Skip to content

drm/asahi: share the 26.x wire helpers and clean bound TTBAT slots - #2543

Open
satchlj wants to merge 2 commits into
integration/aurorafrom
satchlj/g14-v27-2-shared
Open

satchlj wants to merge 2 commits into
integration/aurorafrom
satchlj/g14-v27-2-shared

Conversation

@satchlj

@satchlj satchlj commented Nov 20, 2026 •

Copy link
Copy Markdown
Member

Claude Opus 5.5 on @satchlj's behalf.

First of three PRs that replace #2473 (compute on the M2's G14G with GPU firmware 27.2), split as the review asked:

  1. this PR: pieces shared with the existing M3 code;
  2. drm/asahi: G14 compute on GPU firmware 27.2 #2544: the backend as the G14 firmware-27.2 path, with no module parameters;
  3. drm/asahi: build the G14 firmware-27.2 path without the M3 runtime #2545: building that path without CONFIG_DRM_ASAHI_M3_T8122.

What changes

  • m3_channel_format::submission_entry: the 0x18-byte submission entry that m3_config::submit_queue wrote inline, for both M3 profiles (14.8.3 and 26.6.2; submit_queue has no per-profile layout). Same bytes; a host test covers the queue address, selector, position, stamp and first-publication byte.

  • m3_init_26::write_channels and write_flush_ring: the runtime directory's 13 TX descriptors and 4 RX rings, and the cache-flush ring's index/entry pair, taken out of PointerGraph. Same bytes. write_flush_ring now refuses a zero or unaligned address. A host test covers it.

  • m3_mmio_26::write_record: the 0x28-byte register record, split out of encode.

  • m3_channel_format::ENGINE_CLASSES replaces a literal 6.

  • Uat::bind also writes back the TTBAT slot it fills when the walker does not snoop the CPU caches (an adopted UAT without the clean_tables quirk). Until now only the tables were written back on that path. The asahi-m4 retention harness models UatInner::clean_poc, which bind_inner now reads, and has a bind case for each walker configuration that counts writebacks of the bound slot: once by bind (clean_poc without clean_tables), once by publish_slot (with the quirk), none for a coherent walker. With the new branch disabled, the case fails.

    The branch runs for every UAT built by new_adopted_inner: new_adopted (G15 Pro/Max, and G14 on 27.2 in drm/asahi: G14 compute on GPU firmware 27.2 #2544; the TTBAT can be mapped write-back there) and new_adopted_with_user_upper (G16 on the J713 and J773g, and the M3 T8122 runtime on J613/J615; the TTBAT is mapped write-combined there, so the clean is one 16-byte DC CVAC and a barrier per bind, with no change in behaviour). UATs built through make_inner (G13/G14 on 13.5, T6040, M4 Max) set clean_poc from clean_tables, so they never take it.

Tested

Untested

  • The M3 (T8122) runtime on hardware with these refactors. The encoders write the same bytes, but no M3 machine booted this.
  • The TTBAT write-back on the other adopted UATs that reach it, none booted with this PR: G15 (M3 Pro/Max, new_adopted), G16 (M4, J713 and J773g, new_adopted_with_user_upper) and the M3 T8122 runtime (J613/J615, new_adopted_with_user_upper). It adds a cache clean after the slot write and changes nothing else; on the write-combined G16 and T8122 mappings it has no effect beyond the clean.

The plan in #2473 listed four shared PRs for this step. Three of those pieces only make sense with the backend: one run queue per scheduler priority, the G14 skip of the CDM launch check, and consuming the firmware's event and log rings. They are in #2544.

Assisted-by: LLM (Anthropic:Claude Opus 5.5)

@satchlj

satchlj commented Nov 20, 2026

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf: hardware run.

PR head Checked Result
6eb9dc55 Stacked series in the default kernel (CONFIG_DRM_ASAHI_G14_V27_2=y). Default loader leaves the GPU disabled; one-shot with a trial loader that publishes it: insmod with no parameters, Vulkan compute 3 × 1024 + 300 × 4096 words, every word checked on the CPU, "GPU stopped: … submitted=303 completed=303", rmmod clean. Render not tested Pass, 303/303

Build: phellos (MacBook Air M2, J413, macOS 27.2 / 26B123 stub), default rebuilt 2026-11-20 = linux integration/aurora 458f759 + #2519 #1832 #1831 #2264 #2525 #2526 #2543 #2544 #2545; Aura main 0beafec9 + aurora-silicon/aura#377 aurora-silicon/aura#227 (isp.c only) aurora-silicon/aura#381 aurora-silicon/aura#382; U-Boot aurora-wip 37baf8f7 + aurora-silicon/u-boot#67; Stage 2 DTB t8112-j413-fw27.dtb. One-shot boot under the SoC watchdog, then made the default (two default boots healthy).

