Skip to content

arm64: dts: apple: t602x: describe the encoder DMA apertures as one block - #2373

Open
ludagoo wants to merge 1 commit into
aurora-silicon:integration/aurorafrom
ludagoo:ave/t602x-dma
Open

ludagoo wants to merge 1 commit into
aurora-silicon:integration/aurorafrom
ludagoo:ave/t602x-dma

Conversation

@ludagoo

@ludagoo ludagoo commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor

Base: integration/aurora 0abc2f385e54. One commit (8b313f4d5640 on branch ave/t602x-dma, ludagoo/linux), independent of the driver PRs.

The t602x trees describe two ranges next to each video encoder, *020000 and *044000, as DARTs. On the M2 Max (J475c) apple-dart rejects the *044000 nodes: the stream count it reads there is 385 at 40d044000 and 350 at 507044000, above its limit of 256. The *020000 nodes probe as DARTs. T600x describes the same two ranges as one apple,t6000-ave-dma block, with the encoder's DARTs at *030000 and *040000. (T603x and T8122 use a third form: *020000 as apple,dart-smmu and *044000 as apple,dapf, both disabled. T8112 has no node for either range.)

This does the same as T600x on t6020 (ave0), t6021 (ave0, ave1) and t6022 (ave0 to ave3, both dies): one dma-controller node with the two ranges, the interrupt and the power domain the old nodes had, and the encoders' iommus keep only the two real DARTs. Every node stays disabled. The binding already allows this compatible with exactly these two ranges; a maintainer may prefer an apple,t6020-ave-dma compatible in front, which the binding does not have today.

Tested

  • Mac Studio (M2 Max, J475c), kernel 7.3.0 with the T6021 encoder PRs: both encoders ran H.264 and HEVC with only the *030000 and *040000 DARTs attached, several hundred sessions, no DART faults. The blocks stay disabled there. With them enabled on that board, apple-ave-dma binds and holds the VENC sys domains on (see the J475c PR); this PR enables nothing. media/dts: apple: enable the T6000 AVE DMA fault reporter with runtime PM #1861 (draft) describes the same effect on T6000 ("once the device bound genpd counted it as active and kept the encoder's power domain up for good") and adds runtime PM to apple-ave-dma; with that, a board could enable these blocks.
  • scripts/dtc/apple-dt-check.sh 0abc2f385e54 <this commit> (dtschema 2026.9): all 137 Apple device trees build, 1404 CHECK_DTBS warning lines at base and head, none new and none removed. Re-run for this version on the eight t602x trees only (all 8 build with CHECK_DTBS, 481 schema findings at base and here, none new and none removed); the device tree files are byte-identical to the ones the full check ran on.
  • checkpatch.pl --strict: clean.

t6020 and t6022 were not booted.

Overlap: #2252 (t602x scaler, draft) and #1861 touch the same files and merge cleanly with this commit.

Not done yet

  • No board enables these blocks. Nothing binds apple-ave-dma on t602x unless a board does.

🤖 Generated with Claude Code

…lock

The t602x trees describe the ranges at *020000 and *044000 next to
each encoder as two more DARTs. On the M2 Max (J475c) apple-dart
rejects the *044000 nodes: the stream count it reads from them is 385
at 40d044000 and 350 at 507044000, above its 256 limit. The *020000
nodes probe as DARTs.

T600x describes the same two ranges as one apple,t6000-ave-dma block,
with the DARTs at *030000 and *040000. Do the same on t6020 (ave0),
t6021 (ave0, ave1) and t6022 (ave0 to ave3), and drop the two ranges
from the encoders' iommus. The binding allows exactly these two
ranges. A maintainer may prefer an apple,t6020-ave-dma compatible in
front; the binding has none today.

The encoders on J475c ran with only the *030000 and *040000 DARTs and
showed no DART faults. t6020 and t6022 were not booted.

Assisted-by: Grok 4.7
Assisted-by: Claude Opus 5.5
Signed-off-by: Lucas Goossen <lucas.goossen@hey.com>
@RyanTheTide

Copy link
Copy Markdown
Contributor

Thanks, this is a clean correction. The two ranges that apple-dart rejects become one apple,t6000-ave-dma block, as on T600x, on all three t602x SoCs, with nothing enabled.

We built all eight t602x device trees at 0abc2f385e54 and at this head. All of them build, and the dtc warnings are identical before and after. The code looks good to us. The only open question is the binding: whether to put an apple,t6020-ave-dma compatible in front of apple,t6000-ave-dma. That one is for the maintainer review, and no hardware row is needed for it.

Once this is merged, #2376 can drop its copy of the commit.

@ludagoo

ludagoo commented Nov 19, 2026 •

Copy link
Copy Markdown
Contributor Author

Thanks for the review. The updates are pushed; replies to each of your comments below.

Re: binding: apple,t6020-ave-dma in front?

Thanks for building all eight. No change from me. The blocks are the same layout as T600x's, which is why they use apple,t6000-ave-dma; if the maintainer wants an apple,t6020-ave-dma in front, I will add it together with the binding update.

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.

2 participants