#asahi-dev · 2026-09-06 13:30
yukathe code wouldn't work then I think#asahi-dev · 2026-09-06 13:29
yukam1n1 nvme.c calls 0x100c NVME_DB_IOCQ too, so I guess the tracer is wrong#asahi-dev · 2026-09-06 13:27
yukaOn my M3 Max it seems to actually use approximately 256 entries#asahi-dev · 2026-09-06 13:25
yukalinux calls it APPLE_ANS_IOCQ_DB, m1n1 tracer calls it NVME_REG_DBS_IOSQ#asahi-dev · 2026-09-06 13:25
yukaI'm also confused if 0x100c is related to the submission or the completion queue#asahi-dev · 2026-09-06 13:24
yukanoopwafel: thanks for confirming#asahi-dev · 2026-09-06 13:23
yukaIn the base M3 I see 65 (!) entries of the IO queue actually being used (0x0 - 0x40)#asahi-dev · 2026-09-06 13:22
yukaboth set APPLE_ANS2_MAX_PEND_CMDS_CTRL = 0x1000100 (256)#asahi-dev · 2026-09-06 13:21
yukaFirst thing I see is M3 seems to actually have a larger queue depth#asahi-dev · 2026-09-06 13:21
yukaI looked at two traces, one from my M3 Max with 14.x fw and one from https://noopwafel.net/asahi/ (I think it's M3 with 14.x fw)#asahi-dev · 2026-09-06 13:20
yukaNVMe is super confusing#asahi-dev · 2026-09-05 19:43
yukaBut on the other hand I do not think it is easier to read for a human when the properties are fully sorted, e.g. d-cache-size anf i-cache-size have some other thing in between#asahi-dev · 2026-09-05 19:42
yukawith dt-check-style, there are less different ways to represent the same thing and makes it diffable#asahi-dev · 2026-09-05 19:41
yukaLooking into it again I see it as a machine readable vs human readable tradeoff#asahi-dev · 2026-09-05 14:29
yukasounds like a good idea#asahi-dev · 2026-09-05 12:29
yukaack#asahi-dev · 2026-09-05 12:29
yukait needs the M4 series which touches all the same files and is only in the apple-soc or linux-next for now#asahi-dev · 2026-09-05 12:29
yukathe series does not apply on v7.3-rc1#asahi-dev · 2026-09-05 12:25
yukaOkay, then I just use linux-next as base?#asahi-dev · 2026-09-04 20:12
yuka(which means exactly nothing xD)#asahi-dev · 2026-09-04 20:12
yukaI can see them presenting it as: "It's the best MacBook Neo we've ever released"#asahi-dev · 2026-09-04 20:11
yukathere's the AirPods Max (that's how it was called when it was released), AirPods Max (still the same name, but now with USB-C), and then AirPods Max 2#asahi-dev · 2026-09-04 20:10
yukathey did it before: https://www.engadget.com/audio/headphones/apple-refreshes-its-airpods-max-with-new-colors-and-usb-c-174330979.html#asahi-dev · 2026-09-04 20:05
yukachaos_princess so you think "MacBook Neo (A18 Pro, 2026)" would be more clear in case they decide to release "The New MacBook Neo"?#asahi-dev · 2026-09-04 19:58
yukaI guess it's inconsistent in this version: the comment in the dt-bindings has it with the additional info, and the device tree itself has it without#asahi-dev · 2026-09-04 19:58
yukaYes, but I didn't put that in the model string#asahi-dev · 2026-09-04 19:56
yukaI guessed that the device tree model strings were based off the "about this mac" text, and went with just "MacBook Neo"#asahi-dev · 2026-09-04 19:56
yukafair#asahi-dev · 2026-09-04 19:54
yukado you think they will change this when they release the MacBook Neo (2027) or so?#asahi-dev · 2026-09-04 19:54
yukait just says "MacBook Neo"#asahi-dev · 2026-09-04 19:54
yukaiirc the macbook neo's "about this mac" screen does not show an soc or year#asahi-dev · 2026-09-04 11:43
yukaand pmgr is a bit too basic to leave out ;)#asahi-dev · 2026-09-04 11:43
yukasven I won't send anything for t604x because I don't understand the pmgr situation there#asahi-dev · 2026-09-04 09:44
yukaack#asahi-dev · 2026-09-04 09:41
yukasven jannau do you have a strong opinion on which device tree work should go next? I have the A18 Pro / MacBook Neo minimal (same as M4) device tree lying around and could send it, but it would technically be skipping M4 Pro from the release order.#asahi-dev · 2026-09-04 09:14
yukainteresting, I didn't realize it was broken#asahi-dev · 2026-08-31 19:50
yukathere are still a lot of differences in the trace which I can try to apply but I'm done for today#asahi-dev · 2026-08-31 19:49
yukalowering the max_queue_size to 63 results in the same errors at size: 64, tail/head: 63/62#asahi-dev · 2026-08-31 19:49
yukasorry about the misinfo earlier, macos sets MAX_PENDS to 0x400040 on both t8132 an t8140#asahi-dev · 2026-08-31 18:27
yukawhy is size 65 though#asahi-dev · 2026-08-31 18:26
yukais the tail 64 maybe the problem?#asahi-dev · 2026-08-31 18:25
yukaI/O CQ : 0x1001880c000, size: 65, tail/head: 64/63#asahi-dev · 2026-08-31 17:59
yukais the full error when writing#asahi-dev · 2026-08-31 17:59
yuka"assert failed: [7454]:CQ (Host I/O) DB error, status_reg: 0x4, head: 63, valid_status: 0x4, err_info_0: 0x1, err_info_1: 0x20000"#asahi-dev · 2026-08-31 17:56
yukaI think either value will work fine, if we write it to all the places, but I'll try that now#asahi-dev · 2026-08-31 17:56
yukaI think we're just being inconsistent with how we use the max_queue_size#asahi-dev · 2026-08-31 17:52
yukaso 63#asahi-dev · 2026-08-31 17:51
yukam1n1 sets NVME_MAX_PEND_CMDS_CTRL to "((NVME_QUEUE_SIZE - 1) << 16) | (NVME_QUEUE_SIZE - 1))"#asahi-dev · 2026-08-31 17:51
yukait was a trace of m1n1 in m1n1#asahi-dev · 2026-08-31 17:50
yukaThe trace showing MAX_PENDS = 63 on t8132 was not a trace of macOS at all