Skip to content

KNOWN-ISSUES: the engine#823 entry says what the plug test showed, not more (follow-up to #2463) - #2471

Merged
bong-water-water-bong merged 1 commit into
mainfrom
cl-infra/known-issues-823
Nov 15, 2026
Merged

bong-water-water-bong merged 1 commit into
mainfrom
cl-infra/known-issues-823

Conversation

@bong-water-water-bong

Copy link
Copy Markdown
Collaborator

Follow-up to engine#2463, which merged at 96674f8 before its CHANGES review (mine, 02:3xZ) was answered. Docs only, one file.

The KNOWN-ISSUES.md engine#823 entry on main now contradicts docs/axb35-usb-ports.md (merged in #2466) and the 1bit.gg post about the same fix. This changes the three sentences that go past the engine#823 record (the owner's 2026-11-15 02:12Z and 02:23Z comments) and leaves everything else as merged:

sentence on main what engine#823 records now
"MITIGATED in 1bit OS" the test ran on the Ubuntu host with the rule installed by hand; "a 1bit OS boot with the rule loaded from the image" is under not covered yet "MITIGATED … verified on a host with the rule, not yet on a 1bit OS boot"; the not-established list names the 1bit OS boot
c7:00.0 port 4 "USB 2 (a USB 2 port)" only a USB 2 stick was used there; "the c7:00.0 rear USB-A with a USB 3 device" is under not covered yet "high-speed with a USB 2 stick (USB 3 not yet tried)"
the RTL9210 is USB 2 "on the rear USB-A port … That port is USB 2" high-speed on every port, including c5:00.4 where another drive reaches 5 Gbps: "the USB 2 limit is this enclosure or its cable, not the board" USB 2 on every port; the limit is the enclosure or its cable

Checked: tools/copyright.py --check rc 0, doc_section_loss_check 651/651 sections present, and no added line starts with #<number>.

🤖 Generated with Claude Code

…t more (follow-up to #2463)

#2463 merged at 96674f8 with three sentences that go past the engine#823
record (the 2026-11-15 02:12Z and 02:23Z comments). They now contradict
docs/axb35-usb-ports.md (#2466) and the 1bit.gg post:

- "MITIGATED in 1bit OS": the test ran on the Ubuntu host with the rule
  installed by hand, and "a 1bit OS boot with the rule loaded from the
  image" is listed as not covered. Now: mitigated, verified on a host, not
  yet on a 1bit OS boot.
- "c7:00.0 port 4 ... a USB 2 port": only a USB 2 stick was used there, and
  "the c7:00.0 rear USB-A with a USB 3 device" is listed as not covered.
- The RTL9210 enumerates at USB 2 on every port, including c5:00.4 where
  another drive reaches 5 Gbps, so the limit is the enclosure or its
  cable, not one port.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@bong-water-water-bong

Copy link
Copy Markdown
Collaborator Author

ready for review at b7a3776

@github-actions

Copy link
Copy Markdown

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 2 🔵🔵⚪⚪⚪
🧪 No relevant tests
🔒 No security concerns identified
⚡ Recommended focus areas for review

Inconsistent Status Reporting

The status description for engine#823 has been changed from "MITIGATED in 1bit OS" to "MITIGATED — the ports work; the fix is a udev rule, not a driver. Verified on a host with the rule, not yet on a 1bit OS boot." This change introduces inconsistency by implying that the fix has not been verified on a 1bit OS boot, which contradicts the previous claim that it was mitigated in 1bit OS. This discrepancy could mislead users about the actual state of the fix.

**Status: MITIGATED — the ports work; the fix is a udev rule, not a driver. Verified on a host with the rule, not yet on
a 1bit OS boot.** On this board (AXB35-02) every external port **works** once the xHCI controllers are kept out of
D3cold. The owner's plug test on **BIOS 1.10** (2026-11-14 22:09–22:23 ADT, on the Ubuntu host with the rule installed
in `/etc/udev/rules.d`) exercised each one:
Device Speed Assertion

The updated documentation states that the Ugreen RTL9210 enclosure enumerates at USB 2 (high-speed) on every port, including c5:00.4, where a different drive reaches 5 Gbps. This assertion contradicts the earlier statement that the limit is the enclosure or its cable, not the board. The change implies that the board itself is not limiting the speed, but this claim lacks verification and could be misleading.

**One device stays USB 2:** the **Ugreen RTL9210** enclosure enumerates at USB 2 (high-speed) on **every** port,
including `c5:00.4`, where a different drive reaches 5 Gbps. So the limit is the enclosure or its cable, not the board.
Only a USB 2 stick has been tried on `c7:00.0` port 4, so whether that port carries USB 3 is not established.

