Skip to content

device/tegu, omamobile-device-tegu: the Pixel 9a's modem and GNSS from the published kernel's own drivers - #594

Merged
derekross merged 5 commits into
derekross:mainfrom
W4HO-ham:tegu-modem-cpif-overlay
Nov 11, 2026
Merged

derekross merged 5 commits into
derekross:mainfrom
W4HO-ham:tegu-modem-cpif-overlay

Conversation

@W4HO-ham

Copy link
Copy Markdown
Contributor

A 9a built from the tree can't start its modem from the kernel's own drivers: the board's device tree (kernel #187) has the modem node under the PCIe link and no cpif node, as you pointed out on #590, so omamobile-tegu-modem-up (#372) could only take its bring-up branch. This brings the missing pieces, and the first runs of the in-tree branch through the starter.

What changes

  • device/tegu/overlays/modem-cpif.dtso: the published form of the bring-up tegu-modem-cpif.dtso. It disables pcie@12100000/modem for the link driver and adds CPIF's cpif and cp-shmem nodes. Paths instead of labels and no references into the base tree, so fdtoverlay needs nothing of the base DTB's symbols, and the bring-up activation module isn't needed.
  • device/tegu/device.conf (and the package's copy): DTB_OVERLAYS="modem-cpif".
  • kernel/configs/omamobile.config: CONFIG_EXYNOS_GNSS_IF_SPI=m, the GNSS link driver from kernel omamobile-omarchy 0.1.18: the lock clock follows the bar's 12/24-hour setting #185.
  • omamobile-device-tegu's guards: gnss_spi and gnssif blacklisted next to CPIF's modules. With the GNSS nodes in the device tree, udev loaded both on a boot without the modem; they are of use only with the modem up, and the starter loads them by name after CPIF. No version bump, since three open PRs already claim the next numbers; say which you want.

Binding CPIF powers nothing, its modules stay blacklisted, and nothing here starts the modem: that is still #372's unit, opt-in.

Tested on the palmtop (2026-11-09, 7.3.0-rc6-omamobile-push1-g773b4102f5ec: omamobile at abcd6aac51cf plus kernel #224, kernel #236 and the 9a's GNSS nodes, built with defconfig, this config and #274's CONFIG_PHY_ZUMA_PCIE=m; the published DTB with this overlay; #372's boot unit enabled; on the charger, watched over SSH, the owner away for the first boot):

  • Three boots in a row, the unit by itself each time: "drivers: the running kernel's", CP ONLINE 8 s after the unit started, the phone service and the bridge up, registered at home on LTE, /dev/umts_ipc0 and /dev/gnss_ipc there, no UFS error-handler pass, Wi-Fi up beside it.
  • Two systemctl reboots with the modem up: "CP OFFLINE", "modem stopped (success, the CP OFFLINE before the daemons went)", the latch removed. A boot with the unit masked started nothing and left no latch.
  • Four 150 s GNSS sessions through the Location switch's socket, indoors by a window: 108 of 113, 142 of 143, 105 of 107 and 142 of 144 GGA sentences fixed, up to 24 satellites; each stopped on its idle lease with the CP still ONLINE.
  • One finding for the kernel side: with the GNSS bus on gs101's compatible alone (automatic chip select, as stock has it), spi-s3c64xx cut each 64-byte block into 60 and 4 bytes and 12 reads timed out in 13 minutes. With zumapro's compatible kept, none in two sessions. The GNSS nodes go to a kernel pull request with that setting; I'll link it here.

Not run yet: a boot without the modem on the package with the new guard lines (the check that the two GNSS modules no longer autoload), and a call on these drivers. Both are next, and I'll add them here.

tools/pr-check, tests/device-overlays, tests/stage-boot-files and the package's test/all pass. The record is in MODEM-PLAN, after section P.

…em and GNSS from the published kernel's own drivers

The board's device tree has only the modem node under the PCIe link, so a 9a built from the tree had no cpif node and the modem start could only use the owner's bring-up modules. device/tegu/overlays/modem-cpif.dtso is the published form of the bring-up overlay (paths, no labels): it hands the modem node to CPIF and adds the cpif and cp-shmem nodes, and the descriptor lists it in DTB_OVERLAYS. The shared config builds the GNSS link driver (CONFIG_EXYNOS_GNSS_IF_SPI=m, linux-omarchy-mobile#185), and the 9a's guards keep gnss_spi and gnssif from autoloading, like CPIF's modules, so only the modem start loads them. MODEM-PLAN records the run on the palmtop: three boots by the boot unit on the in-tree drivers, the confirmed stop, four GNSS sessions, and why the GNSS bus keeps zumapro's SPI port configuration.
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

The two things that were still open, and the kernel side:

