Skip to content

Nothing tells the modem the screen is off: an idle, registered CP wakes its link about 49 times a minute #618

Open

Description

@W4HO-ham

Found on the Pixel 9a while testing sleep with the modem up (#601; the same phone service and link driver run on the Pixel 7 Pro).

What was measured (2026-11-09, palmtop idle and registered on LTE, screen off): the CP link came up 49 times a minute (98 cp2ap_wakeup edges, SIT indication 0906 about 51 times a minute), on the bring-up modules and on the in-tree drivers alike.

What exists. omamobile-sitd already has the request: {"action": "screen-state", "on": false} on its socket sends SIT 0902 with 0 and sets the indication filter (0928) to 0. Nothing in the tree sends it. Sent by hand it took the idle link to 3 or 4 link-ups a minute (6 to 8 edges, no 0906), and on: true took it back to 49.

What's wanted. The shell (or whatever owns the display state) telling sitd when the screen goes off and comes on, so the CP is quiet behind a dark screen. It matters for the battery by itself, and it is one of the things sleep with the modem up needs (the CP should be idle when a sleep starts).

Not known. Whether anything the user sees depends on the indications the filter drops (signal bars are the obvious candidate; they aren't shown with the screen off). With the filter at 0 the five sleeps of that boot were still refused by CPIF, so this alone doesn't make a sleep go through.

Record: device/tegu/MODEM-PLAN.md, "Suspend with the modem on, with the sleep fixes" (#601).

Activity

  1. derekross commented on Nov 11, 2026

    @derekross
    Owner

    Thank you, @W4HO-ham, for splitting this out and for measuring both
    directions (49 a minute, 3 or 4 with the request sent, 49 again with it
    back on).

    Your numbers match what @ikegam measured on the Pixel 7 Pro for #556: 47
    optional indications a minute with the screen on and none with it off. So
    it's the same behaviour on both modems, and the request does what we hoped.

    #556 already carries the fix: sitd follows the backlight and sends the
    screen state itself, and a sleep hook asks for "off" before a sleep and
    puts "on" back after the wake if the panel is lit. I'd rather have sitd
    watch the backlight than have the shell tell it, because it then works
    behind the lock and with no session. On your "not known": the filter at 0
    drops the optional signal and dormancy reports and keeps calls, SMS and
    significant registration and data changes, so signal bars are the only
    thing I expect to notice, and only until "on" is sent again. On the 9a that
    matters more than on the 7 Pro, since your modem sends signal reports only
    after it has been told the screen is on.

    This matters on the 7 Pro even without sleep: its modem holds the sleep
    inhibitor, so the phone sits awake behind a dark screen with the link woken
    47 times a minute. I'll ask in #556 whether the screen-state part can land
    ahead of the part that lifts the inhibitor. I'll leave this issue open
    until it does and close it from that merge.

    If you'd like to try #556's sitd on the palmtop before then, the thing I'd
    most like to know is whether the bars come back promptly after screen-on.

    One note on the link in the issue: the MODEM-PLAN section it names comes
    with #601, so it isn't on main yet.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions