Repository navigation
arm64: dts: apple: t6050: apr1 on every T6050, both ave1 DART streams - #2422
RyanTheTide wants to merge 1 commit into
Conversation
The second ProRes engine (apr1) was described only when T6050_J873S was defined. The J873s is an M5 Pro Mac mini, and its ADT (the public GravityLinux adt-collection dump, macOS 27.0) has apr1 only in the children for the M5 Max and Ultra die variants (:sotra_h17c, h17cr1, h17cr2, h17d), not for the M5 Pro (h17s). Its DART, apr1_dart, is already described for every T6050. Move the node out of the board block next to its DART, name its two windows (core, wrapper, in apr0's order) and leave it disabled: unlike apr0 it has no SCODEC aux/ctrl windows, which the T6050 ProRes variant needs. ave1's iommus listed only SID 0 of its two DARTs. That ADT's dart-ave1 has the same SIDs as dart-ave0, 0 and 12 (the temporal filter), and ave0 already lists both: do the same for ave1. ave1 stays disabled. The J873s DTB gains only reg-names on apr1. Every other T6050 DTB gains the disabled apr1 node and the two ave1 streams. Signed-off-by: Ryan Murray <ryan@aurorasilicon.org>
|
Row request for M5 Max owners (@erikwb J716c, @satchlj J775c, @joshuaswarren K114c). We have no T6050 Mac in our lab, and the
Posting results here is perfect. Thank you. |
|
J775D (M5 Ultra Mac Studio) answer to row 1, plus a boot row. Row 1, from the J775D's ADT on 27.2 (26B123): all four encoders are marked preloaded. ave0 and ave1 (die 0) and ave2 and ave3 (die 1) each have Boot row with this PR, #2278 and #2291 (27.0 stub, read-only root, 4 boots):
Assisted-by: LLM (Claude:Opus 5.5) |
|
The J775D answers the first check: every encoder is marked preloaded, and the tree boots with apr1 and ave1 disabled and no errors from them. Still needed: the M5 Pro check (apr1 absent and ave1 present and preloaded in the live ADT on a J714s, J716s, J873s, K114s or K116s), then a review and ready-marking. The encode and ProRes rows belong to the follow-up that enables these nodes. |
Problem
The M5 Max has two video encoders and two ProRes engines (Apple's tech specs). In our T6050 tree the second ProRes engine,
apr1, was described only under#if defined(T6050_J873S). The J873s is the M5 Pro Mac mini, so that was the wrong gate: no Max or Ultra board had the node, and the one board that had it is a Pro. The second encoder,ave1, listed only SID 0 of its two DARTs.ave0lists SIDs 0 and 12.Change
apr1out of the board block. It now sits next toapr1_dart, which is already described for every T6050. Give itreg-names = "core", "wrapper"and keep it disabled. The ADT hasapr1only under the M5 Max and Ultra die variants (:sotra_h17c,h17cr1,h17cr2,h17d) and not under the M5 Pro (h17s). Every single-die T6050 board, Pro or Max, includes the same SoC description (t6050-max.dtsiis the one-die base despite its name). So there is no Max-only file to move the node into, and a disabled node in the shared description is the honest option.ave1:iommusnow lists SIDs 0 and 12 on both DARTs, the same asave0. The ADT'sdart-ave1has the samesidproperty asdart-ave0(0 and 12).apr1has no SCODEC windows (aux,ctrl), which the driver's T6050 variant requires.ave1needs a firmware handoff for a second encoder (see below).Testing
unit_address_vs_reg,avoid_unnecessary_addr_size,alias_paths,simple_bus_reg). Same warning count on every board, and none come from these nodes.reg-namesonapr1. Every other board gains the disabledapr1node, a phandle onapr1_dart, and the two extraave1streams. Everything else is phandle renumbering.apr1orave1. The ProRes binding already allows two windows for a T6050 instance.Still to run (M5 Max owner: J714c, J716c, J775c, K114c, K116c; Ultra J775d)
ave1? From the aura log, or from the ADT in macOS:ioreg -l -p IODeviceTree -n ave1 | grep -E 'segment-ranges|reg'. With aura#334 plus aura#341, the line to look for isFDT: AVE ... from ADT /arm-io/ave1, with twomemory-regionentries on/proc/device-tree/soc*/video-encoder@23a3100000.apr1absent from the live ADT (ioreg -l -p IODeviceTree -n apr1)? Isave1present, and is it preloaded?gst-launch-1.0 videotestsrc num-buffers=600 ! video/x-raw,format=NV12,width=3840,height=2160,framerate=60/1 ! v4l2h265enc ! fakesinkon each node at once, fps for each).apr1: needs a driver variant for the two-window "apr,n1" engine first. That variant needs a Max owner's ADTapr1node and a register dump. Not in this PR.Why there is no driver or boot loader change here
ave-drv.cfixedapple-avedebugfs directory,ave-fw.cone set of CRCs in/chosen) are already fixed generically by media: apple: ave: add the T6021 (M2 Max) encoder #2374 (ludagoo, commits "give each encoder its own debugfs directory" and "read the firmware CRC and IOVA per encoder":aurora,crc32per reservation, severaliommu-addressestuples). T6050 can reuse those as they are.ave1is matched and reserved if the ADT preloads it, and stays disabled in the DT. aura#341 adds the per-reservation CRC that two encoders with different DATA images need. With both merged and rebased, enablingave1(and its DARTs and VENC1 power states) for T6050 is a DT-only follow-up, gated by the preload the boot loader finds and not by a board list. It waits for row 1.apr1is a different engine ("apr,n1": two windows, two interrupts, its own reset). Its init rows are not known, so a variant for it would be guesswork.ave2,ave3,apr2,apr3,jpeg2,jpeg3,scaler2,scaler3) has no nodes, and die 1's VENC/PRORES/SCODEC power states are not described either. That is DT fill work for the DTFILL-T6050 lane and is left out here.ADT facts used (public J873s dump, GravityLinux adt-collection, macOS 27.0)
apr1(compatibleapr,n1, interrupts 1763 and 1762, clock gate 639): only under/arm-io/:sotra_h17cr1_42,:sotra_h17cr2_42,:sotra_h17c_48and:sotra_h17d_97. Its reg is two windows, 0x210b000000 (0x4000) and 0x210a000000 (0x1000000), which are 0x230b000000 and 0x230a000000 after arm-io's ranges.apr0(apr,n0) has five.dart-apr1is a plain/arm-iochild, as in the DT.dart-ave0anddart-ave1: both havesid = 0xc00000000, which is SIDs 0 and 12. Their registers matchave0_dart/ave0_dart1andave1_dart/ave1_dart1.Assisted-by: LLM (Claude:Opus 5.5)