Skip to content

kboot: T6050: hand the M5 Ultra's GPU to Linux instead of refusing its firmware build - #261

Merged
DjDeveloperr merged 3 commits into
aurora-silicon:mainfrom
funkzy:t6050-gpu-ultra-build
Nov 19, 2026
Merged

DjDeveloperr merged 3 commits into
aurora-silicon:mainfrom
funkzy:t6050-gpu-ultra-build

Conversation

@funkzy

@funkzy funkzy commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

What is missing

The T6050 GPU handoff admits the M5 Pro's (G17S) firmware builds, 26A428 and 26B8125 (sithglan's c3c72c0). On an M5
Ultra (Mac Studio, J775D, G17D) the loaded GPU firmware is a different build, on the 27.0 stub (26A428) and on the
27.2 production stub (26B123) alike, so the handoff stops at check 1 and the GPU, its power states and its mailboxes
stay disabled.

What this adds

Two rows in that handoff's build table, one per stub; no second handoff.

  • The 26A428 G17D row has the values of Linux's drivers/gpu/drm/asahi/hw/t6050.rs CONFIG_ULTRA: fw-text 0xa8000
    (the secondary role's text is 0x4000 longer than on the M5 Pro), its boot-filled ranges, and the digest of fw-text
    with them zeroed. fw-data is checked as before.
  • The 26B123 G17D row (second commit) has the same sizes and the same kinds of records as the 26A428 row, at its own
    positions (in the row). Its digest is of the 27.2 GPU firmware's text
    with those ranges zeroed; the same derivation reproduces the 26A428 G17D row's digest.
  • Each row now carries its fw-text size (and its GPU letter, for the log). The table is still tried in order and
    the digest decides; any mismatch leaves everything disabled.
  • The two G17S rows are tried as before, on 0xa4000 bytes from the start of GFX's __TEXT, whatever the ADT says
    of GFX1. The G17D rows are tried only when the ADT puts GFX1's __TEXT right after GFX's and the two reach
    0xa8000, so their check reads no further than the loaded firmware. The matched row's size is what fw-text reserves.
  • All rows enable the same seven nodes. The Ultra's GPU firmware roles, mailboxes and the power states Linux's
    J775D description uses are all on die 0.

Who runs the changed code

  • Only T6050. J775D takes the new rows.
  • On J714s and K116s the two G17S rows are tried as before, in the same order on the same 0xa4000 bytes, and write
    the same properties with the same values, whatever their ADT says of GFX1. The new code reads gfx1-asc's
    segment-ranges. Only if that puts GFX1's __TEXT right after GFX's and reaching 0xa8000 does it allocate a
    0xa8000-byte copy instead of 0xa4000 and, after both G17S rows miss, try the G17D rows as well. The log lines read
    the same.
  • The M5 Max takes its own G17 path (9ebf3ad) and does not reach this code. joshuaswarren's open kboot: T6050: accept the 27.3 (26C5035r) G17C firmware digest #172 adds a 27.3
    build to that path; the two do not overlap.

Size

clang/lld 23.1.4, RELEASE=1 CHAINLOADING=1 ESP_STAGE2=1, against main at 3d517e6: .init content +0x4, .text
content +0x80; _file_vital_end 0x11b980 (+0xf0, of which the 26B123 row is 0x70), under the 0x120000 limit with
18048 bytes of headroom (tools/stage1_room.py). aura.bin 3997696 bytes as on main.

Depends on

Nothing in git. For Linux to use it, the J775D device tree has to describe the GPU the way J714s's does (linux
pull request "arm64: dts: apple: t6050-j775d: a working GPU and CPU clusters for the Mac Studio (M5 Ultra)").
On 27.2, Linux also needs a 26B123 G17D build in hw/t6050.rs; until then its driver refuses that build at probe
and the boot goes on without the GPU.

