upstream: decide Bigscreen Beyond DisplayID parser vs EDID quirk route

[Jesssullivan]

Goal

Select the correct upstream route for Bigscreen Beyond EDID non-desktop handling.

The original goal was a small BIG/0x1234 DRM EDID quirk. That route is now on hold because Sefa Eyeoglu submitted the same quirk to dri-devel in May 2024, and the review thread redirected the problem toward DisplayID primary-use parsing.

Decision gate: Monday evening, 2026-04-27.

2026-04-25 update: Path A now has a positive no-quirk KUnit result on current
drm-misc-next. The likely upstream artifact is no longer a BIG/0x1234
quirk; it is either a regression-test-only patch or no upstream patch because
current upstream already handles this EDID.

Current evidence

• honey exposes the headset at /sys/class/drm/card0-DP-2 on kernel 6.19.5-7.xr.el10.
• EDID first bytes: 00 ff ff ff ff ff ff 00 09 27 34 12 d2 04 00 00.
• That decodes to manufacturer BIG and product 0x1234.
• The DisplayID extension begins 70 20 79 07, meaning DisplayID 2.0 with primary use 0x07, head-mounted VR.
• Durable full EDID capture: Jesssullivan/Dell-7810:data/captures/honey/bigscreen-beyond-edid-2026-04-25.txt.
• No-quirk KUnit proof: Jesssullivan/Dell-7810:data/captures/honey/bigscreen-displayid-kunit-2026-04-25.txt.
• KUnit tree: drm-misc-next commit 03af6c3afc48.
• KUnit result: drm_test_connector_edid_displayid_hmd_primary_use passed.
• This proves hardware identity, DisplayID primary-use evidence, and current upstream parser behavior for this EDID. It does not yet prove a live connector-property result from an unmodified upstream kernel booted on this hardware.

Prior public review

• Sefa initial patch: https://lists.freedesktop.org/archives/dri-devel/2024-May/454216.html
• Jani evidence request: https://lists.freedesktop.org/archives/dri-devel/2024-May/454226.html
• Sefa revised patch: https://lists.freedesktop.org/archives/dri-devel/2024-May/454244.html
• Philipp DisplayID parser objection: https://lists.freedesktop.org/archives/dri-devel/2024-May/454248.html
• Sefa EDID conformance note: https://lists.freedesktop.org/archives/dri-devel/2024-May/454255.html
• Jani DisplayID-header position: https://lists.freedesktop.org/archives/dri-devel/2024-May/454416.html
• freedesktop issue 39: https://gitlab.freedesktop.org/drm/misc/kernel/-/work_items/39

Acceptance criteria

• Prove current drm-misc-next behavior against the captured full EDID.
• Retire or narrow the local quirk lane because current upstream sets non_desktop.
• Decide whether a regression-test-only patch is worth submitting.
• Contact Sefa before any public patch that supersedes or reopens the 2024 work.
• Do not include 0x5095 without separate evidence.
• Send any final patch with a human Signed-off-by and with the 2024 thread represented in the rationale.

Related: #9, #23

[Jesssullivan]

Refreshed upstream check on 2026-04-25 after fetching kernel.org upstream/master to 897d54018cc9:

• drivers/gpu/drm/drm_edid.c still has no BIG, 0x1234, or 0x5095 match in the checked upstream tree.
• This keeps the Bigscreen Beyond non-desktop quirk classified as a small upstream candidate rather than an already-upstreamed patch.
• Acceptance should still require local non_desktop=1 evidence on the target headset path before formatting/submitting the patch.

[Jesssullivan]

Linear tracking is now split out for this upstream submission: TIN-612.

Current grounded state:

• xr/patches/bigscreen-beyond-edid.patch applies cleanly against the case-sensitive linux-xr checkout.
• The current checked upstream drivers/gpu/drm/drm_edid.c still has no BIG/0x1234 or BIG/0x5095 quirk.
• Bounded maintainer lookup for drivers/gpu/drm/drm_edid.c points at DRM misc + DRM maintainers and dri-devel@lists.freedesktop.org / linux-kernel@vger.kernel.org.

Before sending v1 upstream, we still need fresh host evidence:

1. capture the connected headset DRM path on honey
2. capture EDID identity showing BIG and product 0x5095 or 0x1234
3. capture non_desktop=1 with the carry applied
4. format the patch against current upstream or drm-misc-next
5. run scripts/checkpatch.pl and regenerate recipients with scripts/get_maintainer.pl --no-git
6. send with a human Signed-off-by and link the lore/thread URL here

This issue remains the public repo tracker; TIN-612 is the internal execution ticket.

