Skip to content

device/tegu MODEM-PLAN: the SIM test runbook for 2026-11-03 - #167

Merged
derekross merged 6 commits into
derekross:mainfrom
W4HO-ham:tegu-sim-runbook
Oct 30, 2026
Merged

derekross merged 6 commits into
derekross:mainfrom
W4HO-ham:tegu-sim-runbook

Conversation

@W4HO-ham

Copy link
Copy Markdown
Contributor

The palmtop's first SIM arrives on 2026-11-03. This adds the gated runbook for that day to device/tegu/MODEM-PLAN.md, in the plan's style (command, Expect/Pass, stop condition, Rollback, risk per step; read-only steps first): package tests on the laptop, the RIL install, wiring omamobile-sitd to the 9a's hand modem start (its unit Requires= the 7 Pro's omamobile-modem.service, which the 9a doesn't run), the backup check, inserting the SIM with the palmtop on and the CP unpowered (a power-off round on the 9a can cost a slot-B try), modem to ONLINE under a dead-man, a read-only SIM probe, registration and IMS, data and a ping, SMS both ways, then radio off, checksums again and an attended reboot. Calls are out of this test: the 9a has no voice audio route yet (TAS2572 amplifiers, no earpiece path), so they stay a later signalling-only step.

It also records what the 9a has and lacks for the test, what to expect, what stays unqualified after a pass (suspend with the modem on; long ONLINE sessions, after the eight link-cycling hangs of 2026-10-29) and what is recorded without identifiers: no ICCID, IMSI, IMEI, number, carrier name, APN or message text.

Docs only; tools/pr-check passes.

A gated runbook for the first SIM on the palmtop, cheetah's first-SIM-test with the 9a's names and gaps: the RIL is not installed and its unit requires cheetah's modem start, which is masked here, so the runbook installs omamobile-modem-sit and adds a drop-in; the SIM goes in with the palmtop on and the CP unpowered, so no bootloader power-off round costs a slot-B try; calls are out because no audio route exists on the 9a; sessions stay short and attended because of the link-cycling hangs of 2026-10-29. Read-only steps first, a stop condition and rollback for each, and the identifier rules.
…t drop Requires

Tried on the palmtop: a drop-in with an empty Requires= leaves omamobile-modem.service required, since systemd never resets a dependency list. SIM-2 now copies omamobile-sitd.service into /etc without that line. SIM-0 to SIM-2 are done on the palmtop; nothing starts.
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

One correction at 426d3c7, found by running the first steps tonight: a drop-in with an empty Requires= doesn't drop omamobile-modem.service (systemd never resets a dependency list), so SIM-2 now copies omamobile-sitd.service into /etc with that line removed and the 9a's transient modem unit in After=. SIM-0 to SIM-2 are done on the palmtop (omamobile-modem-sit 0.3.3 installed, both units disabled, no APN yet); nothing starts until the card is in.

@derekross

Copy link
Copy Markdown
Owner

Thanks for writing this up so carefully and so early. The layout is exactly what I want for a first SIM: read-only steps first, the SIM inserted with the CP off, calls left out until there's an audio route, and the recording rules spelled out. It's also good to see you caught the drop-in problem in SIM-2 yourself and fixed it. Before 11-03 I'd like five things changed, because this is the riskiest session the palmtop has had:

