#
✕ nick:jannau
#asahi-dev · 2022-05-26 18:43
jannauRBEP_SHUTDOWN for dcp iboot can apparently block the mailbox send for ~33 us while m1n1 still receives syslog messages
#asahi-dev · 2022-05-26 17:34
jannauVinDuv: not related to https://github.com/AsahiLinux/m1n1/pull/183 the display never disconnects
#asahi-dev · 2022-05-26 17:33
jannauI think the display is to blame for this
#asahi-dev · 2022-05-26 17:33
jannauthere's nothing in the syslog messages indicating the first mode set did not work. the second modeset apparently reset the hdmi connection
#asahi-dev · 2022-05-26 17:31
jannau"fixed" by setting the mode twice with 2000 ms delay
#asahi-dev · 2022-05-26 14:41
jannauthere seems to be a problem dcp shutdown in m1n1. without RTKIT_SYSLOG dcp crashes quite often on shutdown when the display didn't turned on, I haven't seen that with RTKIT_SYSLOG yet
#asahi-dev · 2022-05-26 14:38
jannaus/shartly/shortly/
#asahi-dev · 2022-05-26 14:38
jannauthe monitor seems to blame. seems to be an timing or power safe issue. other modes work shartly after powering the display on or if the display is initialized
#asahi-dev · 2022-05-26 13:48
jannauand I haven't tested any modes below 1280x720
#asahi-dev · 2022-05-26 13:47
jannauI get 47 modes, those two modes (ignoring the refresh rate) are the only modes that work
#asahi-dev · 2022-05-26 12:49
jannauso the display still works after display power switch / hdmi unplug / replug
#asahi-dev · 2022-05-26 12:48
jannausetting 1920x1200 with an too smalll fb seems to crash dcp in an interesting way. display scanout still works and nothing stops it on hdmi hotplug
#asahi-dev · 2022-05-26 12:43
jannauthe iboot allocated frame buffer on the m1 ultra is however only large enough for 1920x1080
#asahi-dev · 2022-05-26 12:42
jannausigh, I don't know what should blame more. dcp iboot or my dell UP2414Q. 3840x2160 and 1920x1200 are the only modes which can be set in m1n1
#asahi-dev · 2022-05-24 23:06
jannaumarcan: https://github.com/jannau/linux/commit/8c2b2b7d7281c09f7fb721fa0eeb6dab5116bd07 unbreaks t6002 with dwc3 as module
#asahi-dev · 2022-05-23 17:36
jannauakick #asahi ADD jokerface
#asahi-dev · 2022-05-23 17:17
jannauj`ey: \o/ - tools/dump_pmgr.py is more readable when you know what you're looking for
#asahi-dev · 2022-05-21 10:45
jannausigh, I forgot that I skipped multi-die pmgr handling in hv.py
#asahi-dev · 2022-05-21 10:37
jannauand broke at least trace_dcp.py
#asahi-dev · 2022-05-18 19:02
jannauios major
#asahi-dev · 2022-05-18 16:33
jannaudone thanks
#asahi-dev · 2022-05-17 23:09
jannauonly the merged tree tested so far (single boot on the mac studio)
#asahi-dev · 2022-05-17 23:07
jannaustill missing dt-bindings updates for mac studio (and the macbooks if we haven't done it already)
#asahi-dev · 2022-05-17 23:06
jannauI splitted commits between t600x and t8103 to make the rebasing simpler. t600x is splitted so that we could submit the devicetrees for all devices for 5.19, i.e. nvme is in, cpufreq is not
#asahi-dev · 2022-05-17 23:03
jannaumarcan: https://github.com/jannau/linux/tree/bits/000-devicetree-t6002 rebaed bits/000-devictree with t6002 support
#asahi-dev · 2022-05-17 12:20
jannauthe physical address should be in /carveout-memory-map
#asahi-dev · 2022-05-17 12:20
jannaufor DCP I see 32 bytes per segment, first 8 byte seems to be physical address in 32-bit cells, little endian
#asahi-dev · 2022-05-16 06:54
jannauI want to switch the die macros to upper case, need to check comments for paste errors
#asahi-dev · 2022-05-16 06:47
jannaumarcan: what's the best way to get the ulta dts into asahi/main? commits on top of bits/000-devicetree will get annoying fast. I can prepare a bits/000-devicetree branch which has t600x multi-die support from the beginning
#asahi-dev · 2022-05-16 06:26
jannauoh, I have to implement translation via "ranges" manually with libfdt. so much for fixing dt_disable_missing_devs() quickly
#asahi-dev · 2022-05-15 13:20
jannaupmp seems to fused off on die 1 or at least pmgr knows that it's not to be turned on
#asahi-dev · 2022-05-15 09:56
jannautook me a momnet to realize why the hypervisor usb uart was dropped after I moved the it to /soc/die0
#asahi-dev · 2022-05-15 09:55
jannaumarcan: do you have a preference where devices from die 0 should live for t6002? "/soc" would have the advantage that the resulting devicetree is closer to t6001, "/soc/die0" would look more consistent. src/kboot.c is the main place where it makes a difference
#asahi-dev · 2022-05-13 17:55
jannauthe ones from the m1 max should work https://github.com/AsahiLinux/linux/pull/22/files
#asahi-dev · 2022-05-13 17:07
jannauhw doesn't differ that much. ethernet and sd card reader are the only things I can think of which use separate module
#asahi-dev · 2022-05-13 17:01
jannauthe asahi distro kernel works with the t6002 dts out of the box
#asahi-dev · 2022-05-13 17:00
jannauno
#asahi-dev · 2022-05-13 16:59
jannauit has most stuff built-in for booting via m1n1
#asahi-dev · 2022-05-13 16:58
jannaucd321x is apple's TPS6598X
#asahi-dev · 2022-05-13 16:57
jannaumaz: my dev config if you want to compare https://www.jannau.net/asahi-t6002.config
#asahi-dev · 2022-05-13 16:51
jannaui2c
#asahi-dev · 2022-05-13 16:40
jannauwhich kernel? if upstream it's missing support for the iommu variant? otherwise do you have CONFIG_TYPEC_TPS6598X ?
#asahi-dev · 2022-05-13 16:35
jannaumaz: I think so, host breaks iirc due to typec interaction
#asahi-dev · 2022-05-13 15:37
jannaus/might/probably/
#asahi-dev · 2022-05-13 15:36
jannaumaz: ^^^
#asahi-dev · 2022-05-13 15:36
jannauyou might need https://patchwork.ozlabs.org/project/uboot/patch/20211224232216.56847-1-kettenis@openbsd.org/
#asahi-dev · 2022-05-13 09:47
jannauinstaller builds with ./build.sh, you have to adjust the host in scripts/bootstrap.sh
#asahi-dev · 2022-05-13 09:46
jannaumaz: with https://paste.debian.net/1240766/ the stand-alone m1n1 installation should work
#asahi-dev · 2022-05-12 22:04
jannauI also wonder whether of_translate_dma_address will this for dcp. at least missing "dma-ranges" are handled as identity translation
#asahi-dev · 2022-05-12 21:29
jannauit's just a little annoying that it has to be parsed twice

plus ›