Skip to content

drm/asahi: g15: admit TEXT 125f0497 on J516C and J514C (carry of #340) - #487

Merged
scottjones merged 2 commits into
integration/aurorafrom
carry/340-j514c-firmware
Nov 19, 2026
Merged

scottjones merged 2 commits into
integration/aurorafrom
carry/340-j514c-firmware

Conversation

@scottjones

Copy link
Copy Markdown
Collaborator

Carries #340 by Naeem Malik, who is away. Pushing the requested changes to the PR's own branch was refused, so the same work comes from an omacom branch here. Naeem stays the author of both commits.

Summary

 g15 firmware TEXT allow-list (drivers/gpu/drm/asahi/g15)
   J516C
+    125f0497   (third J516C entry, as in #340)
   J514C
+    125f0497   (second commit: admits the same TEXT on J514C)

The changes since #340's approved head b33057c8 are the ones asked for in its review: the J514C title, and the message and row comment narrowed to the one verified J514C case. Both commits now carry Scott's Signed-off-by after Naeem's.

Evidence

Merge Danger

Door: two-way

Blast Radius: small

Only J516C and J514C with this firmware TEXT are affected; every other board's allow-list is unchanged. Supersedes #340.

malik-na and others added 2 commits November 16, 2026 00:38
Two J516C MacBook Pro 16" M3 Max owners (issue #289) load a GPU
firmware whose TEXT digest is 125f0497c7904fde... (sha256
125f0497c7904fde87b05a9264a88cf03bdb3572a1715cec91bd0f9b78245544,
TEXT 0x64000, DATA 0x140000, marker 0x5cf28). The driver refuses it and
the Mac falls back to llvmpipe.

Add it as a third FirmwareBuild, written like the two existing J516C
entries and still gated to the J516C. The entry name and the comment
record where the digest was measured; they do not claim a distinct
firmware build.

Evidence from #289: with the digest accepted through
asahi.g15_accept_text_prefix, glxinfo reports Apple M3 Max (G15C B2),
and a GLES compute run shows 0 mismatches and no faults in dmesg.

This branch boots on a J516C M3 Max in the Mac lab (job
j1031-185713-2fe6, 3 of 3 tests pass, GPU probed). That Mac loads the
macOS 14.7 stub firmware, so the new digest itself was not exercised
there.

Assisted-by: Claude:claude-opus-5-5
Signed-off-by: Naeem Malik <malik2015naeem@gmail.com>
Signed-off-by: Scott Jones <scottajones@gmail.com>
A 14-inch M3 Max J514C loads the same GPU firmware TEXT that the
previous commit admits on J516C: sha256
125f0497c7904fde87b05a9264a88cf03bdb3572a1715cec91bd0f9b78245544,
TEXT 0x64000, DATA 0x140000, marker 0x5cf28. Its boot chain reports OS
firmware 14.7, iBoot2 iBoot-10151.140.19.700.2 and system firmware
26.6.2 (omacom/omarchy-mac-pkgs#129). Without the digest the driver
refuses the firmware and the Mac falls back to llvmpipe. With it
accepted through asahi.g15_accept_text_prefix, the GPU came up in all 3
boots tried; in the first, glxinfo reports G15C B2 with OpenGL 4.6
accelerated and the kernel log shows no asahi-g15 errors. Idle only;
Vulkan and load were not run.

A row takes one machine, so add a second row for the same digest bound
to apple,j514c. The J516C row stays bound to apple,j516c; the digest
is not open to any other T6031 board. The board images come from the
Config, not from the row, so the row carries only the firmware identity.

The J514C host test counted exactly one J514C row and expected another
board to be refused a digest; with one digest bound to two boards, count
two rows and expect the first row bound to the asking board instead.

Not tested on hardware by this change: the new digest itself was not
exercised in the Mac lab, whose J514C loads the already admitted build.
Vulkan, sustained load and power and thermal qualification of J514C
remain open.

Assisted-by: Claude:claude-sonnet-5-5
Assisted-by: Claude:claude-opus-5-5
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Naeem Malik <malik2015naeem@gmail.com>
[scott: reworded message and row comment to the one verified J514C case]
Signed-off-by: Scott Jones <scottajones@gmail.com>
@LubomirGeorgiev

Copy link
Copy Markdown

Note

Drafted with Claude Opus 5. I read every line and reproduced every number myself before I posted it.

No objection to this merge. It unblocks a J514C today, and a boot override is not something a user should have to type.

One measurement for the record, because a fourth J516C will arrive.

I pulled armfw_g15c.im4p out of Apple's macOS 14.8.3 OS firmware asset with HTTP range requests. This is the generic file from Apple's CDN, not one machine's copy. That asset carries iBoot-10151.140.19, the iBoot the first T6031 G15C row names.

armfw_g15c.im4p   3489705 bytes
file sha256       f404aac59e5bd6539e0727745e374bdfcd08032c204ea940f09d74c8dc96f6c6

rkosftab entries: 2
  a000  LC_UUID 0f22e68b-3667-3c2e-a823-3315fadd3a6a
        TEXT sha256 53342914acdba6e6b178737cf21831639d82a8638c7e4f9266b1bd0bc49b6579
  b000  LC_UUID f9bea4a3-d65b-386d-b884-ecf2df7b159e
        TEXT sha256 ebbec4aff2032e131efb7efb9f9727250e864f53db6e95f5f21538973cd073c5

both entries: text_size 0x64000   data_size 0x140000   marker 0x5cf28

The segment layout matches all three pinned T6031 G15C rows exactly. The digests match none of them. git grep finds neither 53342914… nor ebbec4af… anywhere under drivers/gpu/drm/asahi/.

So c26ec89c…, 1bf2fd34…, 669913fb… and the 125f0497… this PR adds are not digests of any firmware Apple ships. They are digests of one machine's TEXT after the boot chain wrote into it. The same holds on T6030, where #247 has the measurements, and #289 now reports a third J516C with the same symptom.

The allow-list therefore grows once per machine, not once per build. Every new owner needs a report, a review and a kernel release.

What could replace it

The build's own 16-byte LC_UUID value also sits inside the hashed TEXT, at offset 0x4214 in all seven 14.x stub images I measured: two in armfw_g15s, two in armfw_g15c, three in armfw_g15g. Matching that value, plus text_size, data_size and the marker offset, identifies the build from bytes the driver already copies. It covers none of the regions the loader rewrites.

One measurement decides whether this works: does the boot chain leave TEXT+0x4214 alone? A g15_accept_text_prefix boot that also logs the UUID read back from loaded memory answers it in one line of dmesg.

I cannot run that measurement. This Mac has no Linux yet, because the T6030 rows are in no release.

The parser that produced the table above is 90 lines of Python and runs on any armfw_g15*.im4p from macOS. Say the word and I will attach it. Full T6030 data is in #247.

@scottjones
scottjones added this pull request to the merge queue Nov 18, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Nov 18, 2026
@scottjones
scottjones added this pull request to the merge queue Nov 18, 2026
@github-merge-queue
github-merge-queue Bot removed this pull request from the merge queue due to failed status checks Nov 18, 2026
@scottjones
scottjones added this pull request to the merge queue Nov 19, 2026
Merged via the queue into integration/aurora with commit 0428c9a Nov 19, 2026
14 checks passed
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.

4 participants