Validation

  • Rebased onto main at 3d517e6 (from 03d120d: sptm_virt: T8142: macOS on all ten cores, and a trace of its sleep and wake #277 among the new commits, no file in common); both commits'
    diffs are unchanged, and kboot_gpu.c is the same on both bases.
  • The build above, on main at 3d517e6, with main's warnings; clang-format clean on the file.
  • The host pytest suite as CI runs it (tests/python and the two T8122 GPU tests): 750 passed, 56 skipped, as on
    main at 3d517e6. Among them the kboot_gpu.c host tests (test_t6040_gpu_fill, test_t6041_gpu_fill,
    test_siliconforest_handoff, test_m3_graphics_defaults).
  • A host harness, not part of this PR, ran this handoff and main's (03d120d, the same file as 3d517e6's), extracted verbatim, against modelled J714s and
    J775D ADTs and device trees, with a stand-in digest: the logs and device trees are identical in all sixteen J714s
    cases (26A428 accepted, 26B8125 accepted, each with and without a gfx1-asc range and with a GFX1 __TEXT 0x4000
    shorter, other build, an ADT range missing, a mailbox missing, no GPU described, die-1 nodes present, Ultra
    firmware, GFX1's __TEXT below GFX's, after a gap or overlapping it). On J775D, with its ADT's ranges, the G17D
    row reserves 0xa8000 of fw-text and enables the GPU; with GFX1's __TEXT below GFX's, after a gap, overlapping it
    or 0x4000 shorter, the G17D rows are not tried and the handoff is refused as on main.
  • The same harness with the real digest and the real firmware: the real 27.0 and 27.2 GPU firmware text, laid out as the ADT describes them, with stand-in boot data in every boot-filled range. On the J775D
    layout the 27.0 pair is enabled as "G17D 26A428" and the 27.2 pair as "G17D 26B123" (main refuses it at check 1);
    the 27.2 pair with one code byte changed, or with an ADT that gives GFX1's __TEXT 0x54000, is refused at check 1.
    On the J714s layout the 27.0 G17S pair is enabled and the 27.2 G17S pair refused, identically to main.

Hardware

These boots ran main at 03d120d with both commits and the other aura pull requests below (the same files as this
branch), booting integration/aurora at 1f38ff8f4dc9 with the linux pull request below and sithglan's 27.2
dfb_ready_query commit (merged since, in #2063).

  • J775D on the 27.0 stub (26A428), one boot: the loader reports "T6050 G17D 26A428 enabled", Linux probes the GPU
    with both firmware roles' initdata acknowledged, 30 CPUs are online, the GPU's compute, chained-submit, timestamp
    and 46-client checks and a 628-case render CTS subset (571 pass, 57 not supported, two runs) pass with no GPU
    recovery, fault or timeout, and the desktop runs on the GPU on HDMI and DisplayPort.
  • J775D on the 27.2 stub (26B123), one boot: the loader reports "T6050 G17D 26B123 enabled" and reserves 0xa8000 of
    fw-text. Linux, which has no 26B123 G17D build on integration, refuses the firmware at probe and the boot goes on
    without the GPU (30 CPUs, the display on HDMI).
  • The 26B123 G17D row confirmed on hardware: J775D on the 27.2 stub, one boot, with a kernel that admits that build:
    integration/aurora at 1f38ff8f4dc9 with sithglan's #1650 (which carries rolotrealanis98's #1173) unchanged, the
    linux pull request below, and a one-entry G17D 26B123 build in hw/t6050.rs (not part of these pull requests; it
    follows once #1173 lands). The firmware the loader admits is the build the kernel identifies, both firmware roles boot
    and both roles' initdata are acknowledged. Compute, 300 chained submits, timestamps (10 of 10), 46 concurrent
    clients, the 628-case subset (571 pass, 57 not supported) and the desktop on HDMI at 1920x1080@60 read the same as
    the same kernel on the 27.0 stub, with no GPU fault, recovery or timeout; a light render submit on die 0 takes
    32.30 us on 27.2 and 32.31 us on 27.0 (median of five passes). That run uses #1173's 27.2 values as they are; it
    shows the firmware runs with them.
  • Earlier, on main at eb66d6c plus the first commit and the other aura pull requests below, booting
    integration/aurora at fc050503ffcd with only the linux pull requests below, one boot on the 27.0 stub: "T6050 G17D
    26A428 enabled", the GPU's compute and chained-submit checks and 46 concurrent clients pass with no GPU recovery,
    fault or timeout, and the desktop runs on the GPU on HDMI and DisplayPort.
  • With an earlier form of the first commit, which tried the G17D row the same way on J775D:
    • J775D, on main at be94691 plus that form, with the kernel uploaded over USB, booting integration/aurora at
      7c71d6df6563 with the J775D GPU described (and a compute-timestamp change that is no longer in this set): the
      loader reports "T6050 G17D 26A428 enabled", Linux probes the GPU with both firmware roles' initdata
      acknowledged, a render node appears, and the GPU's compute and chained-submit checks, 46 concurrent clients and
      a 628-case render CTS subset (571 pass, 57 not supported) pass, with no GPU recovery, fault or timeout in the
      kernel log.
    • The same with the other aura and linux pull requests below on eb66d6c and fc050503ffcd, one boot: the same
      checks pass (the subset in two runs), with no GPU recovery, fault or timeout in the kernel log.
  • Not booted on J714s or K116s. dj, could you boot it once on J714s, and could someone with a K116s? The modelled
    runs say nothing changes there.