  1. Slot B. SIM-4 cites an AGENTS entry, "Pixel 9a: an unmarked slot B", that isn't on main (the closest is the line from AGENTS, device/tegu: a fallback to slot A on the 9a costs the Linux root #125, merged today, which also puts marking B on hold). Every boot this runbook may cost spends a try if B isn't marked: the final reboot, the power-off round, a hang. We know where zero tries ends on the 9a. Please copy SLEEP-PLAN's precondition (stop if B isn't marked; this runbook doesn't mark it), and add a cap: one hang or unexpected reset ends the session for the day, rather than "reboot and start again from SIM-5" without a limit.
  2. Stopping the profile. Your own M5 notes say a profile stop hard-reset the palmtop once, and that until M6 we don't stop it by hand but reboot with someone there. SIM-5's dead-man, its rollback, the SIM-6 power-off round and SIM-11 all stop omamobile-tegu-modem by hand. Could those become attended reboots (the unit's stop still runs during shutdown), with SIM-11's checksum compare moved to just after the reboot, before anything starts the modem?
  3. Which profile runs. The table and SIM-1 say omamobile-modem 0.1.4, but the plan says the 27th GNSS try ran with 0.1.9's watcher, and 0.1.4 predates the 9a's chroot support. So I think some files were copied by hand over an older package. The NV checks all live in that profile, so before the SIM goes in I'd like a real package install from the branch, pacman -Qkk omamobile-modem clean, and the version (and the profile's sha256) recorded, with SIM-1 checking that exact version.
  4. Link cycling. Your analysis of the 10-29 hangs ties the early ones to the watcher's open and close of the boot device, and your omamobile-modem-sit 0.3.4: read the CP state on one open file #156 says sitd 0.3.3 would cycle the link thirty times a minute on the 9a. The runbook installs 0.3.3 and keeps the old watcher. I'll try to get omamobile-modem 0.1.9: watch the CP state on one open file #148, omamobile-modem 0.1.10: the modem profile starts once per boot #163 and omamobile-modem-sit 0.3.4: read the CP state on one open file #156 merged first. Then please run with omamobile-modem 0.1.10 and omamobile-modem-sit 0.3.4, or note why not.
  5. The start script. omamobile-tegu-modem-up is the step that turns the modem on, and the runbook relies on it to refuse a second start, arm the sleep inhibitor and set the stop. Its source is only in your local logs. Could you commit it as a bring-up script under device/tegu/ so it can be read alongside the runbook?

A few smaller ones:

  • sitd's safety stop for an unreleased call stops omamobile-modem.service, which is masked on the 9a, so it would just time out. Please add "don't call the palmtop's number during the test" and what to do if the journal says a call release wasn't confirmed.
  • The hang stop names RebootWatchdogSec, but that only covers shutdown. The 10-29 hangs ended in the cluster watchdog reset, so please name that.
  • ls /sys/module/cpif | wc -l in SIM-1 prints the number of entries, not 1, when CPIF is loaded. test -d is what you want there.
  • The unit copy in /etc will hide future package versions of omamobile-sitd.service, so please mark it temporary and re-check it after upgrades.
  • For the power-off round, would a plain poweroff from Linux do the same job as going through the bootloader menu?
  • The runbook names the SIM's network, though its own rule says no carrier names. Your call, but it's easy to drop.

None of this is about the plan's shape, which is good. It's that a SIM, a transmitting radio and an unqualified power-off path together leave less room than the GNSS sessions did.

…t, and runs on real packages

The slot-B precondition and a one-hang cap head the runbook, citing the AGENTS line that is on main. Every stop of the modem unit by hand becomes an attended reboot, since a profile stop hard-reset the palmtop once and the power-off path is unqualified; SIM-11's checksum compare moves to after the reboot. SIM-1 now requires a real install of omamobile-modem (0.1.10 once derekross#148 and derekross#163 merge) with pacman -Qkk clean and the profile's sha256 recorded, and omamobile-modem-sit 0.3.4, since 0.3.3's state reads cycle the CP link. The start script the runbook relies on is committed under device/tegu/bringup. Smaller: no calls to the palmtop during the test and what to do if a release isn't confirmed, the cluster watchdog reset named for a hang, test -d for the CPIF check, the unit copy marked temporary, the Linux poweroff question answered, the network no longer named.
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

All of it at 39aa3e0 (main merged in first, no conflicts).

