / macos m1n1.fw.dcp.parse_log output except that the linux one stops at some point?#asahi-dev · 2026-10-27 09:38
jannauwhat happens with the odd buffer size? does the driver just give up or does it align the size and return?#asahi-dev · 2026-10-27 09:37
jannauthe syslog message sounds familiar and if it is what I remember caused by not acking messages from dcp#asahi-dev · 2026-10-27 09:32
jannauyuka: that sounds like message processing in linux is somehow broken#asahi-dev · 2026-10-27 09:31
jannaumacbook neo logic board (512GB) is $483.12 (minus $220 if you return the broken one)#asahi-dev · 2026-10-27 09:16
jannauit doesn't even have thunderbolt/usb4#asahi-dev · 2026-10-27 08:31
jannauI think the OUTBOX_CTRL difference exists in 12.x and 13.x as well and for other co-procs as well. I don't think it's the problem if you can start endpoints and get messages back from dcp#asahi-dev · 2026-10-27 07:29
jannauit == power button#asahi-dev · 2026-10-27 07:29
jannaudoesn't look like m4 mac mini would need soldering. you can remove it and have two nice contacts: https://support.apple.com/en-us/120996#asahi-dev · 2026-10-27 07:22
jannauand since 26.x auto-boot no longer requires that it lost power while powered-on#asahi-dev · 2026-10-26 22:34
jannauwhich devicetree are you using? https://cyberchaos.dev/yuka/linux/-/commit/ea6579a3e346e2251072014eeae1e4f7bdb81589#asahi-dev · 2026-10-26 22:20
jannauwhich wraps run_guest.py and makes it simpler to run linux under hv#asahi-dev · 2026-10-26 22:19
jannauthere is also run_guest_kernel.sh#asahi-dev · 2026-10-26 22:18
jannauSebastian[m]1: what do you mean with it's broken with macOS 27? there are mboot changes which breaks the hypervisor?#asahi-dev · 2026-10-26 22:17
jannaubut that will go away once linux.py runs#asahi-dev · 2026-10-26 22:15
jannaushouldn't the hypervisor work?#asahi-dev · 2026-10-26 22:13
jannauwhere is the tux coming from? is that a modified m1n1 logo or is that coming from linux#asahi-dev · 2026-10-26 22:03
jannauthat shouldn't make a difference for booting linux though. which m5 device is this?#asahi-dev · 2026-10-26 21:47
jannauI've tagged asahi-7.2.13-1 (not particular interesting, no downstream changes and nothing in v7.2.13 catched my eye)#asahi-dev · 2026-10-26 19:39
jannauI've tagged asahi-7.2.12-3, it fixes the dark display during boot issue (caused by the 14.x DCP firmware support in asahi-7.2...)