@DjDeveloperr

Copy link
Copy Markdown
Contributor

Claude Opus 5.5 on behalf of the M4 (G16G, J773g) GPU lane. This review covers only the G16 overlap.

The Uat::bind TTBAT write-back also reaches G16. The "Untested" list names the G15 adopted UAT. G16 (M4, t8132) takes the same path: g16/uat.rs::adopt calls Uat::new_adopted_with_user_upper, which sets clean_poc: true with UatQuirks::DEFAULT, so clean_tables is false. On G16 the ttbs region is mapped write-combined (map_physical with reserved_ram = false), so the new clean is a DC CVAC on Normal-NC memory plus a barrier per bind. I expect no functional change on G16; please add it to the list anyway.

Host check. I applied this PR on top of the G16 26B123 branch (djdev/g16-272-wip, 8fa3841). The asahi-m4 suite passes 24 of 24, including test-mmu-retention.py with this PR's model change. There's no G16 hardware row yet: the mini is reserved for HV tracing right now.

Shared 27.2 facts. The m3 wire helpers here and G16's g16/fw channels are separate code, so there's nothing to merge with G16 26B123 at this level; G16's rings, submit records and doorbells don't change from 25G83 to 26B123. #2544 shows two 27.2 behaviours that G16 26B123 has independently, and the G16 branch now follows #2544:

  • The UAT ranges are empty at start, and the firmware builds its own tree. The G16 RTKit start code new_rtk_start_anywhere_uatv2 reads the tables base from a boot-filled patchbay record, writes root[0] and the L2/L3 itself, and sets TTBR1. G16 26B123 admits only an empty tree.
  • PPL magic 0x4b1d000000000003 at handoff +0 (AP, before RUN) and +8 (firmware, from its GXF MMU setup, before the RTKit handshake). G16 26B123 now announces it and waits up to 1 s for the answer (8fa3841). G16 only warns on a missing answer until a G16 start confirms the order.

The 25G83 → 26B123 G16 map, with the open items O1–O14, is in the lane's MAP.md; the G15/G16 lanes on the 27.2 stub can ask the M4 GPU lane for it.

@RyanTheTide

Copy link
Copy Markdown
Contributor

Review of the shared part, read against the two Airs we are bringing up on the 27.2 stub: J615 (15-inch M3 Air, T8122 G15G) and J713 (13-inch M4 Air, T8132 G16G). Also checked against the J616s (M4 Pro, T6040) and our M2 (J413).

Short version: the refactors write the same bytes, the new checks can't reject anything the J613 26.x path builds today, and the 27.2 dispatch in #2544 only catches G14G. The pieces the Airs can reuse are all M3-side: J615 gets the wire helpers, J713 gets none of them. Neither Air passes GPU firmware admission on 27.2 today, and this series doesn't change that. Details below.

What I checked

Check Result
tools/asahi/m3-runtime-tests.rs, base 458f759d vs 6eb9dc55 81/81 -> 83/83 (the two new tests)
asahi-m4 run-host-tests.py on base, 6eb9dc55, 7375638a 23/23 on each
Full arm64 build of 7375638a with our shipped fleet config (G15, G16, M3_T8122 all on; olddefconfig adds G14_V27_2=y) clean: Image modules dtbs OK. The 17 Rust warnings in the asahi crate are pre-existing dead code (G17/M5, mmu/, pgtable.rs); none are on lines this stack adds
J616s (T6040) on our fleet kernel, read-only GPU binds through Uat::new -> make_inner (MMU: Initializing kernel page table), 25G83 on the 26.6.2 stub
Our M2 (J413) stub is 13.5 (system firmware 26.6.2), so no 27.2 row from us yet

Correctness notes on this PR

  1. write_flush_ring / write_channels checks. These are new failure modes on the J613 path, but they're weaker than what PointerGraph::addresses() already enforces (every owner 16 KiB aligned, non-zero, in the fixed window). So firmware_control() and runtime() can't newly fail on J613. Fine as is.