  • The guard lines, on a boot without the modem (the package with them installed on the palmtop, the unit masked): cpif, /gnss and the SPI device are in the device tree, and none of cpif, shm_ipc, pcie_gs201_cp, gnss_spi, gnssif was loaded; no latch. On the boots with the modem the starter loaded all five by name, as before. A Power-key wake from s2idle on the modem-less boot was clean (kernel AGENTS: on the Pixel 9a a plain fastboot flash of slot B clears its successful flag too #224).
  • A call, and data, on these drivers. An incoming call on the fourth boot: ringing, accepted through the bridge, active for 8 s, hung up, the release confirmed 272 ms after the hang-up, the CP ONLINE and no UFS line after (details go to omamobile-mmd 0.1.3: wait for confirmed call release #461). Mobile data: omamobile-sit-data brought rmnet0 up and three of three pings went out over it. omamobile-sit status each boot: SIM ready, home LTE, IMS voice and SMS, data bearer up.
  • The GNSS nodes are kernel omamobile-shell 0.2.56: fluid home and drawer gestures #237, with the bus setting described above.

That makes four boots by the unit on the in-tree drivers and three confirmed stops at systemctl reboot. Texts and the NV check after them are still to run on this kernel.

@W4HO-ham

Copy link
Copy Markdown
Contributor Author

The last open item, texts on these drivers (2026-11-09, 20:40 MST, a sixth boot of g773b4102f5ec by the unit): one text sent from the palmtop with omamobile-sit sms send (AIMS_SEND_SMS: error=0) arrived on the owner's phone, and his reply reached the palmtop 22 s after the send ("SMS stored and acknowledged" in sitd, then the bridge's SMS/0 received, taken by the Messages app). In MODEM-PLAN with 725c0503.

So on the published kernel's own drivers the 9a has now done: registration, a call, texts both ways, data, GNSS fixes, and confirmed stops, with the NV partitions unchanged.

One related fix went to #590 rather than here, because it is in #372's starter: the start now loads phy_zuma_pcie itself (until then only the Wi-Fi step did, and only with Wi-Fi's firmware on the phone). It ran on the palmtop with the firmware aside for a boot.

@derekross

Copy link
Copy Markdown
Owner

Thank you for this. It closes exactly the gap I pointed at on #590, and it does so cleanly: paths instead of labels so the base needs no symbols, the modules kept behind the guards with the test extended to the two GNSS ones, and the check on a boot with the unit masked that none of the five loaded. The finding about the GNSS bus (60 and 4 byte pieces on gs101's compatible, none with zumapro's) is the kind of thing that saves the next person a week; thank you for carrying it to kernel #237.

I checked the Pixel 7 Pro side: the two GNSS modules match compatibles that no device tree in omamobile has yet, and the option's dependencies are already on, so the shared fragment only gains two modules nothing loads. I also compiled the overlay and applied it to a DTB from today's tree, where it applies.

Three things before I merge:

  1. The overlay needs a kernel with device/husky: UFS-PLAN: U4, the Omarchy image boots #187. On the DTB of the last tagged kernel (kernel-2026.11.08, before the modem node) fdtoverlay fails with FDT_ERR_NOTFOUND, and bin/omamobile images then stops at "applying the overlays failed" without saying why. That's the safe failure, but please add a line to the device.conf comment naming the kernel it needs. A question with it: what happens on the owner's build device/tegu, DEVICES, AGENTS: the 9a on 2026-10-27 and 10-28 #100? Does its device tree have pcie@12100000/modem, and with the overlay applied there, does the starter take the kernel's drivers or the bring-up modules?
  2. "The published kernel". Section P lists what actually ran: the tip plus kernel AGENTS: on the Pixel 9a a plain fastboot flash of slot B clears its successful flag too #224, omamobile-audio 0.5.8, device-cheetah 0.1.36: louder speakers under the firmware's protection #236, omamobile-shell 0.2.56: fluid home and drawer gestures #237's nodes and kernel config: the zuma-family PCIe PHY (CONFIG_PHY_ZUMA_PCIE) #274's PHY config, all still open. So a 9a built from the tree after this merge has the cpif node but no PHY driver and no GNSS nodes yet. Could P's first sentence say that, instead of "with nothing of the owner's kernel"? I'll word the merge subject the same way, and I'll look at kernel config: the zuma-family PCIe PHY (CONFIG_PHY_ZUMA_PCIE) #274 alongside this.
  3. Who was there. The description says the first boot ran with the owner away, watched over SSH. Section P only says "on the charger". Please put it in P as well, for each boot. It was the first start of the link driver with allow_hw=1 through the starter on the 9a, and I'd like the record to show how it ran.

On the version: you were right to leave it. I'll set it at merge; this one takes device-tegu 0.2.11, and #372 and #590 follow.

If you have a tools/phone-smoke report from one boot with the overlay and the unit masked, it would be good to have here.

