Skip to content

DEVICES: the Pixel 9a's Suspend cell to works (opt-in), with the runs on the published branch - #846

Merged
derekross merged 1 commit into
derekross:mainfrom
W4HO-ham:devices-tegu-suspend-works
Nov 18, 2026
Merged

derekross merged 1 commit into
derekross:mainfrom
W4HO-ham:devices-tegu-suspend-works

Conversation

@W4HO-ham

@W4HO-ham W4HO-ham commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

The 9a's Suspend cell said "works once that no-wake is found and fixed and a series has run from the published kernel as it is, with kernel #224 in it". Both are done now:

After the review, this pull request:

The cell keeps "kernel #247 comes with the next release" and the 300 mA drain caveat.

@derekross

Copy link
Copy Markdown
Owner

Thank you for this, and for asking the release question up front. You've done both halves of what the cell asked for: #247 is merged, and the sleeps on the published branch as it is all woke. Before the word changes I'd like four things in the cell and its notes, and I owe you two decisions.

  1. The word. The legend makes "works" mean on by default, and the cell's first clause says the 9a's auto-sleep toggle ships off. By the legend that is "works (opt-in)", with the reason the toggle is off, the same form the Modem cell uses. If we'd rather have plain "works", the way there is to ship auto-sleep on in device-tegu, which is a separate change.
  2. The notes. The cell links device/tegu MODEM-PLAN: the 9a sleeps with the modem up on a test kernel with the sleep guard (2026-11-13) #758 for the 86 sleeps and the two hangs, but those runs are in device/tegu MODEM-PLAN: the 9a sleeps with the modem up on a test kernel with the sleep guard (2026-11-13) #758's comments from the 16th, after its last push, and SLEEP-PLAN.md on main doesn't have them (its status line still says device-panther 0.1.3: USB comes back after a replug, and the battery plan #247 isn't in a release). The matrix only cites merged notes. Could you put the 11-16 runs and the hangs (with the NV check after each) into MODEM-PLAN, add a row to SLEEP-PLAN S1's table, and link those from the cell?
  3. One clause back. "So sleep tests stay attended" has gone, and AGENTS.md's 9a rules haven't changed: a new sleep test still runs attended first, and with the modem up the inhibitor stays, your palmtop under Pixel 9a: on the owner's palmtop, sleep tests with the modem up may run unattended #809 excepted. A short clause saying so keeps the cell from reading as "anything goes now".
  4. The Modem cell. The Suspend cell says the awake hangs belong there, and the Modem cell doesn't mention them. Please add them in this pull request: twice on 2026-11-16 on the published branch, once in everyday use with Wi-Fi connected, cause open.

One wording point: the 86 sleeps were each ended by the modem within 29 s, so none reached its alarm; please say that, so nobody reads them as 86 timer wakes.

My two decisions, and I'd like your view on the second. On the release: the 9a has no release image, so what a 9a owner builds is the published branch, and I'm inclined not to wait for the tag as long as the cell keeps "kernel #247 comes with the next release". On the series: the three failures were all found in timer-woken sleeps with the modem off, and nobody has run that plain series from the published branch as it is yet. A sleep-series.sh run of a few hundred from 26867fc4e or later, under the sleep entry's conditions, would settle the cell's old condition without argument. Would you be willing to run it?

I'll also open an issue for the awake hang on the kernel side, so it has a home outside #758.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Yes to all four, and yes to the series.

  1. The word. "works (opt-in)" for this pull request, with the reason the toggle ships off. I'd still like to get to plain "works", so once the series is in I'll send the device-tegu change that ships auto-sleep on as its own pull request, and you can judge it there.
  2. The notes. I'll put the 11-16 runs and both hangs, with the NV check after each, into MODEM-PLAN, add the row to SLEEP-PLAN S1's table, bring its status line up to date, and link those from the cell instead of device/tegu MODEM-PLAN: the 9a sleeps with the modem up on a test kernel with the sleep guard (2026-11-13) #758.
  3. The clause. It goes back in: a new sleep test still runs attended first, and with the modem up the inhibitor stays, my palmtop under Pixel 9a: on the owner's palmtop, sleep tests with the modem up may run unattended #809 excepted.
  4. The Modem cell. The awake hangs go in: twice on 2026-11-16 on the published branch, once in everyday use with Wi-Fi connected, cause open.

The 86 will read as what they were: each ended by the modem within 29 s, and none reached its alarm.

On the release I agree: the cell keeps "kernel #247 comes with the next release". On the series: I'll run it tonight, timer-woken, on the charger, with the modem off for the boot, from derek/omamobile as it is, a few hundred sleeps with the switches that showed the third failure (PMDEBUG=1 TRACE=2 PMPRINT=0), and post the result here.

…ell to works (opt-in), with the runs on the published branch

The cell's old condition was a fix for the third no-wake and a series from the published kernel as it is. Kernel derekross#247 is merged, and on derek/omamobile at 26867fc4e (derekross#224, derekross#232, derekross#247, derekross#258) 300 of 300 sleeps with a 60 s alarm woke with the modem off on the night of 2026-11-17: 296 on the alarm and 4 within seconds on the touchscreen's interrupt with nobody at the palmtop. On 2026-11-16, 86 of 86 sleeps entered with the modem up woke, each ended by the modem within 29 s, so none of those reached its alarm.