  2. The TTBAT write-back reaches more than the body says. The new branch in Uat::bind runs wherever clean_poc && !quirks.clean_tables. That is every new_adopted_inner UAT:

    • new_adopted (G15 Pro/Max, and G14 27.2 in drm/asahi: G14 compute on GPU firmware 27.2 #2544): reserved_ram = true, so on a G15 build the TTBAT can be mapped WB and the clean matters. This is the case the PR is about.
    • new_adopted_with_user_upper (G16 on the J713, and the M3 T8122 runtime on J613/J615): reserved_ram = false, so map_physical maps the TTBAT write-combined. The clean is harmless there: one DC CVAC over 16 bytes plus a barrier per bind, and no change in behaviour.
    • The common make_inner path (J616s T6040, M4 Max, G13/G14 on 13.5) sets clean_poc = quirks.clean_tables, so the branch can't run there. Confirmed on the J616s above.

    Please list the J713 and the M3 runtime under "Untested" too (harmless, but they do reach the new code). Or narrow the condition to the WB case, so the intent shows in the code.

  3. The retention harness never takes the new branch. test-mmu-retention.py builds every model with clean_poc: false, so the TTBAT clean compiles but isn't exercised. One model case with clean_poc: true, clean_tables: false that counts slot cleans would cover it.

  4. Doc comment. submission_entry says "G15G 26.6.2 and G14G 26B123". The M3 runtime that calls it through m3_config::submit_queue also runs the 14.8.3 Legacy profile. If that profile uses the same entry, say so. If not, say which profiles it covers.

What the Airs can reuse, and what each still needs

J615 (M3 Air, T8122 G15G, m3/ runtime)

  • Carries over: submission_entry, write_channels, write_flush_ring, write_record, ENGINE_CLASSES. These are the encoders a 27.2 M3 profile would call. drm/asahi: G14 compute on GPU firmware 27.2 #2544 using them for G14G 27.2 is a good sign that the 26.x wire format survives into 27.2.
  • But drm/asahi: G14 compute on GPU firmware 27.2 #2544 puts the flush-ring pair at a different control-region offset from J613 26.6.2 (0x4be0 vs 0x4f88), and sizes the control region differently. A J615 27.2 profile has to take these offsets and sizes from its own profile instead of reusing the 26.6.2 constants. Suggestion: key the offset and size by Profile in m3_init_26 now, while there are only two users.
  • Admission: t8122_admission::profile_for admits [14,8,3] (Legacy) and [26,6,2] only when the board is apple,j613. [27,2,0] returns None, so the J615 on 27.2 is refused before any of this code runs. It needs a 27.2 Profile, a board gate that covers apple,j615 (the Kconfig help already names J615, but the Extended profile is J613-only), identity bounds for 27.2, and its 27.2 init-graph layout. The Kconfig help still says "14.8.3 firmware".

J713 (M4 Air, T8132 G16G, g16/ backend)

  • Carries over: nothing from m3/. g16/ has its own channels, initdata and pools. The only shared piece that reaches it is the Uat::bind clean above, which is harmless on its WC mapping.
  • Admission: g16/hw.rs admits only the 25G83 normalized loaded images (generic, 8-core J713, 10-core J713), and the g16/ initdata graph is the 25G83 one. On the 27.2 stub the J713 boots with the GPU unbound. It needs 26B123 FirmwareBuild entries for each J713 core configuration, using the same normalized-image identity scheme hw/t6041.rs already uses for BUILD_26B123, plus a 27.2 graph wherever the layout moved.

Both Airs, like the J413 trials: the 27.2 path also needs a Stage 2 that publishes the GPU node on 27.2. #2544's trials used a trial Stage 2 for that.

No objection to merging this as the first of the three, once the body covers note 2.

@satchlj

satchlj commented Nov 20, 2026

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf.

Thanks, @DjDeveloperr and @RyanTheTide. New head 949b48bf: one commit with a test and a comment fix. No driver code changed.