Question

On J775D the handoff also writes the leakage calibration words (9f501f3). Linux's Ultra configuration reads none.
Should the G17D rows skip them, or are they harmless to leave?

Part of the M5 Ultra platform pack

Each pull request below stands on its own; together they bring the Mac Studio (M5 Ultra, J775D) up from public
branches to the desktop.

Not covered by this pack: the USB-A ports. Their controller's driver is built in and asks for
asmedia/asm2214a-apple.bin while the kernel starts, before any root filesystem is mounted, so on this machine the
file has to be in the initramfs. The loader does not supply it.

…s firmware build

The T6050 GPU handoff admits the M5 Pro's (G17S) firmware builds: 26A428
(27.0) and, since sithglan's c3c72c0 ("kboot: T6050: admit the 27.2
(26B8125) G17S firmware of the K116s"), 26B8125 (27.2). On an M5 Ultra
(J775D, G17D) the loaded GPU firmware is a different build, so the
handoff stops at check 1 and the GPU, its power states and mailboxes
stay disabled. This adds that build to the same table; it adds no
second handoff.

The G17D row has the values of Linux's drivers/gpu/drm/asahi/hw/t6050.rs
CONFIG_ULTRA: fw-text 0xa8000 (the secondary role's __TEXT is 0x4000
longer than on the M5 Pro), its boot-filled ranges, and the digest of
fw-text with them zeroed. Each row now carries its fw-text size, and its
GPU letter for the log. fw-data is checked as before: gfx-data-base to
the end of gfx1-data-shared-ro spans 0x184000 on J775D too.

The table is still tried in order and the digest still decides. The
two G17S rows are tried as before, on 0xa4000 bytes from the start of
GFX's __TEXT, whatever the ADT says of GFX1. The G17D row is tried only
when the ADT puts GFX1's __TEXT right after GFX's and the two reach
0xa8000, so its check reads no further than the loaded firmware; only
then is the copy 0xa8000 bytes. The matched row's size is what fw-text
reserves. All rows enable the same seven nodes: the
Ultra's GPU firmware roles, mailboxes and the power states Linux's J775D
description uses are all on die 0.

Boards: only T6050 runs this code. J775D takes the new row. On J714s and
K116s the two G17S rows are tried as before, in the same order on the
same 0xa4000 bytes, and write the same properties with the same values.
The G17D row is tried there only after both miss and only if the ADT's
GFX1 __TEXT follows GFX's to 0xa8000; otherwise the one addition is a
read of gfx1-asc's segment-ranges. Their log lines read the same.