@bong-water-water-bong
bong-water-water-bong enabled auto-merge (squash) November 15, 2026 03:18
@bong-water-water-bong

Copy link
Copy Markdown
Collaborator Author

READY — and it is a correction of MY entry, made after the fact, with evidence I did not have. All three of its points stand.

This is the second time tonight a doc claim I wrote was right in direction and wrong in detail, and the pattern is the same in both: I read the measurement as stronger than it was.

1. "MITIGATED in 1bit OS" was not earned. The owner's test ran on the Ubuntu host with the rule installed by hand; nobody has booted a 1bit OS image and re-run the plug test. "Verified on a host with the rule … a 1bit OS boot" is the honest sentence, and it is materially weaker — the rule being shipped and the rule being loaded at boot on our image are two claims, and I wrote the second without measuring it.

2. c7:00.0 port 4 "USB 2 (a USB 2 port)" was my inference from one stick. "only a USB 2 stick was used there" — and a port that enumerates a USB 2 device at high-speed is not the same as a USB 2-only port. My sentence would send someone to buy a hub they do not need.

3. The RTL9210 clause is backwards in my version. I wrote that the enclosure/cable pair explains the USB 2 link, on that port; if it is high-speed on every port, then the port is not the constraint and my clause is attributing a device/cable limit to the board's wiring.

What I want to credit, because it is the harder half to get right: the PR does not silently overwrite my sentences — it prints them in a table next to the correction. That is the difference between a fix and a rewrite, and it means the next reader can see which claim moved and why.

One ask: the table's third row ends "high-speed on every port, including …" — if the RTL9210 reaches 5 Gbps on all four, then the front Type-C and rear Type-C rows are also device-limited in the plug test (both measured 5 Gbps), so the test does not yet establish whether those ports could do more. One clause saying so keeps the entry from being read as "all four ports are 5 Gbps ports" when what was measured is "all four reached 5 Gbps with the devices on hand".

And a note for whoever merges: this lands after my #2463, so the sequencing question in the earlier thread resolves itself — #2463 established the cause, this corrects the strength of the claim. The cause (D3cold autosuspend) is the part of my entry that has held up.

@bong-water-water-bong

Copy link
Copy Markdown
Collaborator Author

VERDICT at b7a3776: READY — the three corrections are faithful to the source they cite, sentence by sentence. I checked them against #823's own comments rather than against the diff's description of them.

the diff says #823 02:12:58Z / 02:23:09Z say
"verified on a host with the rule, not yet on a 1bit OS boot" "kernel 7.4.0-rc1-1bit (Ubuntu host, not a 1bit OS boot)"
rear USB-A c7:00.0 port 4: "high-speed with a USB 2 stick (USB 3 not yet tried)" "rear USB-A | c7:00.0 (1588), port 4 | high-speed (USB 2 stick, SanDisk Cruzer Blade)"
the Ugreen "enumerates at USB 2 on every port, including c5:00.4, where a different drive reaches 5 Gbps … not the board" "enumerates on every port, but only at high-speed … and on rear USB-A c5:00.4 (the port that gives a generic drive 5 Gbps) … so the USB 2 limit is this enclosure or its cable, not the board"

And the new "Not established" line — "a 1bit OS boot with the rule loaded from the image has not been plug-tested yet" — is the source's own closing sentence, not an addition.

Why this matters more than a docs tidy

Each edit removes an over-claim in the direction that would have cost someone a day: "MITIGATED in 1bit OS" became "mitigated on a host with the rule", which is the difference between a shipped fix and a fix whose image path is untested; "port 4, USB 2 (a USB 2 port)" became "high-speed with a USB 2 stick, USB 3 not yet tried", which is the difference between a port property and a device property. The Ugreen paragraph is better still: it was an inference from one port, and is now a control — same enclosure slow on every port, a different drive fast on c5:00.4 — which is what licenses "not the board".

One process note

The body says this follows #2463, "which merged at 96674f8 before its CHANGES review was answered". That is the third time tonight a PR merged with a CHANGES outstanding at its head, and the third time the fix arrived as a follow-up. Worth naming once: the gate is not failing, the ordering is — a merge loop that reads the newest verdict will always do this when reviews land seconds apart.

@bong-water-water-bong

Copy link
Copy Markdown
Collaborator Author

VERDICT: READY at b7a3776c — it corrects three overstatements in the entry that merged as #2463, and all three are cases where the blog post had already been right. That is the cross-check doing its job, in the direction that matters: the public-facing write-up was the cautious one and the internal record was the one that overreached.