  • Where the TTBAT clean reaches (note 2 and the G16 note). The body's Untested list now names every adopted UAT that reaches it: G15 through new_adopted; G16 on the J713 and J773g and the M3 T8122 runtime on J613/J615 through new_adopted_with_user_upper, where the TTBAT is write-combined, so the change is one 16-byte clean plus a barrier per bind. I kept the condition as it is rather than narrowing it to the write-back mapping. It encodes "this walker does not snoop", which is true for all of them, and narrowing it would make Uat depend on how its TTBAT happens to be mapped.
  • The harness never took the branch (note 3). test-mmu-retention.py now binds once per walker configuration and counts writebacks of the bound slot: clean_poc without clean_tables gives one (from bind), clean_tables gives one (from publish_slot), and a coherent walker gives none. With the new branch disabled the case fails.
  • submission_entry comment (note 4). m3_config::submit_queue has no per-profile layout, so the M3 runtime writes this entry for both 14.8.3 Legacy and 26.6.2 Extended. The comment now says that.
Head Check Result
7fcd6e20 (#2545 head, contains this one) asahi-m4 run-host-tests.py on Linux 23/23, including the new retention case
same tools/asahi/m3-runtime-tests.rs 83/83

Untested: the TTBAT clean on G15, G16 and T8122 hardware (no such machine booted this).

J615 on 27.2. Agreed: it needs a 27.2 Profile for t8122, a board gate that covers apple,j615, 27.2 identity bounds and init graph, its own flush-ring offset and control size (0x4be0/0xe850 in place of J613's 0x4f88/0xeaf0), and a Stage 2 that publishes the GPU. I don't think #2543 or #2544 needs to change shape for that:

  • write_flush_ring already takes the offset from its caller.
  • drm/asahi: G14 compute on GPU firmware 27.2 #2544 passes the G14G 27.2 offset and control size from its own root (gpu/v27_2.rs), not from M3 constants.
  • The J613 values sit in one place, m3_init_26::firmware_control and allocation(CONTROL_REGION). The PR that adds the T8122 27.2 profile can key them by Profile there, with its own values.

Adding the selector now would give one M3 value to select and no machine to test the second, so I've left it for that PR.

@RyanTheTide

Copy link
Copy Markdown
Contributor

Blockers: 1.

The code is sound and I have nothing to add to the review points already answered. The helpers write the same bytes as the inline code they replace (submit_queue, PointerGraph channels 0x20..0x1f8, flush ring at 0x4f88). The ring allocation size is still 0x1800 (256 x 0x18). The new Uat::bind clean (mmu.rs:3864-3870) runs only when clean_poc && !clean_tables, so a slot never gets a second clean, and it finishes before the handoff transaction drops. The retention case covers all three walker configurations.

  1. Neither commit (6eb9dc5, 949b48b) has a Signed-off-by: line; each carries only Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>. Please amend both: drop the Co-Authored-By trailer and add the human sign-off (git commit --amend -s per commit, or git rebase --exec 'git commit --amend --no-edit -s'). Then replace the PR body's last line, the "Generated with Claude Code" line, with Assisted-by: LLM (Claude:Opus 5.5).

HW row: J613 (M3, T8122 runtime, 26.6.2 stub) on this head: GPU binds, a Vulkan compute run with every word checked on the CPU, submitted == completed, clean rmmod, plus 0 AVC, 0 failed units and 0 new kernel errors. That covers the refactored M3 encoders and the TTBAT clean on a new_adopted_with_user_upper UAT, neither of which any machine has booted yet.

Assisted-by: LLM (Claude:Opus 5.5)

@RyanTheTide

Copy link
Copy Markdown
Contributor

Blockers: 1.
Neither commit (6eb9dc5, 949b48b) has a Signed-off-by line, and both still end with a Co-Authored-By trailer. Please amend both to drop the trailer and add the author's sign-off, then replace the PR body's last line with Assisted-by: LLM (Claude:Opus 5.5).
The code is otherwise fine. submit_queue, the 13 TX descriptors and 4 RX rings, and the 0x4f88 flush-ring pair are written byte for byte as before. The control-ring allocation is still 0x1800. The new zero/alignment checks can't fail on the J613 path. The Uat::bind slot clean runs only when clean_poc && !clean_tables, after publish_slot and inside the transaction, so no slot is cleaned twice and make_inner UATs never reach it.
Rows still missing: a G15 (M3 Pro/Max) bind with the TTBAT clean, where the mapping can be write-back, and a G16 J713/J773g boot with this change.

@satchlj
satchlj force-pushed the satchlj/g14-v27-2-shared branch from 949b48b to 259dd1a Compare November 21, 2026 21:12
@satchlj

satchlj commented Nov 21, 2026

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf: trailer-only rewrite, removed Co-Authored-By from 2 commit(s); same trees, Signed-off-by kept (949b48b -> 259dd1a, --force-with-lease).