Size (clang/lld 23.1.4, RELEASE=1 CHAINLOADING=1 ESP_STAGE2=1): .init
content +0x4, .rodata +0x28, .text's section size unchanged, with
_file_vital_end 0x11b748 (+0x80) under the 0x120000 limit.

Tested: the build above; clang-format clean on the file; the
kboot_gpu.c host tests that run on the build host pass
(test_t6040_gpu_fill, test_t6041_gpu_fill, test_siliconforest_handoff,
test_m3_graphics_defaults). A host harness, not part of this change,
ran this handoff and main's, extracted verbatim, against modelled J714s
and J775D ADTs and device trees: the logs and device trees are
identical in all sixteen J714s cases (26A428 accepted, 26B8125
accepted, both with and without a gfx1-asc range and with a GFX1
__TEXT 0x4000 shorter, other build, an ADT range missing, a mailbox
missing, no GPU described, die-1 nodes present, Ultra firmware, GFX1's
__TEXT below GFX's, after a gap or overlapping it). On J775D, with its
ADT's ranges, the G17D row reserves 0xa8000 of fw-text and enables the
GPU; with GFX1's __TEXT below GFX's, after a gap, overlapping it or
0x4000 shorter, the G17D row is not tried and the handoff is refused at
check 1, as on main.

On hardware: J775D, with this and the M5 Ultra's other loader changes
on main, the kernel uploaded over USB, booting integration/aurora with
the J775D GPU described (its own pull request): the loader reports
"T6050 G17D 26A428 enabled", Linux probes the GPU with both firmware
roles' initdata acknowledged, a render node appears, the GPU's compute
and chained-submit checks and 46 concurrent clients pass, and the
desktop runs on the GPU. An earlier form of this change, which tried
the G17D row the same way on J775D, also passed a 628-case render CTS
subset (571 pass, 57 not supported) on main alone. On this machine main
starts 27 of the 30 CPUs and hands over no display yet; both are
outside this change. Not booted on J714s or K116s.

Signed-off-by: Jeff Queisser <funkzy@users.noreply.github.com>
@funkzy
funkzy force-pushed the t6050-gpu-ultra-build branch from 6fe5ba3 to 0d421b1 Compare November 18, 2026 04:44
…instead of refusing its 26B123 build

The production boot stub is 27.2 (26B123). Its GPU firmware for the
M5 Ultra (J775D, G17D) is another build than 27.0's (26A428), so the
T6050 handoff stops at check 1 there and the GPU, its power states and
mailboxes stay disabled.

Add that build as a fourth row of the same table, with 27.0's G17D
sizes (fw-text 0xa8000, fw-data 0x184000) and its own boot-filled
ranges and digest. The row is validated on the J775D with the 27.2
stub: the handoff enables the GPU and Linux runs it.

Only J775D-shaped ADTs reach the row: it is a G17D row, tried only when
GFX1's __TEXT follows GFX's to 0xa8000 and after the rows before it
miss. On every other board nothing changes.

Signed-off-by: Jeff Queisser <funkzy@users.noreply.github.com>
@funkzy
funkzy force-pushed the t6050-gpu-ultra-build branch from 0d421b1 to 182519c Compare November 18, 2026 07:47
@RyanTheTide

Copy link
Copy Markdown
Contributor

@funkzy, thanks for the J775d GPU and boot work. Four rows from a J775d would close the remaining gaps there. Nothing here blocks this PR; I'm putting them here because this is where your J775d work lives.

