Skip to content

device/dipper: record corrected FTS5 RAM test - #704

Merged
derekross merged 10 commits into
derekross:mainfrom
quatrixone:codex/dipper-fts5-safe-ram
Nov 16, 2026
Merged

derekross merged 10 commits into
derekross:mainfrom
quatrixone:codex/dipper-fts5-safe-ram

Conversation

@quatrixone

@quatrixone quatrixone commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

What changed

Record the corrected FTS5 RAM boot, stale-DTB test, and direct-DTS boot. Add a sourceable enable overlay and a pre-boot/reserved-memory stop gate.

Hardware verification

  • Corrected overlay image 1bf0ffd5302fff7fa75f49382567063d7e03bd52083baa81a88107a47dcbdbd2: physical contact, changing X/Y and ABS_MT_TRACKING_ID=-1 release were captured on event2; UFS partitions were ro=1; /dev/sda was absent; the reserved range was present; qcom_wdt bound.
  • Direct-DTS image ac93a5d9f6787153cf620704f192c22da91cf2041789f16f4214111b8458d226: stmfts registered on event2, all 85 UFS partition flags were ro=1, /dev/sda was absent, the reserved range was present, and no Oops/BUG/panic/WARN was seen. No physical touch was captured on this boot.
  • Stale-DTB image e37196952fca14d5a9152871927039b2ac1d311ec500710f8611f9ac6eff8072 omitted both removed-region nodes and exposed the ranges as System RAM. This was detected by reviewing /proc/iomem after the touch test. Its exact base-panel DTB source commit was not retained.
  • Missing disk flags, watchdog feeder state and touch X/Y min/max are marked not read in NOTES rather than inferred. No UFS partition was written.

Safety gate

Before any future boot, decompile the packed DTB and confirm both removed-region nodes. Before touching the screen or doing any other activity, verify /proc/iomem reserves 0x85fc0000–0x8a8fffff; otherwise stop and run reboot-to bootloader immediately.

This remains a draft pending review and rebuild from the proposed clean sources.

@derekross

Copy link
Copy Markdown
Owner

Thanks for writing this down the way you did. Marking the first image as superseded and saying exactly why (the stale base-panel.dtb, the two removed-region nodes missing, both ranges showing as System RAM) is the kind of record that makes the rest believable, and so is "no physical touch event was captured" on the last boot.

Two things before this leaves draft:

  1. Where the device tree came from. The earlier rows in the table name a pull request and a commit; the three new ones say "FTS5 overlay" and "enabled directly in the Dipper kernel DTS". The published Dipper DTS has no touch node, there's no kernel pull request with one, and the overlay isn't in this repository, so nobody can rebuild what ran. You've already named the cure in the notes (the node goes into the kernel DTS disabled, and the overlay here only enables it). Could you open that as a draft pull request on linux-omarchy-mobile, put the overlay under device/dipper/overlays/, and then name the commits in the three rows (the base kernel commit too)? I'd like that in place before the phone boots another image built from a local tree.

  2. A stop for what the first boot ran with. Gate 4 in BRINGUP-M1.md asks to record the reserved-memory lines, but nothing there says stop when they're wrong, and the first image went on to a touch test with two firmware ranges handed to Linux as RAM. Please add to gate 4: stop and reboot-to bootloader at once if /proc/iomem doesn't show 0x85fc0000–0x8a8fffff reserved, checked before anything else. And one check before fastboot boot, on the laptop: decompile the DTB that was packed and look for both removed-region nodes. One sentence in the notes on how the stale DTB got into the image, and whether you saw the missing reservation during that boot or afterwards, would complete it.

Smaller things:

  • the new rows don't give sdb–sdf and the sda disk flag as the older rows do, and the last one doesn't say whether qcom_wdt and its feeder ran; add them, or "not read";
  • the lowest and highest X and Y you saw would be worth a line (more useful later than "inside the panel");
  • two older sentences in NOTES still say the FTS5 driver hasn't been tested with Dipper's node; a "(tested 2026-11-12, see above)" on each keeps a reader from stopping there;
  • this conflicts with device/dipper: add device runtime package #714 in the NOTES table (both add a row at the top). I'd take this one first and rebase device/dipper: add device runtime package #714 on it; the status line's "newest RAM boot" then moves to device/dipper: add device runtime package #714's.