[Jesssullivan]

Read-only host evidence update from 2026-04-25:

• honey is reachable on 6.19.5-7.xr.el10.
• The headset path is currently /sys/class/drm/card0-DP-2, status=connected, with a 256-byte EDID.
• First 16 EDID bytes: 00 ff ff ff ff ff ff 00 09 27 34 12 d2 04 00 00, supporting BIG/0x1234 on the live headset path.
• Still missing before v1: a clean DRM connector-property capture proving non_desktop=1. This host does not expose /sys/class/drm/card0-DP-2/non_desktop; drm_info, modetest, and edid-decode are not installed; unprivileged debugfs state is not readable.

No service restart, reboot, or rke2 interaction was performed. Linear TIN-612 has the same evidence note.

[Jesssullivan]

Patch hygiene correction from local checks:

• GNU patch -p1 --dry-run < xr/patches/bigscreen-beyond-edid.patch succeeds against current xr/main.
• git apply --check xr/patches/bigscreen-beyond-edid.patch fails at drivers/gpu/drm/drm_edid.c:220. Treat the current carry as fuzz-applicable, not submission-ready.
• scripts/checkpatch.pl --no-tree xr/patches/bigscreen-beyond-edid.patch reports one warning: prefer a lore.kernel.org link over the current mail-archive.com reference.

So the v1 prep has two gates now: capture non_desktop=1 from DRM properties, then regenerate a clean upstream-formatted patch against current upstream or drm-misc-next.

[Jesssullivan]

Route documentation PR: #31

This adds xr/patches/bigscreen-beyond-edid.route.md and links it from the patch README, upstream-status docs, and top-level README.

The route records the current state explicitly: live honey EDID identity supports BIG/0x1234, non-desktop=1 evidence is still missing, the release carry is GNU-patch usable but must be regenerated before v1, checkpatch needs a lore link, and the submission should use the bounded DRM maintainer route with a human Signed-off-by.

No restart, reboot, or rke2 interaction is part of the evidence route.

[Jesssullivan]

Route documentation landed in PR #31: #31

Merged commit: 3beda6220731b8ac23e98acc5f88cf2f8fc2dad4.

This does not complete the upstream submission. Remaining gates are still: capture non-desktop=1 from the live headset DRM connector path, regenerate a clean upstream/drm-misc patch, resolve the checkpatch lore-link warning, and send with a human Signed-off-by.

[Jesssullivan]

Evidence gate update from 2026-04-25:

• Durable PR: TIN-612: record Bigscreen EDID evidence gate #32
• honey libdrm capture now proves the connected headset path reports non-desktop=1.
• Host/kernel: honey, 6.19.5-7.xr.el10
• DRM device/connector: /dev/dri/card0, DP-2, connected
• EDID first16: 00 ff ff ff ff ff ff 00 09 27 34 12 d2 04 00 00, supporting BIG/0x1234
• DRM property: property=non-desktop value=1
• Durable host evidence path: Jesssullivan/Dell-7810:data/captures/honey/drm-connector-properties-2026-04-25.txt

Patch hygiene is also corrected in PR #32:

• git apply --check xr/patches/bigscreen-beyond-edid.patch now passes.
• patch -p1 --dry-run < xr/patches/bigscreen-beyond-edid.patch passes.
• scripts/checkpatch.pl --no-tree xr/patches/bigscreen-beyond-edid.patch reports 0 errors and 0 warnings after replacing the stale mail-archive reference with a lore link.

Remaining upstream gate: regenerate/send the v1 patch from an upstream or drm-misc topic branch with a human Signed-off-by, then link the lore/thread URL here. No restart, reboot, or rke2 interaction was performed.

[Jesssullivan]

PR #32 landed: #32

Merged commit: 0c5ef24c7dcd447d0e76f421d308d05eb5d6cc6c.

Current state after merge:

• The honey BIG/0x1234 evidence gate is recorded with DP-2 reporting non-desktop=1.
• The durable Dell evidence path is recorded: Jesssullivan/Dell-7810:data/captures/honey/drm-connector-properties-2026-04-25.txt.
• The carry patch is now git apply --check clean and checkpatch.pl clean.
• The route doc explicitly bounds the evidence: local hardware proof covers 0x1234; 0x5095 remains deferred/public-report-only unless a Beyond 2e capture is obtained or v1 narrows to 0x1234.
• The route doc explicitly notes that an unpatched honey baseline was not captured; the expected upstream argument is the EDID quirk-table mechanism.

Remaining before this issue is complete: generate the actual upstream/drm-misc topic patch, choose 0x1234 only vs both IDs with caveat, add human Signed-off-by, send v1, and link the lore/thread URL here.