… kernel and attendance as they were

The descriptor's comment names the kernel the overlay needs (the modem
node of linux-omarchy-mobile#187) and how an older DTB fails. Section P
no longer calls its kernel the published one: it was the tip plus four
things then open, two of them merged since. It lists the six boots with
their times from the journal and who was at the palmtop for each (the
first two: nobody), and says what the overlay does on the owner's daily
kernel: it doesn't apply, and the starter takes the bring-up modules.
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

All three are in (6e4f8e7).

  1. The kernel it needs. The device.conf comment (both copies) now names it: a kernel whose zumapro-tegu.dtb has the modem node of kernel device/husky: UFS-PLAN: U4, the Omarchy image boots #187, so one built after the kernel-2026.11.08 tag, and says how an older DTB fails.

    Your question about build device/tegu, DEVICES, AGENTS: the 9a on 2026-10-27 and 10-28 #100: the overlay doesn't apply there either, and nothing changes. I tried it on the laptop today: device/tegu, DEVICES, AGENTS: the 9a on 2026-10-27 and 10-28 #100's base DTB is the integration tree's and has no modem node under pcie@12100000, so fdtoverlay fails with the same FDT_ERR_NOTFOUND. That kernel is built by my kit with the kit's own overlay list, not this descriptor's. Its running tree does have a disabled modem node, cpif and cp-shmem, from the bring-up overlay, and its module tree has none of the CP drivers (modinfo cpif: not found), so the starter takes the bring-up modules: "drivers: the bring-up modules in /var/lib/omamobile-tegu/modem", read on it this morning. That's in section P now.

  2. "The published kernel". P's first sentence says what it was: the tip plus four things all open that day. A new bullet says where that stands: kernel omamobile-audio 0.5.8, device-cheetah 0.1.36: louder speakers under the firmware's protection #236 and omamobile-shell 0.2.56: fluid home and drawer gestures #237 merged today, AGENTS: on the Pixel 9a a plain fastboot flash of slot B clears its successful flag too #224 and kernel config: the zuma-family PCIe PHY (CONFIG_PHY_ZUMA_PCIE) #274 open, and until kernel config: the zuma-family PCIe PHY (CONFIG_PHY_ZUMA_PCIE) #274 a build from the tree has the cpif node but no PHY driver, so the link can't come up there.

  3. Who was there. P has it per boot, with the times from journalctl --list-boots: six boots by the unit, the modem up at 18:32, 18:50, 19:12, 19:26, 20:26 and 20:37 MST. The first two ran with the owner out of the house, watched only over SSH; that includes the first start of the link driver with allow_hw=1 through the starter. The third with him home but not at the palmtop. The fourth was the call, which he made and watched. The fifth and sixth he was at the laptop for, and the texts on the sixth went to and from the phone in his hand. (The sixth was "20:40" in the text before: that was the texts' time, not the boot's.)

The tools/phone-smoke report from a boot with the overlay and the unit masked comes tonight: the palmtop is on the daily kernel for the owner's day, and the next boot with the overlay is this evening's.

@derekross

Copy link
Copy Markdown
Owner

Thank you. All three are answered the way I hoped: the descriptor now says which kernel the overlay needs and how an older one fails, section P says plainly what the test kernel was, and the record of who was there for each boot is exact, including the part that isn't comfortable to write down. The answer about build #100 settles my question too: the overlay doesn't apply there, nothing changes for the owner's daily kernel, and the starter keeps taking the bring-up modules.

I merged it onto today's main here (it is clean beside #274's config line and #372's package changes), ran the package's tests on the result, and compiled and applied the overlay again: it applies to a DTB from today's tree and fails on the tagged one exactly as your comment says.

I'm merging this first, as device-tegu 0.2.14. That number matters now: kernel #237 put the GNSS nodes into the tree today, so the two new blacklist lines have to reach the phones with the config line.

One small thing for your next push to #590 rather than a new one here: section P says #274 is still open and that a tree build has no PHY driver until it lands. It was merged a few minutes before your commit, so after this merge a 9a built from the tree has everything the in-tree start needs. Please bring that sentence up to date together with the DEVICES cells.

The smoke report from a boot with the overlay and the unit masked is still welcome whenever the palmtop is next on that kernel.