  1. Slot B: the runbook now opens with SLEEP-PLAN's precondition (stop if B isn't marked; this runbook doesn't mark it), citing the AGENTS, device/tegu: a fallback to slot A on the 9a costs the Linux root #125 line on main, and a cap: one hang or unexpected reset ends the session for the day; SIM-4's text says the same instead of the entry that wasn't on main.
  2. Stopping the profile: no step stops omamobile-tegu-modem by hand any more. SIM-5's dead-man is an attended systemctl reboot, its rollback and SIM-11 are attended reboots (the unit's stop runs at shutdown), the power-off round is an attended systemctl poweroff, and SIM-11's M0.1 compare now runs right after the boot, before anything starts the modem. The M5 hard reset is cited where the rule is stated.
  3. Which profile runs: the table says the 0.1.4 install has newer profile files copied over it, and SIM-1 now requires a real install of omamobile-modem from main with pacman -Qkk clean and the version plus the profile's sha256 recorded; it names 0.1.10 and says what to do if omamobile-modem 0.1.9: watch the CP state on one open file #148 and omamobile-modem 0.1.10: the modem profile starts once per boot #163 haven't merged by then (wait, or 0.1.8 with the old watcher off, recorded as such). Nothing was installed for this.
  4. Link cycling: the run is on omamobile-modem 0.1.10 and omamobile-modem-sit 0.3.4 (omamobile-modem-sit 0.3.4: read the CP state on one open file #156); the table says why (0.3.3's state reads would cycle the link about thirty times a minute; 0.1.9's single open file retires the old watcher).
  5. The start script is in the PR as device/tegu/bringup/omamobile-tegu-modem-up (the palmtop's copy, SPDX MIT, no identifiers), and the table links it.

Smaller: "nobody calls the palmtop's number during the test" with what to do if the journal says a release wasn't confirmed (radio off, record, end with SIM-11's reboot; sitd's safety stop would only time out on the masked unit); the hang stop names the cluster watchdog reset, with RebootWatchdogSec kept only for a hung shutdown; test -d /sys/module/cpif; the unit copy in /etc is marked temporary and SIM-1 re-checks it against the package's unit after every upgrade; on the Linux poweroff: not tried on the 9a under our kernel, and the plain reboot call hung before, so the runbook prefers systemctl poweroff with a 90 s stop condition and keeps the bootloader menu as the fallback; the network's name is gone from the section (the older M4 paragraph above it still names it, untouched here).

…atcher-off fallback

The runbook stops the modem only by an attended reboot, so the start
script's header no longer says to stop it with systemctl. SIM-1's fallback
of 0.1.8 with the old watcher off can't be followed: 0.1.8's profile_start
starts the watcher unconditionally, and turning it off would be a hand edit
that SIM-1's own pacman -Qkk check rejects.

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

Copy link
Copy Markdown
Owner

Thank you, this is a careful answer to every point, and the runbook now reads the way I'd want for a first SIM: one spare cycle, never a hand stop, the NV compare right after the reboot, a real package install. Thanks too for committing the start script; reading it next to the runbook was useful straight away.

I've pushed two small edits to your branch (c3c5fa6):

  • The script's header said "Stop it with systemctl stop omamobile-tegu-modem before a reboot", which is the hand stop the runbook now rules out. It now says the modem is stopped only by an attended reboot, whose shutdown runs the unit's stop.
  • SIM-1's fallback "install 0.1.8 and keep the old watcher off" can't be done with 0.1.8 as packaged: profile_start in profiles/modem.conf starts the watcher unconditionally, so turning it off would need a hand edit, which your own pacman -Qkk step rejects. I dropped that option and kept "record why and wait", with a note saying why.

On my side, I'll try to get #148, #163 and #156 merged well before the 3rd.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

Both edits are right: the header now matches the runbook's no-hand-stop rule, and the 0.1.8 fallback couldn't have passed the package check anyway. Thanks for aiming #148, #163 and #156 at before the 3rd; I'll rebase #163 the moment #148 is in.

@derekross
derekross merged commit 4fae9d5 into derekross:main Oct 30, 2026
2 checks passed
robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…s#167)

The Pixel 9a's first SIM session as a gated runbook: read-only steps first, the SIM inserted with the CP off, calls left out until there's an audio route. Slot B must be marked (the runbook doesn't mark it) and one hang ends the day. The modem is never stopped by hand: every stop is an attended reboot or poweroff, with the NV checksum compare right after the boot. The modem profile is a real package install (0.1.10 with omamobile-modem-sit 0.3.4) with pacman -Qkk clean and its sha256 recorded. The start script is committed under device/tegu/bringup/.

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