#asahi-dev · 2024-09-20 10:22
chadmedcant figure out a reliable repro but sometimes the shader just... stops? and you need to disable and reenable colour management to fix it#asahi-dev · 2024-09-20 10:21
chadmedcolour management on plasma 6.2 is even more broken actually ffs#asahi-dev · 2024-09-19 12:28
chadmedi wonder how that squares with commercial interest in keeping the door open to "easy" x86 emulation#asahi-dev · 2024-09-19 11:31
jannaumoney quote from https://lpc.events/event/18/contributions/1721/ (Multi-sized THP performance benchmarks and analysis on ARM64): "We would recommend the use of a kernel with 16K page size as an optimal solution ..."#asahi-dev · 2024-09-19 11:30
jannaubah, I misremembered and thought usb3 was already working#asahi-dev · 2024-09-19 10:56
svenfwiw, in the rewrite branch not even usb3 will work right now #asahi-dev · 2024-09-19 10:23
jannauwe could probably start with supporting thunderbolt only on the first die#asahi-dev · 2024-09-19 10:22
jannauso we need the mmio resourse mapped. I don't think we can use the platform resource helpers for it unless we change the multi-die support. Can currently only think of ugly hacks orcomplicated solutions#asahi-dev · 2024-09-19 09:55
svenbut i habe to be able to read and write to that location #asahi-dev · 2024-09-19 09:54
sveniirc i just need a single mmio location #asahi-dev · 2024-09-19 09:45
nickchanokay I hacked some stuff up to have linux booting again I maybe can give the device tree series another spin tomorrow#asahi-dev · 2024-09-19 09:32
jannaudo you need full mmio access? dcp has a similar problem but it just needs to pass a pmgr mmio address to the coproc. Solved there via "pmgr_dcp" and "apple,bw-scratch"#asahi-dev · 2024-09-19 09:28
sven*required#asahi-dev · 2024-09-19 09:28
svenjannau: ugh, thats rewuired for thunderbolt #asahi-dev · 2024-09-19 08:52
nickchanit seems to be only a small change though#asahi-dev · 2024-09-19 08:52
nickchannot doing much with linux right now because in 18.1b4 apple broke the jailbreak used for booting again an I need to fix that first#asahi-dev · 2024-09-19 08:42
jannausven: I'm currently adding support to all other platforms to 20240806-atcphy-rewrite. do you need the "usb4pmgr" mmio reg? it's unused in atc.c as far as I can see and would pose a problem on multi-die systems. The mmio addresses on the 2nd die are shifted via "ranges" but the system uses only pmgr on die 0#asahi-dev · 2024-09-19 03:30
chadmedargh never mind its not fixed#asahi-dev · 2024-09-19 03:27
chadmedcolour management flickering is fixed with plasma 6.2 but not the blur effect black rectangles#asahi-dev · 2024-09-18 14:32
axboebooted and it fixes it as expected#asahi-dev · 2024-09-18 14:32
axboelol#asahi-dev · 2024-09-18 14:28
jannauI missed that patch initially because it is not in v6.11..v6.10 - doh!#asahi-dev · 2024-09-18 14:27
axboethanks jannau - I did see that commit, but for some reason missed that it isn't just removing the func != NULL check#asahi-dev · 2024-09-18 14:23
jannauaxboe: upstream regression fix pushed to asahi-wip#asahi-dev · 2024-09-18 14:02
jannaulooks like kernel regression as the sysfs visibility/writeability is checked before the use count is increased#asahi-dev · 2024-09-18 13:49
jannaucaused be6299c6e55e971ffc060495708740a59aa0e45b have to find out why psy->use_cnt is 0#asahi-dev · 2024-09-18 13:21
jannauI can reproduce on j493#asahi-dev · 2024-09-18 13:15
axboewell, just running a one-liner to default to 80 for now#asahi-dev · 2024-09-18 12:55
axboeinitial poking, ->property_is_writeable() is never called#asahi-dev · 2024-09-18 12:49
mpsaxboe: same for me with m1pro#asahi-dev · 2024-09-18 12:27
axboem2 max fwiw#asahi-dev · 2024-09-18 12:27
axboeif not known, I'll poke a bit, nothing immediately jumped out at me#asahi-dev · 2024-09-18 12:27
axboejannau: fwiw 6.11 works for me, only thing I noticed is that the charge_* configurations under macsmc-battery/ are RO rather than RW#asahi-dev · 2024-09-17 19:28
jannauI've pushed asahi-6.10.9-2 to https://github.com/AsahiLinux/linux/releases/tag/asahi-6.10.9-2 (containing a fix for PCIe on imacs)#asahi-dev · 2024-09-17 13:54
jannauwe need driver updates at least in so far as the host interface changed from i2c to spmi#asahi-dev · 2024-09-17 13:53
jannaupossibly interesting for ace3 (M3 Pro/Max and probably later): https://www.youtube.com/watch?v=QlaJQEa3_zQ (I haven't watched yet as I'm traveling#asahi-dev · 2024-09-17 13:50
aissenthanks, I'm listening to unnamed RE, but have a huge backlog, I'll bump this episode to the top of the list.#asahi-dev · 2024-09-17 12:14
jannauace2/ace3 (usb-c port controller) RE podcast https://unnamedre.com/episode/71#asahi-dev · 2024-09-17 04:17
nickchans/19562a2ca2c66030f8459a242b985dc8e690e431/f12d2897402cee92c404deb310c3d3b20676d80a/#asahi-dev · 2024-09-17 03:59
nickchanit is not required and i don't plan to send it anymore for the time being#asahi-dev · 2024-09-17 03:58
nickchannext version of the device tree patches will also drop all mention of the a10 workarounds#asahi-dev · 2024-09-17 03:56
nickchanaarch32 on a non-4K system is not an actual issue, anyways#asahi-dev · 2024-09-17 03:55
nickchanat least, I probably will not bother sending it unless I run into an actual issue with this cursed setup#asahi-dev · 2024-09-17 03:55
nickchanhttps://lore.kernel.org/asahi/ZuhPIdnx36yXJhHi@arm.com/#asahi-dev · 2024-09-17 03:54
nickchannot particularly interested in sending these though since Catalin doesn't quite like the workarounds in principle #asahi-dev · 2024-09-17 03:52
nickchangiven the MPIDR mess this wires it up as an errata in case more workaround is needed#asahi-dev · 2024-09-17 03:52
nickchanfrom commit 51d15c7c849c1745e47ada8fe2c46d35f67b952f to 19562a2ca2c66030f8459a242b985dc8e690e431#asahi-dev · 2024-09-17 03:51
nickchanjannau: since you included it in your tree here is a new version of the a10 fusion arch workarounds https://github.com/asdfugil/linux-apple/tree/apple/bits/000-cpufeature-a10#asahi-dev · 2024-09-16 13:15
mazyou may need a small hack in cpu_logical_map and co though.#asahi-dev · 2024-09-16 13:13
mazbut this is still doggy as fsck.