Main moved since this branch was cut, and the branch set no number of its
own, so the squash would have shipped its changes under a version phones
already have.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@derekross
derekross merged commit cb8ebdf into derekross:main Nov 11, 2026
2 checks passed
W4HO-ham added a commit to W4HO-ham/omarchy-mobile that referenced this pull request Nov 11, 2026
…ig lines are merged

derekross#274 (the PCIe PHY's config line) and derekross#594 (the modem-cpif overlay and the GNSS link driver's config line) are in, so the 9a's Modem and GNSS cells no longer call them "in review" or warn about a tree build without the PHY's driver, and section P says a kernel built from the tree now has everything the in-tree start needs. omamobile-device-tegu 0.2.15, one above main.
@W4HO-ham

Copy link
Copy Markdown
Contributor Author

The tools/phone-smoke report from a boot with the overlay and the unit masked, as promised. It is from 2026-11-11T02:46Z, three minutes into the boot, on the build of this evening's sleep series: the sources of the 212-wake series (push1-gff4fdf58035e, today's published tree plus kernel #224) with the function tracer built in, the published DTB with the public-ramoops and modem-cpif overlays applied by the kit, omamobile-tegu-modem-boot masked, the modem not started.

phone-smoke: tegu

Kernel 7.3.0-rc6-omamobile-push1-dbg-gff4fdf58035e, tools at d1fb7b6a, 2026-11-11T02:46Z.

Check Result Detail
boot (required) ok kernel 7.3.0-rc6-omamobile-push1-dbg-gff4fdf58035e, slot b, up 183s
storage guard (required) ok 62 partitions; only userdata and slot B's boot images writable
root (required) ok ext4 on userdata, 90G free
kernel health (required) ok no oops or BUG; 1 WARN, 54 error lines in dmesg (39 known, 15 new)
new dmesg errors not ok samsung-pinctrl N: irq number not available x6; maxtcpc N: Failed to get vbus regulator handle x3; module goodix_fp: .gnu.linkonce.this_module section size must match the kernel's built struct module size at run time x2
systemd units not ok failed: tegu-battery-hold.service tegu-gpu-dvfs.service tegu-nfc-clkpad.service
USB gadget (required) ok bound to 11210000.usb
display ok card1-DSI-1 connected, 1080x2424
GPU ok render node on panthor
session (required) ok Hyprland and the Omarchy shell running
lock ok PAM file present, the shell can lock (passwordPam true, locked true)
touch ok Synaptics TouchComm Touchscreen
buttons ok gpio-keys
Wi-Fi ok wlan0 on brcmfmac, link up
Bluetooth ok hci0, powered true
battery ok 89% Full, 31 °C
sound cards skip none
sensors ok Accelerometer AmbientLight Proximity
cameras not ok libcamera lists none (or a camera app holds them)
thermal ok hottest zone 41 °C

Packages: hyprland 0.57.0-3.1, libcamera 0.7.2-4.22, mesa 1:26.2.4-1.4, omacam 0.3.3-1, omamessages 0.2.2-1, omamobile-apps 0.2.2-1, omamobile-audio 0.5.13-1, omamobile-device-tegu 0.2.17-1, omamobile-fingerprint 0.1.8-1, omamobile-firstboot 0.1.0-1, omamobile-gnss 0.2.8-1, omamobile-keyring 20261026-1, omamobile-mirror 0.1.1-1, omamobile-mmd 0.1.4-1, omamobile-modem 0.1.17-1, omamobile-modem-sit 0.3.36-1, omamobile-omarchy 0.1.37-1, omamobile-owl-widgets 0.1.2-1, omamobile-sensors 0.2.1-1, omamobile-shell 0.2.101-1, omamobile-telephony-tegu 0.1.0-1, omaphone 0.2.3-1, omarchy 4.5.0-3, omatel-agent 0.2.4-2, quickshell 0.3.2.r7.g2b4501a-1.1

The required checks pass. The rows that don't are the palmtop's bring-up state meeting this kernel, not the overlay: the three failed units are the palmtop's own bring-up units (not in the repository), goodix_fp is a bring-up module built for another kernel, and this kernel shows no sound card and no cameras on the palmtop. The palmtop's omamobile-device-tegu is a hand-built 0.2.17 from the bring-up branches, not main's 0.2.14.

robertzas pushed a commit to robertzas/omarchy-mobile that referenced this pull request Nov 14, 2026
…a's modem-cpif overlay and GNSS link driver (derekross#594)

Adds device/tegu/overlays/modem-cpif.dtso to the Pixel 9a's DTB_OVERLAYS (it needs a kernel with linux-omarchy-mobile#187), CONFIG_EXYNOS_GNSS_IF_SPI=m to the shared fragment (two modules that bind only to the 9a's nodes), and gnss_spi and gnssif to the 9a's guards, so they don't load by themselves now that the kernel has the 9a's GNSS nodes (linux-omarchy-mobile#237). Nothing here starts the modem; that stays derekross#372's opt-in unit.

MODEM-PLAN records six boots on the owner's palmtop on 2026-11-09, on the published kernel tip plus kernel derekross#224, derekross#236, derekross#237 and derekross#274's config: registration, a call, texts, data, GNSS fixes and confirmed stops, with the NV partitions unchanged. The Pixel 7 Pro gains two modules that never load.

Co-Authored-By: Claude Fable 5.2 <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