  1. HDMI (the J775d would be the first T6050 to drive it): an HDMI monitor modeset + hotplug x3, modetest -M apple -c showing HDMI-A connected, and aplay -l listing the HDMI audio card (a silent sink is fine).
  2. Speakers: the ADT values for the amp nodes (i2c addresses, speaker order), so we can add the sound card DT. The speakers stay at the safe level, so please don't play anything through them. Confirming that the amps probe is enough.
  3. Second engines (the Ultra has two dies): two concurrent AVE encodes on two video nodes, both at full rate. Also a scaler run, 1080p -> 720p NV12, with PSNR > 40 dB against swscale.
  4. Mesa on G17D: a dEQP-VK smoke run (pass/fail counts) with mesa#362 applied.

dmesg plus the command output, here or in a new issue, is perfect.

@DjDeveloperr
DjDeveloperr merged commit 727e92d into aurora-silicon:main Nov 19, 2026
15 checks passed
@funkzy

funkzy commented Nov 19, 2026

Copy link
Copy Markdown
Contributor Author

Thanks. Rows from the J775D (Mac Studio, M5 Ultra), one Linux boot through the 27.2 stub (26B123):

  1. HDMI. The J775D drives HDMI with kboot: T6050: give the M5 Ultra Mac Studio its displays #264. apple-drm binds both external DCPs, and HDMI-A-1 comes up at 1920x1080@60 with 13 modes (modetest -M apple -c/-e below). The HDMI port's DCP is dcpext4, on die 1.

    • Hotplug x3 is still to do: it needs someone at the machine. I'll add it here.
    • No HDMI audio card: aplay -l finds no sound cards, because the J775D tree has no DP-audio node yet. The ADT has dp-audio1..8, and dcpext4's audio points at dp-audio5 (the HDMI path is on #2449).
  2. Speakers. The amp probes: 0-0038 (sn012776) is bound to tas2764, and the jack codec 1-004b to cs42l84. Nothing was played.

    • The sound card stays deferred on its 'Headphone Jack' CPU DAI, since the AOP and its ADMAC are disabled in the J775D tree.
    • The ADT has one amp, and the integration tree already describes it with the same address and GPIOs. Values below.
  3. Second engines. Not run: both apple,t6050-ave nodes and both apple,t6050-scaler nodes are status = "disabled" in the J775D tree, so nothing probes. The modules are built, and ffmpeg's v4l2m2m encoders and v4l2-ctl are installed. The two encodes and the scaler run need those nodes enabled for the J775D first.

  4. Mesa on G17D with mesa#362, dEQP-VK (vulkan-cts-1.4.7.1):

    No GPU recovery, fault or timeout.

HDMI: modetest and dmesg
J775D HDMI on 27.2 (aura #264 display handoff). Monitor connected at boot; nothing unplugged.

$ ls -l /dev/dri/by-path
platform-2300000000.gpu-card -> ../card1
platform-2300000000.gpu-render -> ../renderD128
platform-display-subsystem-card -> ../card0
card0: driver apple-drm device display-subsystem

$ for c in /sys/class/drm/card0-*; do ...; done
card0-DP-1 status=disconnected enabled=disabled
card0-HDMI-A-1 status=connected enabled=enabled modes: 1920x1080 1920x1080 1920x1080 1280x720 1280x720 1280x720 1024x768 800x600 720x576 720x480 720x480 640x480 640x480

