diff --git a/device/tegu/NOTES.md b/device/tegu/NOTES.md index 1f77d41e9..58fa7422a 100644 --- a/device/tegu/NOTES.md +++ b/device/tegu/NOTES.md @@ -13,6 +13,97 @@ hash/build ID; this dirty release is not a published kernel package. `[V]` marks verified observations; `[I]` marks an inference still to test. Raw captures, recordings, backups and vendor files remain off repository. +## 2026-11-02, evening: calls, Phone and Messages, echo, a restart and MMS + +All on the daily kernel (`7.3.0-rc6-omamobile-tegu-s2mpg15-m1-g991042e2eff3-dirty`) +with the modem started at boot by `omamobile-tegu-modem-boot` (#372). + +- [V] **Calls and texts from the apps:** Phone and Messages through + `omamobile-mmd` (#371) placed and answered calls with speakerphone audio + both ways and sent and received texts. Messages showed "service isn't + running" until the user's `omatel-agent` unit was started and enabled. +- [V] **Signal and the airplane icon:** the modem sends its signal reports + only after it is told the screen is on, and radio state 10 means on; + with both in `omamobile-modem-sit` 0.3.15 and the bridge 0.1.2, the bar + shows signal and no airplane. +- [V] **A sitd restart in the same boot (#376), a deliberate test:** sitd's + unit says a second start-up isn't supported, and this checked what + happens. It failed at once on + leftovers in `/run/omamobile-sit` (the LTE-only setting's saved original, + then the bearer records), though the modem had dropped the bearers when + sitd stopped. With 0.3.18 a restart gets data and signal back, but IMS + doesn't register again (no registration indication after the PDN + information is accepted), so calls and texts need a reboot after a + restart. [I] The modem's IMS stack is never told the old session ended. +- [V] **Echo cancellation (#374, #378):** the owner's tuning files from + the factory image upload with the AoC's status 0. The AoC's reply + rewrites the second header byte (0x01 sent, 0x3e back), which the first + check took for an unknown outcome, so that call was silent (fixed in + 0.3.19). With the tuning alone the far end still heard its own voice; + with `Audio DSP Mode` set to `Telephony_APMG3` (0.3.20) it heard no echo + for the first few seconds and a new low hiss, then the echo again with + the hiss. [I] The echo reference or the order of the setup is wrong. +- [V] **MMS (#380):** with the carrier's MMS APN equal to its internet APN, + setting up the separate MMS bearer (CID 3) took the IMS bearer down: + afterwards the modem listed bearers 1 and 3, not 2. sitd held the + reservation as uncertain, as designed, and a reboot cleared it. 0.3.22 + sends MMS over the internet bearer instead; not yet tried. +- [V] **Wi-Fi after a sleep with the keep flag:** in a boot with the + modem's boot start disabled and the CP never started (checked before the + sleep test), after the shell's auto-sleep with + `/etc/omamobile/wifi-keep-across-sleep` in place, + brcmfmac didn't attach to the chip again; a reload, a bind and a PCI + remove and rescan all left no interface. A reboot brought it back. + +## 2026-11-02: the SIM, the in-tree modem stack with GNSS, and a pruned module tree + +Boots as before: test kernels pushed, checked, and the daily kernel pushed +back. Details are on each pull request. + +- [V] **The SIM (omarchy-mobile #356, the daily kernel):** LTE + registration, mobile data over `rmnet0` and texts both ways, after five + data setups refused with status 36 and a wait after the data permission + in `omamobile-modem-sit` 0.3.8. The record is in + [MODEM-PLAN](MODEM-PLAN.md), "SIM test results". +- [V] **GNSS driver first boot (kernel #185, G1, + `7.3.0-rc6-omamobile-push1-g30e9d94b4a91`):** `gnss_spi` and `gnssif` + bound with the modem off, `/dev/gnss_ipc` appeared, no request + interrupts, wakeup armed only while the device was open, and an unload + and reload were clean. Then a 10-minute session on the daily kernel with + #185's modules built for it: 538 of 556 GGA sentences fixed. +- [V] **The modem's two-lane PHY (kernel #186, bench, the modem off):** + the first bench locked every PLL on one cycle and then hung the SoC at + the next cycle's first ELBI write: power-off left the controller on the + PHY's stopped clock. With the clock parked on the oscillator first, as + stock does, five cycles over two boots locked every time + (`7.3.0-rc6-omamobile-push1-g909c8c8121bb`, which also showed both HSI2 + GPIO bus clocks held on with kernel #172's second commit). +- [V] **The in-tree modem stack (kernel #187):** the gs201-cp host with + zumapro's channel 0 through #186's PHY linked to the modem's ROM three + times at Gen1 x1 (`7.3.0-rc6-omamobile-push1-g1b270b5cc447`); then, with + a cpif fix for an uncached link without IO coherency, CPIF booted the CP + to ONLINE and a 10-minute GNSS session through #185 gave 543 of 549 GGA + fixed, 1343 link power cycles, no reset (`7.3.0-rc6-omamobile-push1-gbaf1b52d1de5`). + The first try's cpif probe had failed ("PCIe IOCC config and dt are + inconsistent"). +- [V] **A pruned daily module tree:** after six test kernels in one day, + the push back to the daily kernel replaced its module tree with the + kit's archive, without the hand-installed `extra/` and `updates/` + modules, and the next boot stopped at the boot logo (display driver + missing). The cause: `omamobile-modules-cleanup` keeps the running tree, + the next boot's and the two newest others, so during a test boot it + removed the older daily tree; the push back then found no marker and + sent the archive. Restored from the backup in `/var/lib/omamobile-tegu/`; + on the palmtop the cleanup now keeps 50 trees (a local drop-in) until a + pin for named releases ships. Until then, check `ls + /usr/lib/modules/$(uname -r)/extra` after any push back to the daily + kernel. +- [I] **A leftover bring-up unit:** `/etc/systemd/system/omamobile-tegu-audio.service` + (2026-10-28, before `omamobile-device-tegu` took over its steps) now + fails at every boot: its script loads NFC's module by hand after the + package already has. The package's units bring up the same hardware, so + it should go, as AGENTS.md asks for bring-up files a package replaces. + ## 2026-11-01, afternoon and evening: Wi-Fi from the package, the wake and the replug, storage idle, the panel ID Boots as in the next section (published-style kernels, pushed, checked,