From the review of derekross#846: the word is "works (opt-in)", since auto-sleep ships off on the 9a (omamobile-device-tegu doesn't install /usr/share/omamobile/auto-sleep-default); the cell says again that a new sleep test runs attended first and that the modem's inhibitor stays, the owner's palmtop under derekross#809 excepted; and it cites merged notes, not derekross#758's comments.

MODEM-PLAN gains the 2026-11-16 section: the runs on the test kernel and on the published branch under derekross#809's terms, the route finding that became derekross#837, and the palmtop hanging awake twice at a power-up of the modem's link (once between the sleeps of a series, once in everyday use with Wi-Fi connected), with the NV check after each. The Modem cell names those hangs. SLEEP-PLAN's S1 table gains a row for the 86 modem-up sleeps and one for the night's 300, and its status line is brought up to date.
@W4HO-ham
W4HO-ham force-pushed the devices-tegu-suspend-works branch from 1a0595b to 85ca091 Compare November 18, 2026 11:06
@W4HO-ham W4HO-ham changed the title DEVICES: the Pixel 9a's Suspend cell to works DEVICES: the Pixel 9a's Suspend cell to works (opt-in), with the runs on the published branch Nov 18, 2026
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

The series ran last night, and the four changes are pushed (85ca091). The title and description are updated too.

The series. sleep-series.sh from main on derek/omamobile at 26867fc4e as it is (7.3.0-rc6-omamobile-push1-g26867fc4e826, the build of the 11-16 runs), with the modem off for the boot (its boot unit masked), on the charger, 60 s alarms, PMDEBUG=1 TRACE=2 PMPRINT=0. This build has no tracer, so TRACE=2 recorded nothing; the stall panic and the 60 s runtime watchdog were on. The NV partitions read SAME against the backup before it.

  • 300 of 300 woke, 20:59 to 04:14: suspend_stats 300 and 0, no UFS line, no oops, no stall, no watchdog timeout, one boot throughout.
  • 296 were ended by the alarm after 59.2 to 60.3 s (pm_wakeup_irq 53, the PMIC).
  • 4 were ended early by the touchscreen's interrupt, after 2.9 to 5.2 s (pm_wakeup_irq 58, spi0.0): the 29th, 199th, 247th and 298th. Nobody was at the palmtop; it lay on the charger with a full pack. The touchscreen has been a wake source since device-tegu 0.2.20 (the double tap), so this is the first long series with it armed, and I don't know what it saw. In everyday use sleepd would put the palmtop back to sleep 5 s later. I'll look at whether it follows the charger.
  • I followed the first two cycles from the laptop; the rest ran with nobody there.

The changes.

  1. The word is "works (opt-in)", with the reason: omamobile-device-tegu doesn't install /usr/share/omamobile/auto-sleep-default.
  2. The 11-16 runs and both hangs are in MODEM-PLAN, with the NV check after each (the one after the morning hang is from 11-17, all SAME), and the cell cites that section. SLEEP-PLAN S1 has a row for the 86 and one for last night, and its status line is current.
  3. The clause is back: a new sleep test runs attended first, and with the modem up the inhibitor stays, my palmtop under Pixel 9a: on the owner's palmtop, sleep tests with the modem up may run unattended #809 excepted.
  4. The Modem cell names the two awake hangs, cause open.

The 86 now read as each ended by the modem within 29 s, none reaching its alarm.

For plain "works" I have the device-tegu change ready: it installs /usr/share/omamobile/auto-sleep-default, as device-cheetah does. I'll send it as its own pull request once this one is in, unless you'd rather see the two together.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

On the four early wakes: nothing in last night's record ties them to the charger.

upower had the pack at fully-charged from 21:00:53, a minute into the series, until the end. Its voltage history has a reading at every wake, and it falls from 4.317 V to 4.303 V over the seven hours, in steps of 3 mV or less after the first two cycles. Across the 29th, 199th and 247th the reading stays within 1 mV (4.308, 4.305 and 4.304 V before and after). For the 298th the history ends two minutes before it.

The journal has no line from the touchscreen's driver at any of them, only sleepd's resumed, woken by IRQ 58 … gpn0 0 Level spi0.0, so what the controller reported is still open. The next thing I'd try is a series with the touchscreen's input events logged, to see whether such a wake comes with a touch or with none.

@derekross

Copy link
Copy Markdown
Owner

Thank you. All four changes are in as I asked, and the series is the one the old cell was waiting for: 300 of 300 from the published branch as it is, 296 on the alarm, with the wake source and the length of every one written down. The 86 now read as what they were, the hangs are in the Modem cell with the NV check after each, and the cell cites merged notes. I compared the row with main's cell by cell (only Suspend and Modem change), checked that the release tag doesn't have #247 yet, and followed the anchors. Merging it.

On how the night ran: the row says the first two cycles were followed from the laptop and the rest ran alone with your yes for that night, on a build that already had its first sleeps behind it. That is within the sleep entry as I read it, and writing it into the row is exactly what I want from a record.

Three things for afterwards:

@derekross
derekross merged commit 1207c79 into derekross:main Nov 18, 2026
2 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.

2 participants