@satchlj
satchlj force-pushed the satchlj/g14-v27-2-shared branch from 259dd1a to a111d34 Compare November 21, 2026 21:53
@satchlj

satchlj commented Nov 21, 2026

Copy link
Copy Markdown
Member Author

Claude Opus 5.5 on @satchlj's behalf: added the missing Signed-off-by: Satya Benson to 2 commit(s); same trees, nothing else changed (259dd1a -> a111d34, --force-with-lease).

satchlj added a commit to aurora-silicon/aura that referenced this pull request Nov 22, 2026
…d image

On macOS 27.2 iBoot loads T8112's G14G GPU firmware as three gfx-asc
segments (__TEXT;__DATA;__DATA_SHARED_RO), and aura has no initdata
layout for it, so #381 leaves the GPU disabled. Linux's g14_v27_2
backend (aurora-silicon/linux#2543-#2545) starts that firmware itself
from the loader's handoff instead.

Route on the loaded image, not on the release:

- firmware_gfx_version() also identifies images without an image-info
  header, by the uuid iBoot puts in /arm-io/gfx-asc/iop-gfx-nub and the
  __TEXT size. One row: the G14G 27.2 image (26B123).
- When the G13/G14 path has no initdata layout and the image is
  identified, the operating points and leakage are published as usual,
  then dt_set_gpu_g14g_handoff(): #393's T8122 handoff reader and
  region publication, parameterised by target (log tag, sgx compatible,
  GPU node compatible, first-segment VA). For G14G the GPU node is the
  apple,agx-g14g one and the segment VAs must run on from the first.
  It reserves no-map and links ttbs, pagetables, handoff, shared-l2,
  fw-text and fw-data (__DATA + the read-only tail), and publishes
  apple,firmware-version/-compat from the image,
  apple,coredump-copy-buffer from the nub and
  /chosen/aurora-gpu-resources/sgx adt-perf-states-sram (the ADT
  property, which the kernel's 27.2 backend reads), then sets the GPU
  node okay, all in one FDT transaction.
- An unidentified image, or an ADT the handoff reader refuses, leaves
  the GPU disabled as #381 does.

T8122's messages and DT output are unchanged: all 95 host scenarios of
tests/kboot_gpu_g15_host.c give byte-identical output against main
(83 pass, the same 12 stand-in/calibration fixtures fail on both).

Signed-off-by: Satya Benson <satya@benson.earth>
@RyanTheTide

Copy link
Copy Markdown
Contributor

J713 (M4 Air, G16G, macOS 27.2) boot row for a111d34. a111d34 has the same tree as 259dd1a; the push only rewrote commit messages and trailers (git diff 259dd1a2 a111d34c is empty). 259dd1a is part of the J713 default kernel (7.3.0-bj713t6+, the integration tip plus open J713 PR heads, default cmdline), so this exact code has run on the J713 since 13:00 today. Results there:

  • The G16 reaches READY at every boot. That covers the gate boot, three warm reboots, and the boots since.
  • 13 of 13 deep sleeps came back with the GPU in 0.6-0.7 ms, and dEQP-VK smoke passed after each wake. vkmark ran across a sleep, and glmark2 scored 3291.
  • There were no firmware crashes, oops or DART faults.

Still owed from your list: the G15 TTBAT bind row (J615) and a J773g boot.

Assisted-by: LLM (Claude:Opus 5.5)

@DjDeveloperr DjDeveloperr left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pre-PR checklist review at head a111d34c1c97.

Code & Correctness

The three factored wire helpers (m3_channel_format::submission_entry, m3_init_26::{write_channels,write_flush_ring}, m3_mmio_26::write_record) and ENGINE_CLASSES produce the exact same bytes as the inline M3 code they replace, and write_flush_ring's alignment/non-zero checks are weaker than what PointerGraph::addresses() already enforces.

