#asahi-dev · 2026-08-31 16:29
yukayes#asahi-dev · 2026-08-31 16:09
yukaI guess... the M4 was running 26.6 for sure, the A18 Pro might have been 27, need to check#asahi-dev · 2026-08-31 16:04
yukaAnd then A18 Pro gets its own compatible where that is set to false (like for previous gens)#asahi-dev · 2026-08-31 16:04
yukaor something like that#asahi-dev · 2026-08-31 16:04
yukahas_zero_based_max_pend#asahi-dev · 2026-08-31 16:03
yukaSo I guess I will introduce another property on apple_nvme_hw#asahi-dev · 2026-08-31 16:02
yukaAPPLE_NVMMU_NUM_TCBS is set to 63 on both#asahi-dev · 2026-08-31 16:01
yukaIt's set to 63 in the M4 trace, and 64 in the A18 Pro trace#asahi-dev · 2026-08-31 16:01
yukait's not true#asahi-dev · 2026-08-31 16:01
yukathe other day I said the MAX_PEND_CMDS_CTRL was being set to 64 in the M4 trace#asahi-dev · 2026-08-31 16:00
yukaAha#asahi-dev · 2026-08-30 12:12
yukaI will paste it later#asahi-dev · 2026-08-30 12:11
yukaSome assertion#asahi-dev · 2026-08-30 12:11
yukaSo I need to investigate more#asahi-dev · 2026-08-30 12:11
yukaHowever the trace I have does write 64#asahi-dev · 2026-08-30 12:11
yuka(On t8132 which I'm targeting first)#asahi-dev · 2026-08-30 12:10
yukaBtw, I tested nvme again and observed what ivoszbg[m] already described: when trying to write to nvme it crashes unless I change the value written into MAX_PENDS to 63 (instead of 64).#asahi-dev · 2026-08-30 12:09
yukaWould a Tested-by help for PSCI over EFI?#asahi-dev · 2026-08-29 12:50
yukaand/or u-boot#asahi-dev · 2026-08-29 12:50
yukaYeah I can figure out how to add it to /chosen/bootargs, just wasn't sure if it would actually be picked up by the kernel#asahi-dev · 2026-08-29 12:39
yukathank you jannau#asahi-dev · 2026-08-29 12:38
yukaOf course we can't force you to do anything in your fork, but it would be appreciated so that nobody accidentally pulls or reads those commits who does not want to be involved with binary re#asahi-dev · 2026-08-29 12:37
yukaivoszbg[m] also please make it obvious in the commit message that it is binary re derived work, something like [TAINTED] prefix#asahi-dev · 2026-08-29 12:35
yukaI have a lot of nodes just commented out in mine#asahi-dev · 2026-08-29 12:34
yukapmgr is also a bit tricky on t604x and t6050#asahi-dev · 2026-08-29 12:32
yukaor something like that#asahi-dev · 2026-08-29 12:32
yukaYes, exactly. You know how it works. Now you have to document it so that I can write the upstreamable code.#asahi-dev · 2026-08-29 12:31
yukaI figured out the new spmi and sart literally by just looking and poking at it, but the irq regs are not as obvious#asahi-dev · 2026-08-29 12:29
yukaThat would be very helpful#asahi-dev · 2026-08-29 12:28
yukaWould you edit this wiki page and add the new irq regs you figured out? https://asahilinux.org/docs/hw/soc/spmi/#asahi-dev · 2026-08-29 12:27
yukaso you decompiled stuff?#asahi-dev · 2026-08-29 12:27
yukaI assumed this to be much more work#asahi-dev · 2026-08-29 12:26
yukaThe new SPMI interface needs significant work like figuring out where all the irq regs are?#asahi-dev · 2026-08-29 12:26
yukaHow#asahi-dev · 2026-08-29 12:26
yukaWait#asahi-dev · 2026-08-29 12:26
yukaI need to measure all the pstates for t8132 / t8140 / t6040 / t6050 (the machines I have access to) and put all of it onto a branch#asahi-dev · 2026-08-29 12:25
yukabut you're also still missing cpu frequency scaling, right?#asahi-dev · 2026-08-29 12:24
yukabut I'm not sure a single core is enough for software rendering#asahi-dev · 2026-08-29 12:24
yukathat's fair#asahi-dev · 2026-08-29 12:24
yukaNo way a normal person can afford these machines otherwise, right?#asahi-dev · 2026-08-29 12:23
yukaI was kindly allowed to buy it from the OpenCollective funds#asahi-dev · 2026-08-29 12:23
yukaYes#asahi-dev · 2026-08-29 12:23
yukaIf you have any improvements to them, send patches ^^#asahi-dev · 2026-08-29 12:22
yukahttps://cyberchaos.dev/yuka/linux/-/commit/9079027844a8c0477f6117b7e3c711edd7bb3f5c#asahi-dev · 2026-08-29 12:22
yukaWell, I already have a t6050 device tree in there, but only with the basics#asahi-dev · 2026-08-29 12:18
yukaivoszbg[m] but yeah, I've been trying to collect the work-in-progress device trees for M4 and M5 devices in my repo to reduce the duplicated work and friction for people experimenting with it.#asahi-dev · 2026-08-29 12:16
yukam1n1 wouldn't even let you boot that#asahi-dev · 2026-08-29 12:16
yukaI don't think you will upstream a device tree with only one cpu node#asahi-dev · 2026-08-29 12:15
yukaWell, we have the two maintainers here, but I do agree with the approach of not rushing it when there are still major issues to work out ;)#asahi-dev · 2026-08-29 12:11
yukaSo t8132 -> t604x -> t8142 -> t6050