$ modetest -M apple -c        (libdrm 2.4.134; EDID blob and property values left out)
opened device `Apple display controller DRM driver` on driver `apple` (version 1.0.0 at 0)
Connectors:
id	encoder	status		name		size (mm)	modes	encoders
62	61	connected	HDMI-A-1       	530x300		13	61
  modes:
	index name refresh (Hz) hdisp hss hse htot vdisp vss vse vtot
  #0 1920x1080 60.00 1920 2008 2052 2200 1080 1084 1089 1125 148500 flags: ; type: preferred, driver
  #1 1920x1080 59.94 1920 2008 2052 2200 1080 1084 1089 1125 148352 flags: ; type: driver
  #2 1920x1080 50.00 1920 2448 2492 2640 1080 1084 1089 1125 148500 flags: ; type: driver
  #3 1280x720 60.00 1280 1390 1430 1650 720 725 730 750 74250 flags: ; type: driver
  #4 1280x720 59.94 1280 1390 1430 1650 720 725 730 750 74176 flags: ; type: driver
  #5 1280x720 50.00 1280 1720 1760 1980 720 725 730 750 74250 flags: ; type: driver
  #6 1024x768 60.00 1024 1048 1184 1344 768 771 777 806 64996 flags: ; type: driver
  #7 800x600 60.00 800 840 968 1056 600 601 605 628 39790 flags: ; type: driver
  #8 720x576 50.00 720 732 796 864 576 581 586 625 27000 flags: ; type: driver
  #9 720x480 60.00 720 736 798 858 480 489 495 525 27027 flags: ; type: driver
  #10 720x480 59.94 720 736 798 858 480 489 495 525 27000 flags: ; type: driver
  #11 640x480 60.00 640 656 752 800 480 490 492 525 25200 flags: ; type: driver
  #12 640x480 59.94 640 656 752 800 480 490 492 525 25175 flags: ; type: driver
65	0	disconnected	DP-1           	0x0		0	68

$ modetest -M apple -e
opened device `Apple display controller DRM driver` on driver `apple` (version 1.0.0 at 0)
Encoders:
id	crtc	type	possible crtcs	possible clones
61	60	TMDS	0x00000002	0x00000001
68	0	TMDS	0x00000001	0x00000002

(-M appledrm: "failed to open device 'appledrm'"; -D /dev/dri/card0 fails too, since -D takes a bus id, not a path.
The driver name is "apple".)

dmesg:
[    3.219551] apple-dcp 2b2e00000.dcp: DCP index:1 dptx target phy: 0 dptx die: 0
[    3.223053] apple-drm display-subsystem: bound 2b2e00000.dcp (ops dcp_comp_ops [appledrm])
[    3.223080] apple-dcp 42b2e00000.dcp: DCP index:1 dptx target phy: 4 dptx die: 1
[    3.223217] apple-dcp 42b2e00000.dcp: no boot framebuffer
[    3.226263] apple-drm display-subsystem: bound 42b2e00000.dcp (ops dcp_comp_ops [appledrm])
[    3.284447] apple-dcp 2b2e00000.dcp: DCP booted
[    3.284613] apple-dcp 42b2e00000.dcp: DCP booted
[    3.287909] apple-dcp 42b2e00000.dcp: dcp_enable_dp2hdmi_hpd: DP2HDMI HPD connected:1
[    4.268018] apple-drm display-subsystem: [drm] fb0: appledrmfb frame buffer device
[    6.330638] apple-dcp 42b2e00000.dcp: dcp_poweron: DP2HDMI HPD connected:1
HDMI audio
J775D HDMI audio, same boot.

$ aplay -l
aplay: device_list:279: no soundcards found...

$ cat /proc/asound/cards
--- no soundcards ---

$ ls /dev/snd
timer

No HDMI/DP audio card: the J775D device tree in this kernel has no dpaudio node (no node under /proc/device-tree
named dpaudio, dp-audio or hdmi-audio). The ADT does carry the DP audio blocks: dcpext4 (the HDMI port's DCP) has
`audio = d9030000` -> /arm-io/dp-audio5 (see hdmi-adt.txt).
The internal sound card (macaudio) is not registered either; see speaker-amps.txt.
Speakers: Linux and ADT values
$ ls /sys/bus/i2c/devices; name, driver, of_node per device
0-0038: name=sn012776 driver=tas2764 of_node=/soc@200000000/i2c@505014000/codec@38
1-004b: name=cs42l84  driver=cs42l84 of_node=/soc@200000000/i2c@505018000/codec@4b
i2c-0: PA Semi SMBus adapter (505014000.i2c)
i2c-1: PA Semi SMBus adapter (505018000.i2c)

$ lsmod
snd_soc_tas2764        49152  0
snd_soc_cs42l84        49152  0
snd_soc_macaudio       81920  0
snd_soc_core          262144  3 snd_soc_cs42l84,snd_soc_macaudio,snd_soc_tas2764

dmesg:
[    2.796221] cs42l84 1-004b: Cirrus Logic CS42L84, Revision: 36
[   14.314868] platform sound: deferred probe pending: snd-soc-macaudio: parsing CPU DAI of link 'Headphone Jack' at /sound/dai-link@0/codec
(tas2764 prints nothing on a successful probe; it is bound to 0-0038.)