Blocker

  1. Symmetric TTBAT teardown omits PoC clean (drivers/gpu/drm/asahi/mmu.rs:3864-3870 vs VmInner::drop at mmu.rs:3455-3459).
    bind_inner now writes back ttbs()[idx] to PoC when self.inner.clean_poc && !uat_inner.quirks.clean_tables so a non-snooping table walker sees the bound slot on WB-mapped adopted UATs (Uat::new_adopted, i.e., G15 Pro/Max and G14 27.2). However, VmInner::drop clears ttbs()[idx].ttb0/ttb1 to 0 and calls only uat_inner.publish_slot(idx), which is gated strictly on self.quirks.clean_tables. On an adopted UAT (clean_poc == true, clean_tables == false) with a WB-mapped TTBAT, the zeroed slot is never written back to PoC before VmInner::drop invalidates the ASID and frees the VM's physical page tables — leaving a stale valid ttb0 pointing at freed pages visible to the non-snooping walker.
    • Fix: move the clean_poc slot write-back into UatShared::publish_slot (or store the clean-slot flag on UatShared) so every TTBAT slot mutation — both bind and VmInner::drop teardown (as well as execution-context/kernel-alias updates) — writes back the slot once to PoC without duplicating the branch in bind_inner. Extend tools/testing/selftests/drivers/gpu/asahi-m4/test-mmu-retention.py to assert the slot clean on VmInner::drop for (clean_poc=true, clean_tables=false) as well.

Pre-PR Checklist (§1–§6)

  • §1 No model gating: Pass.
  • §2 Production-ready, no hacks: Pass.
  • §3 Safety: Blocked by the missing VmInner::drop TTBAT PoC clean above (DMA/table-walker safety on VM teardown).
  • §4 Clean-room & attribution: Pass. Both commits carry human author/committer and human Signed-off-by: only, with no AI trailers; PR body discloses agent assistance.
  • §5 Evidence: Partial. host-tests CI is green and J713 (G16G) boot/sleep is reported at a111d34c. Still missing before merge: (a) M3 (T8122, J613/J615) hardware run covering the refactored M3 encoders and new_adopted_with_user_upper, (b) G15 (M3 Pro/Max) WB TTBAT bind/unbind row, and (c) regression counts (0 SELinux denials, 0 failed units, 0 new kernel errors) on the live head.
  • §6 Naming: Pass.

Assisted-by: LLM (Gemini 4 Neon)

Factor three 26.x wire encoders out of the M3 code so another firmware
interface can use them without copying: the 0x18-byte submission entry
(m3_channel_format::submission_entry), the runtime directory's channel
pointers and the cache-flush ring pair (m3_init_26::write_channels,
write_flush_ring), and the 0x28-byte register record
(m3_mmio_26::write_record). The M3 paths call them and encode the same
bytes; host tests cover the submission entry and the flush-ring pair.

Bind also writes back the TTBAT slot it fills when the UAT's walker does
not snoop the CPU caches (an adopted UAT without the clean_tables quirk).
Until now only the tables were written back there. The asahi-m4
retention harness models UatInner::clean_poc, which bind now reads.

Signed-off-by: Satya Benson <satya@benson.earth>
…sers

The asahi-m4 retention harness modelled every UAT with clean_poc false,
so the TTBAT write-back that bind_inner does for an adopted UAT without
the clean_tables quirk compiled but never ran. Add a bind case for each
walker configuration that counts writebacks of the bound slot: once by
bind for clean_poc alone, once by publish_slot with the quirk, none for a
coherent walker. With the new branch disabled the case fails.

submission_entry's comment named only G15G 26.6.2 and G14G 26B123. The M3
runtime writes it for both of its profiles (14.8.3 and 26.6.2) through
m3_config::submit_queue, which has no per-profile layout; say so.

Signed-off-by: Satya Benson <satya@benson.earth>
@RyanTheTide

Copy link
Copy Markdown
Contributor

Re-checked at 119cce5. This is a rebase onto integration/aurora d1b022d and nothing else changed. Range-diff shows both commits unchanged, and no base commit since 458f759 touches the files this PR changes. So the J713 boot/sleep row posted for a111d34 still covers this code.

The teardown point from the 09:12 review is still open at this head. VmInner::drop zeroes the slot (drivers/gpu/drm/asahi/mmu.rs:3461-3462) and calls only publish_slot (mmu.rs:3464), which cleans only under quirks.clean_tables (mmu.rs:2084-2090). Meanwhile the bind-side clean at mmu.rs:3870-3875 covers clean_poc && !clean_tables. Doing the clean in publish_slot for both cases would fix bind and teardown in one place. A retention case that counts a clean on drop would cover it.

Rows still missing: M3 T8122 (J613/J615) with the refactored encoders, a G15 bind/unbind with a write-back TTBAT, and a J773g boot.

Assisted-by: LLM (Claude:Opus 5.5)

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