#asahi-dev · 2026-10-21 23:24
MichaelReeves[m]yay!#asahi-dev · 2026-10-21 22:34
kit_ty_kate❤️#asahi-dev · 2026-10-21 22:27
jannauI've tagged asahi-7.2.11-1 and pushed it to github#asahi-dev · 2026-10-21 22:00
jannauwe're will be at M17 or so before we save a single megabyte (with a single binary for all versions / tiers). avd-fw is tiny (64k). for me it's more about the number. I'd be happy with one binary per version#asahi-dev · 2026-10-21 21:39
JamesCalligeros[m]When we still have to carry gigabytes of useless firmware from every other vendor anyway#asahi-dev · 2026-10-21 21:39
JamesCalligeros[m]In service of saving a couple mb#asahi-dev · 2026-10-21 21:38
JamesCalligeros[m]The firmware gets uploaded on every boot so “a few reads” is actually “a few reads every time a stream starts and stops”#asahi-dev · 2026-10-21 21:37
JamesCalligeros[m]I dont think i really like this approach#asahi-dev · 2026-10-21 21:10
sofushuh i remembered it as more, there is also the smaller address space quirk, but thats peanuts#asahi-dev · 2026-10-21 21:01
jannauI see one use to define a different value for AVD_OP_EXEC_FIFO_MASK (AVD_VER == 5 && AVD_TIER == 0)#asahi-dev · 2026-10-21 20:55
sofusbut i dont think its enough to matter#asahi-dev · 2026-10-21 20:54
sofusthat sould be for the neo. in avd-next the tiers are used a few other places i think#asahi-dev · 2026-10-21 20:52
jannauoh, and a different SP for v5 tier 0#asahi-dev · 2026-10-21 20:51
jannauaren't the tiers be just different tunables?#asahi-dev · 2026-10-21 20:47
sofuswe could collapse the tiers at the cost of a few extra reads. We could do the same for the versions but i would prefer not to (most registers differ)#asahi-dev · 2026-10-21 20:35
jannausaves space on millions of disks, makes it easier to fill CONFIG_EXTRA_FIRMWARE#asahi-dev · 2026-10-21 20:33
jannaupushed the reworked firmware change. the idea would be to add version / tier fields to `struct avd_config`, have the kernel driver fill them out and use that in the firmware to use the same binary for multiple tiers (and maybe versions)#asahi-dev · 2026-10-21 20:31
sofusUnkown on m1, but otherwise yes. What would the benefit be by reducing the amount?#asahi-dev · 2026-10-21 20:03
jannausofus: are there registers that identify the avd version / tier? if not what do you think of providing the information via firmware (written by the kernel driver). The idea bind that would be to reduce the number firmware binaries#asahi-dev · 2026-10-21 17:41
jannauyes, found it as well#asahi-dev · 2026-10-21 17:40
chaos_princesshttps://github.com/AsahiLinux/linux/blob/asahi/drivers/soc/apple/apple-pmgr-misc.c#L89 - here#asahi-dev · 2026-10-21 17:38
jannaumaybe it's #address-cells / #size-cells although the node has no children#asahi-dev · 2026-10-21 17:37
jannauI don't see how that makes a difference though. I don't see there that it parsed#asahi-dev · 2026-10-21 17:35
jannauapparently caused by removing "apple,dcs-min-ps = <7>;" from pmgr_misc#asahi-dev · 2026-10-21 17:06
jannaususpend/resume is broken on m1 max #asahi-dev · 2026-10-21 10:56
MichaelReeves[m]it is p3s#asahi-dev · 2026-10-21 10:56
MichaelReeves[m]yes#asahi-dev · 2026-10-21 10:56
Alex0[m]Was warthog the p3s related one or can you remind what was for?#asahi-dev · 2026-10-21 10:52
MichaelReeves[m]I expect it should work fine, but will probably need some changes as MM changes frequently#asahi-dev · 2026-10-21 10:52
MichaelReeves[m]when asahi-wip is tagged, I'll rebase on it and make warthog-7.2#asahi-dev · 2026-10-21 10:47
chadmedUSB is also working again #asahi-dev · 2026-10-21 10:46
chadmedjannau: gpu driver looks fixed on j415 here now. still issues with dmabuf-wayland on kwin (and dwl) but i believe these are unrelated to the kernel driver #asahi-dev · 2026-10-21 10:37
chadmedcan do so now #asahi-dev · 2026-10-21 10:35
jannaunot sure what the usb-c/usb4/tb issue was but I can no longer reproduce it#asahi-dev · 2026-10-21 10:34
jannauI intend to tag the current asahi-wip (9907031896a7) later today. please test#asahi-dev · 2026-10-21 10:19
chadmedone problem at a time#asahi-dev · 2026-10-21 09:53
sobek[m]https://github.com/galadran/we-have-a-problem.com/issues/45 (not to mention outdated macOS versions)#asahi-dev · 2026-10-21 09:53
sobek[m]I think the example of the compromised intermediate certificates, where HTTPS isn't quite trustworthy anymore outside browsers (macos curl git or python all use a currently out of date root of trust, only safari / web browser use a properly fixed stack), show why curl https:// | sh isn't good enough for security, and we kind of need a better way of ensuring legitimate provenance.#asahi-dev · 2026-10-21 08:43
JamesCalligeros[m]Was anyone able to test 96khz playback on a 13” m1 macbook pro#asahi-dev · 2026-10-21 08:04
maznow, if you have a solution that would avoid trapping all that pile of crap, I'm all ears.#asahi-dev · 2026-10-21 08:01
maznickchan: you have the wrong end of the stick. If an SError can be generated in a guest and not be synchronised by an ESB, this is a HW bug, and TGE has nothing to do with that.#asahi-dev · 2026-10-21 06:57
chadmedwell, ive been compiling to bytecode to test obvious things but havent tried to do any installs yet#asahi-dev · 2026-10-21 06:56
chadmedwhatever's on gh is fully untested btw#asahi-dev · 2026-10-21 06:56
chadmedill get around to it eventually, i dont think theres that much pure refactoring to do #asahi-dev · 2026-10-21 06:53
MichaelReeves[m]nickchan: ah yes good point, maybe it's OK to send all of them then. jannau was the one who said only send tested SoC's so I'd like to hear what they say first though#asahi-dev · 2026-10-21 06:52
MichaelReeves[m]chadmed: no problem! thank you for helping make it#asahi-dev · 2026-10-21 01:57
nickchannot going to fight anything IMPDEF in the kernel atp so add every MIDR :/#asahi-dev · 2026-10-21 01:56
nickchanBut linux doesn't do that (yet) so the solution becomes adding every MIDR#asahi-dev · 2026-10-21 01:56
nickchanas Discussed previous the vgic behavior on M2+ is not really an errata but the necessary result of HCR_EL2.TGE being set between the guest does a SError-genereating action and that SError being ready to be delivered and the proper way to handle it is to check for IMPLEMENTATION DEFINED symdrome in ESR_EL1#asahi-dev · 2026-10-21 01:56
nickchanthe midrs