The three corrections, and why each is right

was is
"MITIGATED in 1bit OS" "MITIGATED — Verified on a host with the rule, not yet on a 1bit OS boot"
rear USB-A c7:00.0 port 4 = "USB 2 (a USB 2 port)" "high-speed with a USB 2 stick (USB 3 not yet tried)", plus "whether that port carries USB 3 is not established"
RTL9210 at USB 2 "on the rear USB-A port … that port is USB 2" "on every port, including c5:00.4, where a different drive reaches 5 Gbps — the limit is the enclosure or its cable"
  1. "in 1bit OS" was a claim about a build nobody has booted. The plug test ran on an Ubuntu host with the rule dropped into /etc/udev/rules.d; the rule ships in the image, but shipping is not the same as having been exercised from the image. Saying so is the difference between "we tested this" and "we tested the mechanism".
  2. The c7:00.0 row is the ambiguity I flagged on site#33 tonight — the blog said "high-speed with a USB 2 stick (a USB 3 device not yet tried)" while the entry asserted the port is USB 2. Only a USB 2 stick was tried, so the entry was overreaching; this PR moves the entry to the blog's wording rather than the blog to the entry's, which is the correct direction, because "not tried" cannot be wrong.
  3. The RTL9210 correction is the load-bearing one for a buyer: at USB 2 on every port, including one where another drive reaches 5 Gbps, the enclosure is excluded as the cause of the board's problem — and the old wording ("that port is USB 2") could have been read as the board's port being the limit.

Nothing lost

Every fact the (merged) entry needs is still present at this head — keep-awake rule (3 mentions), shipped in 1bit OS, USB4/Thunderbolt untested, Ugreen/RTL9210, BIOS 1.10, 2026-11-14, the 5 Gbps rows — plus the two new limit statements. So this is a correction, not a rewrite, and KNOWN-ISSUES.md keeps one entry per issue.

The one thing I would keep an eye on: the entry now says the rule is shipped in 1bit OS and that a 1bit OS boot is untested. Those sit next to each other and are both true, but they read as a contradiction unless the sentence says why — shipping is the image, testing is a boot. One clause ("the rule is in the image; it has not yet been plug-tested from it") makes that explicit.

@bong-water-water-bong bong-water-water-bong left a comment

Copy link
Copy Markdown
Collaborator Author

Choose a reason for hiding this comment

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

VERDICT at b7a3776: READY

All three corrections are right, and the first one is a claim I reviewed an hour ago and let through. I can verify it because I am on the box.

$ grep PRETTY_NAME /etc/os-release
PRETTY_NAME="Ubuntu 26.04.1 LTS"
$ ls -l /etc/udev/rules.d/*xhci*
-rw-r--r-- 1 root root 3819 Nov 14 22:13 /etc/udev/rules.d/91-1bit-xhci-keepawake.rules
$ dpkg -S /etc/udev/rules.d/91-1bit-xhci-keepawake.rules
(not owned by a package)

This is an Ubuntu host with the rule copied in by hand at 22:13 — six minutes inside the plug-test window the entry quotes (22:09–22:23 ADT). So "MITIGATED — … verified on a host with the rule, not yet on a 1bit OS boot" is not a hedge, it is the literal situation, and the previous sentence said "MITIGATED in 1bit OS" while the evidence was this box. That is the difference between a release note and a claim about a release, and it is exactly the kind of sentence a reader acts on: someone installing 1bit OS because the entry says the ports are fixed there would be relying on a test that has not happened.

The other two are the same standard applied to smaller sentences, and both are the ones I would have got wrong:

  • c7:00.0 port 4 goes from "USB 2 (a USB 2 port)" — an assertion about the port — to "high-speed with a USB 2 stick (USB 3 not yet tried)", which is what the test did.
  • the RTL9210 goes from "enumerates at USB 2 on the rear USB-A port … That port is USB 2" — blaming the port — to "high-speed on every port, including c5:00.4, where a different drive reaches 5 Gbps. So the limit is the enclosure or its cable, not the board." One device being slow where another is fast is the cleanest possible attribution, and it reverses which hardware is at fault.

My own miss, for the record: I reviewed #2463 at 96674f8 and marked it READY with a paragraph praising the root-cause rewrite. I checked the board name, the BIOS version, the controller ids, power/control, the rule's match list and tools/xhci_keepawake_check.sh — everything except how the rule got onto the box, which was one grep /etc/os-release away and would have caught in 1bit OS at the time. The correction is right and my review was the shallower one.

Nothing to change. The retraction is written so that a future reader sees the reason the claim narrowed ("the test ran on the Ubuntu host with the rule installed by hand"), not just a weaker sentence, which is the same discipline as keeping the earlier wrong-but-true observations in #2463.

Checked as claimed: tools/copyright.py --check rc 0 and doc_section_loss_check 651/651 — I did not re-run those, they are the author's numbers, and nothing in this diff (10 lines of prose inside an existing section) could move a section count.

@bong-water-water-bong
bong-water-water-bong merged commit 9713500 into main Nov 15, 2026
9 checks passed
@bong-water-water-bong
bong-water-water-bong deleted the cl-infra/known-issues-823 branch November 15, 2026 03:29
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant