Repository navigation
Conversation
The 25G83 device tree advertises the AGX-specific compatible followed by asc-mailbox-v4. Both bind the same mailbox operations. Admit this exact pair while retaining all resource and identity checks. Exercise the production rejection predicate with public-DTS fixtures and resource mutations; allow optional local DT captures for reproduction. Signed-off-by: Satya Benson <satya@benson.earth>
chriskearney
left a comment
chriskearney
left a comment
There was a problem hiding this comment.
Reviewed d2c35b6 against 2dd4aab. The added exact compatible pair already occurs in the public 25G83 DTS; both compatibles select apple_mbox_asc_hw. Register, IRQ ordering/names, cell-count and identity checks are retained. The production-expression host test passes against DTBs compiled from the J613/J615 legacy and 25G83 sources; reverting the entry fails. No new kernel CLI opt-in is added.
Important dependency: on 2dd4aab, T8122 firmware identity still admits only 14.8.3, so a genuine 26.6.2 boot is refused before this mailbox check. This entry becomes useful with the separate Extended profile in draft #303. The #303–#307 stack does not contain this pair and needs explicit composition; this approval does not review or approve that stack. Please make that dependency clear in the PR description.
Source approval only. The #301/#302 registry conflict was composed locally preserving both tests (ManifestTest and both suites pass); full composed CI/package checks remain required before merge. No Actions run, kernel build or independent physical test is credited here. The author's diagnostic/manual-switch boot with additional kernel and bootloader changes is not proof for these PRs alone or the packaged default path. Firmware admission, J615 experimental status and existing thermal/TCMz limits remain unchanged. A reversed-pair negative case would be useful optional test hardening.
|
Thanks @chriskearney. I added the dependency on the 26.6.2 profile (#303–#307) to the description, and noted that the stack will be rebased onto #301/#302 before leaving draft. The reversed-pair negative case is a good idea; I'll fold it in when I rebase the stack, so this PR doesn't need another approval cycle just for that. Written by Claude on @satchlj's behalf. |
|
Please hold this PR from the merge queue for this RC, alongside #301. Our existing 14.8.3 path is working and tested on both M3 Air models (J613/J615), and we are retaining that validated path unless a concrete user-visible benefit or compatibility requirement justifies expanding RC scope. The source approval and green normal CI stand. This is an RC deferral, not a new source-correctness objection. I have asked for the missing benefit/requirement on #301. Rationale: #301 (comment) |
The 25G83 device tree gives the GPU mailbox
apple,t8122-agx-asc-mailboxfollowed byapple,asc-mailbox-v4, but the T8122 admission table does not list that exact pair, so the GPU is refused with "unexpected GPU mailbox description". Both compatibles bind the same mailbox operations. Add the pair without relaxing the register, interrupt, cell-count or identity checks.With this change, a diagnostic boot of J613 (T8122) on 26.6.2 firmware passed mailbox admission and, together with the related fixes, reached native Apple M3 rendering and the internal display.
The host test compiles the production rejection expression, uses fixtures from the public DTS, and checks that resource and identity mutations are still rejected.
Dependency: on its own, on
integration/aurora(2dd4aab), this changes nothing at runtime for a 26.6.2 boot: T8122 firmware identity there still admits only 14.8.3, so 26.6.2 is refused before this mailbox check. The entry becomes reachable with the extended 26.6.2 profile in the draft stack #303–#307. That stack is currently based on 2dd4aab and does not contain this pair; it will be rebased onto #301 and #302 before leaving draft.