@quatrixone
quatrixone force-pushed the codex/dipper-fts5-safe-ram branch from ebcb080 to 1d3138f Compare November 13, 2026 16:42
quatrixone added a commit to quatrixone/omarchy-mobile that referenced this pull request Nov 13, 2026
Preserve the kernel PR review rows, stale-DTB explanation, and pre-boot/reserved-memory stop gate when the package PR is based on derekross#704.

Co-Authored-By: Codex <codex@users.noreply.github.com>
quatrixone added a commit to quatrixone/omarchy-mobile that referenced this pull request Nov 13, 2026
Preserve the kernel PR review rows, stale-DTB explanation, and pre-boot/reserved-memory stop gate when the package PR is based on derekross#704.

Co-Authored-By: Codex <codex@users.noreply.github.com>
quatrixone added a commit to quatrixone/omarchy-mobile that referenced this pull request Nov 14, 2026
Preserve the kernel PR review rows, stale-DTB explanation, and pre-boot/reserved-memory stop gate when the package PR is based on derekross#704.

Co-Authored-By: Codex <codex@users.noreply.github.com>
quatrixone and others added 8 commits November 15, 2026 16:38
The first FTS5 test image used a stale base DTB and exposed reserved memory as System RAM. Record the corrected RAM-only boot, read-only UFS checks and physical contact event, including the touch-data gaps.
Rebasing PR derekross#704 changed the overlay source commit. Update the
hardware record so its test configuration points at the source in
the current PR history.

Assisted-by: Codex
@quatrixone
quatrixone force-pushed the codex/dipper-fts5-safe-ram branch from d7aa2b3 to eec33c9 Compare November 15, 2026 15:40
@quatrixone
quatrixone marked this pull request as ready for review November 15, 2026 15:41
@derekross

Copy link
Copy Markdown
Owner

Thank you for this, it answers both of my points. The node is now proposed disabled in linux-omarchy-mobile#249, the overlay here only switches it on, the rows name the base commit, and gate 4 checks the packed DTB for both removed-region nodes before the boot and /proc/iomem before anything else on the phone. The sentence on how the stale DTB got in, and that the missing range was only noticed after the touch test, is exactly what I was hoping for.

One thing before I merge, and it's about wording:

  1. Which overlay the boots ran. The two overlay rows now say the "enable overlay source" is this PR's commit 5184f623 (and f25195d1 for the superseded boot). Both of those hold the seven-line file that only sets status = "okay", and both were written after the boots. The overlay the boots actually used defined the whole touch node; your earlier version said so ("The test overlay still defines the touchscreen node itself"), and that sentence went in this push. Could you say in both rows that the booted overlay defined the node itself, and that device/husky: battery bring-up overlays, the dead-man tool and the results #249's node plus this overlay rebuilds it, written afterwards? If you still have the 1bf0ffd5 image on your laptop, one more offline check would let the notes say more than that: extract its DTB, decompile the touchscreen@49 node, and diff it against a DTB built from device/husky: battery bring-up overlays, the dead-man tool and the results #249 with this overlay applied. Then the record says "identical", or names what differs.

A small one: after the squash, 5184f623 is only reachable through this pull request. The overlay itself lands on main with this PR, so naming the path (and its blob, f09381d5) would keep the row valid through any rebase.

After that I'll merge this, and #714 can rebase on it.

@derekross

Copy link
Copy Markdown
Owner

Thank you, this is what I was after. Both rows now say the booted overlay defined the node itself and that its source wasn't kept, and they name the rebuild through #249 plus the overlay's path and blob. That stays valid after the squash. I'm merging it.

When you rebase #714 on it, one optional wording point. "rebuilds the tested node" sounds as if you compared the two. If the 1bf0ffd5 image is still around, the DTB diff I mentioned would let it say "identical". Otherwise "is meant to rebuild" says what the record supports.

@derekross
derekross merged commit 6fa2458 into derekross:main Nov 16, 2026
2 checks passed
quatrixone added a commit to quatrixone/omarchy-mobile that referenced this pull request Nov 16, 2026
Preserve the kernel PR review rows, stale-DTB explanation, and pre-boot/reserved-memory stop gate when the package PR is based on derekross#704.
quatrixone added a commit to quatrixone/omarchy-mobile that referenced this pull request Nov 16, 2026
Preserve the kernel PR review rows, stale-DTB explanation, and pre-boot/reserved-memory stop gate when the package PR is based on derekross#704.
quatrixone added a commit to quatrixone/omarchy-mobile that referenced this pull request Nov 18, 2026
Preserve the kernel PR review rows, stale-DTB explanation, and pre-boot/reserved-memory stop gate when the package PR is based on derekross#704.
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.

2 participants