$ cat /sys/kernel/debug/devices_deferred
sound	snd-soc-macaudio: parsing CPU DAI of link 'Headphone Jack' at /sound/dai-link@0/codec

Live DT status: i2c@505014000, codec@38 (compatible "ti,sn012776", "ti,tas2764"), i2c@505018000, codec@4b and
/sound ("apple,j775d-macaudio", "apple,macaudio") are enabled; aop@290c00000 and dma-controller@293800000 (ADMAC) are
status = "disabled". So the amp and the jack codec probe, and the sound card stays deferred: its CPU DAI is the AOP
audio node, whose parent is disabled. No card is registered.

/arm-io/i2c1                         (phandle 0x94; the speaker amp bus)
  compatible            = "i2c,t8101", "i2c,s5l8940x", "iic,soft"
  device_type           = "i2c"
  reg                   = 00400105 03000000 00400000 00000000
  interrupt-parent      = d0010000   (aic)
  interrupts            = 60050000
  clock-gates           = bf000000
  clock-ids             = 04000000
  function-device_reset = 71020000 54535241 bf000000   (pmgr)
  gpio-iic_scl          = 3c000000 02010100 41500000
  gpio-iic_sda          = 3d000000 02010100 41500000
  filter-tunable        = fff3ffff a9433343
  trigger-level-tunable = 0000ffff 00003344
  tbuf-tunable          = 0000ffff 00003444
  use-table-div         = 01000000
  #address-cels         = 01000000   (spelled so in the ADT)
  #size-cells           = 03000000

/arm-io/i2c1/audio-speaker           (phandle 0x95; the amp)
  compatible            = "audio-control,sn012776"
  device_type           = "audio-control"
  reg                   = 38000000 c4090000 00000000 00000000   (addr 0x38, 2500)
  interrupt-parent      = 59020000   (gpio0)
  interrupts            = 6a000000 01000000   (gpio0 pin 106)
  function-reset        = 59020000 4f495047 6b000000 01000100   (gpio0 pin 107)
  speaker-config        = 00140002
  bop-config            = 01320222 832d8002 06324630 02063840 3002063e 3730ffe6
  amp-dcblocker-config  = 01410000
  speaker-protection    = 00000000
  iboot-audio-volume    = (empty)
  private               = "1"

/arm-io/i2c2                         (phandle 0x96; the jack codec bus)
  compatible            = "i2c,t8101", "i2c,s5l8940x", "iic,soft"
  reg                   = 00800105 03000000 00400000 00000000
  interrupts            = 61050000
  clock-gates           = c0000000
  gpio-iic_scl          = 3e000000 02010100 41500000
  gpio-iic_sda          = 3f000000 02010100 41500000
  (the other properties as i2c1, with function-device_reset = 71020000 54535241 c0000000)

/arm-io/i2c2/audio-codec-output      (phandle 0x97; the headphone jack codec)
  compatible            = "audio-control,cs42l84"
  device_type           = "audio-control"
  reg                   = 4b000000 c4090000 00000000 00000000   (addr 0x4b, 2500)
  interrupt-parent      = 59020000   (gpio0)
  interrupts            = 86000000 01000000   (gpio0 pin 134)
  function-reset        = 59020000 4f495047 8b000000 01000000   (gpio0 pin 139)
  function-codecinput_master = 4d040000 31445541 63327061   (-> /arm-io/iop-audio-controller/audio-codec-input/audio-codec-input, compatible "audio-data,cs42l84-input")
  mic-config            = 0100ffff ffffffff
  enable-ringsense      = 01000000
  ringsense-inverted    = 01000000
  enable-dcid           = 01000000
  internal-mclk-src     = 01000000
  bits-per-sample-subset= 18000000
  samplerate-default    = 00000000 80bb0000
  PowerOffAtSleep       = 01000000
  private               = "1"

