Skip to content

tools/phone-smoke: zuma's two exynos-clkout lines are known - #584

Closed
aninmucek wants to merge 1 commit into
derekross:mainfrom
aninmucek:phone-smoke-clkout-known
Closed

aninmucek wants to merge 1 commit into
derekross:mainfrom
aninmucek:phone-smoke-clkout-known

Conversation

@aninmucek

Copy link
Copy Markdown
Contributor

From #509's review: the two exynos-clkout lines go into tools/phone-smoke's known list, in their own PR, with the SoC they come from.

Which SoC. On today's omamobile, clk-exynos-clkout has PMU entries for gs101 and gs201 (5d0c3dcd1, 2026-10-12) and for zumapro (93a73f872, 2026-10-31), but none for google,zuma-pmu. So on zuma it can't match its parent and its probe fails with -22: the Pixel 8 Pro (husky's NOTES) and the Pixel 8a (akita's BRINGUP-M1 and BRINGUP-M2). raven's MODEM-PLAN also mentions the -22 on gs101 (2026-10-27), most likely from a kernel without 5d0c3dcd1. With the gs101 entry, a current gs101 kernel shouldn't print it, so the comment names zuma only.

The change:

  • Two entries in known_errors. The comment names the SoC and the phones, the missing match, and the fix that would remove them: a zuma entry, once a zuma board needs the clock.
  • Two cases in tests/phone-smoke. The lines as a Pixel 8a prints them are known, and another errno from the same probe stays new, so the entry can't hide a different failure.

On the phone (the Pixel 8a's M2.6 root), the kernel health row now reads 4 error lines in dmesg (4 known, 0 new), and the "new dmesg errors" row is gone. Every other row is as in #509's report.

Checks:

  • tests/phone-smoke: 10 ok.
  • tests/all: 520 ok, 0 not ok.
  • tools/pr-check: passes.

🤖 Generated with Claude Code

On zuma, clk-exynos-clkout has no entry for google,zuma-pmu (it has the
gs101, gs201 and zumapro PMUs), so it can't match its parent, and its
probe fails with -22. Every zuma boot logs the two lines, and the smoke
report from the Pixel 8a listed them as new dmesg errors (derekross#509). List
them as known, with the SoC and the fix that would remove them. The test
covers both lines, and checks that another errno from the same probe
stays new.

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

Copy link
Copy Markdown
Owner

Thank you, Darek. You did exactly what I asked on #509, and the description is careful about which SoC prints these lines. Looking at it properly, though, what I asked for was too weak, and I can't merge it this way. That's on me, not you.

The known list in tools/phone-smoke is one list for every phone, so these two lines would be "known" on the Pixel 7 Pro and the Pixel 9a as well, and there they are a real failure. The 7 Pro's NFC controller takes its reference clock from this driver (nfc@8 in gs201-cheetah.dts, clocks = <&pmu_system_controller 1>, enabled on the phone by the nfc-test overlay), and so does the 9a's (zumapro-tegu.dts). If the clkout driver doesn't bind, NFC's clock never appears, and a driver waiting for a clock defers quietly, so these two lines are most likely the only sign. It has happened: my 7 Pro's logs from before the gs201 entry went in (2026-10-12) carry exactly these lines. I fed them to both versions of the tool: main reports 0 known, 2 new and a "new dmesg errors" row, your branch reports 2 known, 0 new. Your "another errno stays new" test is a good idea, but the case that matters is the same errno on another SoC.

There's also a second thing: linux-omarchy-mobile#209 (the 8 Pro's NFC clock pad) adds the google,zuma-pmu entry to the driver, and its author booted it on the 9th. Once it merges, zuma stops printing the lines, on the 8a too, and from then on they'd be a regression there as well.

So I'd like to hold this one until #209 lands and then close it, with the 8a's smoke report showing the two lines as new until that kernel is on the phone; that's honest and short-lived. If you'd rather have a clean report before then, the entries need to apply on zuma only: read the exact string google,zuma from /proc/device-tree/compatible on the phone, give --errors-from a way to set it so the test doesn't read the laptop's device tree, add a case where the same two lines on a gs201 board stay new, and name linux-omarchy-mobile#209 in the comment as the fix. Your call; either is fine with me.

@derekross derekross closed this Nov 10, 2026
@aninmucek
aninmucek deleted the phone-smoke-clkout-known branch November 10, 2026 09:51
@aninmucek

Copy link
Copy Markdown
Contributor Author

Thank you, that makes sense, and thanks for testing it against the 7 Pro's old logs. I'd only been looking at akita and husky, so I missed that the known list applies to every phone, and on the 7 Pro and the 9a these two lines mean NFC's clock is missing. Closing it is the right call.

Since linux-omarchy-mobile#209 has merged, I'll leave phone-smoke as it is. The 8a's M2.7 report in #607 comes from main's tool, so it shows the two lines as new. That kernel was built just before #209 went in. Once the next akita kernel is on the phone, I'll check that the lines are gone and note it in akita's results.

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