/product/audio (excerpt)
  acoustic-id           = 2f000000   (AID 47)
  enabledChannels       = 07000000

Our kernel DT matches: codec@38 on i2c@505014000 with reg 0x38, interrupt gpio 0x6a, shutdown-gpios gpio 0x6b;
codec@4b on i2c@505018000 with reg 0x4b, interrupt gpio 0x86, reset gpio 0x8b.
AVE and scaler
J775D video encoders (AVE) and scaler, same boot. Not run: neither probes.

Live DT (/proc/device-tree):
  /soc/video-encoder@2283100000  compatible "apple,t6050-ave"                       status = "disabled"
  /soc/video-encoder@23a3100000  compatible "apple,t6050-ave"                       status = "disabled"
  /soc/scaler@2285000000         compatible "apple,t6050-scaler", "apple,scaler-v23"  status = "disabled"
  /soc/scaler@23a1000000         compatible "apple,t6050-scaler", "apple,scaler-v23"  status = "disabled"

So no platform device exists for any of the four, and nothing binds:
$ ls /sys/bus/platform/devices/ | grep -E 'video-encoder|scaler|ave'
(none)
$ lsmod | grep -E 'apple_(ave|scaler)'
(none; the kernel ships apple-ave.ko, apple-ave-dma.ko and apple-scaler.ko as modules, CONFIG_VIDEO_APPLE_AVE=m,
CONFIG_VIDEO_APPLE_SCALER=m, but with the nodes disabled they are never loaded)
$ ls /dev/video* /dev/media*
ls: cannot access '/dev/video*': No such file or directory
ls: cannot access '/dev/media*': No such file or directory
$ v4l2-ctl --list-devices
(no output)
dmesg: no ave or scaler line.

Tools present (nothing installed for this): v4l2-ctl, ffmpeg (h264_v4l2m2m, hevc_v4l2m2m and the other v4l2m2m
encoder wrappers), gst-launch-1.0. The concurrent two-engine encode and the 1080p -> 720p NV12 scaler PSNR run need
the nodes enabled for the J775D first.
dEQP-VK
J775D dEQP-VK, mesa #362 r3 (386634c2da97), same boot (27.2 firmware).
deqp-vk: vulkan-cts-1.4.7.1 (releaseId 0x0d18b347). Device: Apple M5 Ultra (G17D B1), Borealis,
Mesa 26.3.0-devel (git-386634c2da).

| set                       | cases | Pass | Fail | NotSupported | other |
|---------------------------|------:|-----:|-----:|-------------:|------:|
| dEQP-VK.api.smoke.*       |     8 |    8 |    0 |            0 |     0 |
| dEQP-VK.info.*            |    23 |   22 |    0 |            1 |     0 |
| render subset (628 cases) |   628 |  571 |    0 |           57 |     0 |

info NotSupported: dEQP-VK.info.device_group_peer_memory_features ("Need a device group with at least 2 physical devices").

Render subset: all dEQP-VK.draw.renderpass.simple_draw.*, every 20th other dEQP-VK.draw.renderpass.* case and every
40th dEQP-VK.renderpass.suballocation.* case of the CTS caselist; 628 cases, list sha256 372186cfeddebfceacad9b4706bcdf27bcec382740257879839dc7e6156f21e0 (the subset
used for mesa #362's validation). The runs took 0 s, 0 s and 8 s.

Kernel counters across the three runs: GPU recovery 0, firmware restart 0, faults 0, timed out 0, WARNING/BUG/Oops 0.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants