Repository navigation
Conversation
|
Thanks for this. It's a lot of careful work, and I like how much of it fails closed: the uncertainty marker for the AoC tuning, the bearer reservations kept for attended recovery, the MMS refusal when the APNs are shared, and the call sink staying muted when cleanup can't be confirmed. Getting audio both ways with the AoC confirming the tuning, and finding the DSP mode, is real progress even though the echo comes back. Since these are drafts, this is direction rather than a merge review, for #374, #375, #376, #378 and #380 together. Ordering. #356 and my #381 (modem-sit 0.3.20, mmd 0.1.1) are on main now and touch the same files (
#376. The NR-mode part is the same change as in #381; yours also checks the saved file, so keep yours. On the bearer records: your restart test ran on 0.3.9's start-up, which switches the radio off and on, so "the modem lists no bearer" was measured after that cycle. #381 skips the cycle when the radio is already on, so please re-test after the rebase. And #374 and #378. #380. Some of this reaches the 7 Pro, so I'll run it there before it merges: the SIT transport, the binary-SMS change, the #374's failed CI run was a mirror download error, so a re-run should clear it. Before any of these leave draft, each needs a |
853247c to
e95b059
Compare
|
Thanks for going through all five. The chain is rebased on main (with #364, #367, #371 and #372 under it), the versions are left for you, and each point is in the PR it belongs to: Ordering. The radio-state and signal-decode commits are dropped from #371 and #372. The screen-on at bring-up is 9a-only now (#372): tegu's sitd drop-in sets #376. I kept my NR-mode check and folded in the IMS restart fix (DelPdnInfo for the earlier start's PDN before registering again). #374 and #378. CALL-AUDIO-TUNING records the three calls of 2026-11-02 with the kernel release and what the far end heard. It also says the reply's second byte is the counter, which the check skips, and the kernel link now points at #380. The wording is fixed: the address is kept out of #374's CI should pass on this push. Phone-smoke reports for each come with the next palmtop session. |
da9464f to
e9e74ee
Compare
efc3ad0 to
e8cd1b3
Compare
Select the TAS2572 route only for google,tegu-aoc-snd-card. Save the controls, verify sink mute, configure the stock microphone/TDM routes, and start EP4 capture and EP5 playback as mono S16_LE at 8 kHz with 80-frame periods and two periods. Applying ALSA hardware parameters prepares each PCM once; do not prepare the modem stream twice. After PCM startup resets the amplifier profile, set both TAS digital gains to index 180 (-20 dB) and verify read-back before unmuting. A gain refusal unwinds the route and leaves call signalling available. Close PCMs before restoring controls, keep mute on cleanup failure, and retain the cheetah CS35L41 route and the default card=none configuration. Add fake-ALSA fault tests and the voice-window audit. The existing SRAM reservation is retained; no CPU mapping or phone operation is added.
Outgoing and incoming speakerphone calls with audio both ways, first on the daily kernel and then on in-tree drivers from the open kernel PRs, where data and a text ran too. The Modem cell says so, and what's left: echo cancellation, an earpiece, the modem's start at boot and the kernel PRs' merge.
Trace the existing app, session agent, ModemManager-compatible bridge and sitd path. Record the tegu installation and unit prerequisites still missing on main, and an attended bridge-only test that preserves the running modem stack. Document the passing offline bridge and apps package suites, activation and audio limits, and the eventual board-specific image integration.
Keep sitd's canonical name while moving the modem dependency into device-owned policy drop-ins. Tegu checks an already running modem and CP ONLINE; cheetah and raven retain their existing starter relationship. Require the device policy capability during RIL upgrades. Add an opt-in telephony package with a bridge policy and attended helper. It checks live sitd readiness, refuses legacy unit shadows and competing managers, and starts only mmd and the graphical user's agent. The service recovery check permits live calls and registration loss; no helper enables or recovers the modem. Remove mmd's service-level conflict so a bridge start cannot stop another modem manager. Package exclusion and D-Bus name ownership remain. Retain the previously developed SIM, safe-release and audio changes in this stack. Validate the helper, dependency policies and composed units offline; run all seven affected/dependency package suites and build the new architecture-independent package without installing dependencies.
…start the bridge, and a plan that reads as one Record the palmtop run of Phone and Messages (kernel, versions, what passed, and that opening an app didn't start the agent), say which start fits which case, cite derekross#363 instead of a branch, and drop the lines that read like a task report.
…in 0.1.3, with this pull request The plan and the README dated the removal of the unit's Conflicts=ModemManager.service to mmd 0.1.1, and the bullet above still listed it as built; this pull request is what removes it, in 0.1.3.
…WantedBy and tmp.mount, and a refusal isn't retried With derekross#372's boot unit on and this package installed, the bridge's ExecStartPre refused every start: sitd wanted omamobile-mmd.service (the bridge's own WantedBy, once enabled) and both units wanted tmp.mount (systemd adds it). Neither pulls the RIL or the CP into a start, so both are allowed now. A refusal of --service-check exits 3, which the drop-in names in RestartPreventExitStatus=, so a refused bridge stays down instead of retrying every five seconds.
…ot, opt-in The 9a's modem only started by hand, from a bring-up script outside the package. omamobile-tegu-modem-up is now packaged: the kernel's own CP drivers when it has them, else the bring-up modules; the NV check, then a sleep inhibitor, then the drivers; a latch written before the drivers and cleared at CP ONLINE, so a start that never got there stops later boots until a human looks; the once-per-boot refusal; and with --phone the phone service and, when installed, the Phone and Messages bridge. omamobile-tegu-modem-boot.service runs it at boot, installed but never enabled by the package. A scratch-root test covers the driver choice, the order, the latch and the refusals.
# Conflicts: # tests/all
|
Thank you for bringing the four drafts back onto main, and for doing it so that each one's diff is only its own change. I checked this one: against today's main it is the tuning upload, its tests, the extractor's six files and the plan, nothing from the PRs that have merged underneath it, and the defaults are still off. CI is green and it merges cleanly. It stays a draft for now, for the reasons you list yourself. One thing I'd like cleared up in the plan before the trial: When the palmtop is free of the sleep series: the speaker trial first, the earpiece only after it passes and away from the ear, then the smoke report from the head that ran. Two housekeeping points for when it leaves draft, no hurry: the title still says 0.3.16 and the description still says it is stacked on #372 and that nothing has run. And 0.3.35 is also claimed by #669, #389 and #376, so I'll set the number at merge. |
|
It comes from calls with
This branch's copy predates those calls, which is why it still says "no phone trial" and "not run": this head by itself has never run on the palmtop, since the first call needed #378's counter fix to be anything but silent. The description says that now, and the title has lost its number. Not tried on a call yet: the last two commits of #378 (the modem mixer's ASP on, and the tuning sent after the voice streams start), and the earpiece route. The speaker trial comes first, with the owner at the palmtop and on the far end, then the smoke report from the head that ran. |
|
The speaker trial ran this evening, and the result is negative again: the far end hears no echo for about the first 2 s of a call and its own voice for the rest. What ran. The calls. Two, from the palmtop to the owner's iPhone, he on the far end: 25 s and 14 s. Both had audio both ways. sitd logged "AoC tegu speakerphone prepared at conservative TAS gain" at each start and "AoC call routes restored" at each end, and no refusal, so both uploads were acknowledged. In his words: he could not hear himself for about the first 2 seconds, then for the rest of the call he could. So #378's last two changes (the modem mixer's ASP on, tried on 11-03, and the tuning sent after the voice streams start, tried today) leave it where it was: the cancellation works for about 2 s and stops. The earpiece route wasn't tried, since the speaker trial didn't pass. A slip to own on #378's branch: my push
phone-smoke: teguKernel
Packages: hyprland 0.57.0-3.3, libcamera 0.7.2-4.23, 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 0.5.18-1, omamobile-device-tegu 0.2.16-1, omamobile-fingerprint 0.1.17-1, omamobile-firstboot 0.1.4-1, omamobile-gnss 0.2.9-1, omamobile-keyring 20261026-1, omamobile-mirror 0.1.1-1, omamobile-mmd 0.1.4-1, omamobile-modem 0.1.20-1, omamobile-modem-sit 0.3.37-1, omamobile-omarchy 0.1.43-1, omamobile-owl-widgets 0.1.2-1, omamobile-sensors 0.2.1-1, omamobile-shell 0.2.125-1, 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 |
|
A wrong time in my comment above: the calls were at 03:45Z (20:45 MST), not 02:45Z. The smoke report's 01:58Z is right. |
|
Thank you for running the speaker trial with the owner on the far end, and for reporting a negative result as carefully as a positive one: what was built and from which head, that #677 had to go on top to build it, both uploads acknowledged, and his own words for what he heard. Five calls now tell the same story, so "the cancellation holds for about 2 s and then stops" is a finding I trust. Leaving the earpiece untried because the speaker trial didn't pass is the right call. This stays a draft, and I think you see it the same way. For when you come back to it:
One question: the extractor now stops when any of the six tuning files is missing from the image, and every 9a owner runs it, also with the feature off. On which factory builds have you seen all six? If there is any doubt, a warning would be safer than a stop while this is experimental. A note so it doesn't surprise you: #626 also adds an earpiece route for the 9a, with different levels and a live switch, and the two can't both merge as they are. Two people have now built this for the same phone, so I'd rather not pick from here: could you and #626's author compare the two on the palmtop, levels and switch included, and tell me which design should carry on? I'll take the one you both stand behind, and I'll say the same on #626. The number is mine to set at merge; 0.3.35 is claimed five times over at the moment. |
… is a warning, not a stop Every 9a owner runs the extractor, with the call tuning off too, and the six files have been seen on one factory build only (BP3A.250905.014). A missing one now leaves the set out whole, since the loader needs a route's three files together, says so, and the rest is extracted and checksummed as before.
tests/all is main's, with the call-tuning extractor's test on its own two lines at the end, in the one-suite-per-line block.
…s it really comes The first call with echo cancellation on the palmtop went silent. The AoC accepted the tuning upload with status 0, but its reply rewrites the second header byte (0x01 sent, 0x3e back), and the check compared the first six bytes, so a successful upload was taken as an unknown outcome and the call audio stayed muted. The check now compares the first byte, the size and the command ID, and the fake AoC in the tests rewrites that byte as the real one does. A refused call now also logs why.
…mode for echo cancellation With the tuning uploaded and confirmed (0.3.19), the far end still heard its own voice: the AoC's output DSP stayed in Ambient during calls, and the uploaded tuning is for its APMg3 telephony mode. With tegu_aec on, the call route now sets Audio DSP Mode to Telephony_APMG3 before the upload and puts the old mode back after the tuning is turned off. On the palmtop (a speakerphone call to an iPhone, 2026-11-02) this changed what the far end heard: no echo for the first few seconds and a new low hiss, then the echo came back with the hiss still there. So the mode takes effect, but the cancellation doesn't hold yet; the echo reference or the order of the setup is the next thing to check.
…t from our own kernel header Record the three calls with the kernel release and what the far end heard, describe the command from struct CMD_AUDIO_OUTPUT_SET_PARAMETER_BULK in our GPL header on omamobile rather than from vendor library names, say the reply's counter byte isn't compared and why, and drop the link to a kernel commit that isn't on linux-omarchy-mobile.
The FT AEC reference changes nothing for the built-in microphone, the modem mixer's ASP stays off during our calls, stock applies wideband tuning to narrowband voice, and stock runs a speaker compensation tuner we don't.
With the tuning confirmed and the DSP in its telephony mode, the echo still came back after a few seconds, and AoC Modem Mixer ASP Mode (block 19, where the tuning goes) read ASP_OFF during the calls. With tegu_aec on, the route now sets it to ASP_ON before the upload and restores it after the call. Not yet tried on a call.
With the tuning sent before the streams, the far end heard no echo for 2-3 s and then the echo again, with ASP on and at stock's microphone gain alike (2026-11-03). The AoC starts its voice path at the playback start ("Start voice call" in the kernel log, after our upload), which may set it up afresh. The tuning now goes up after both streams have started, while the call sink is still muted; a failed upload closes the streams and keeps the sink muted as before. Not yet tried on a call.
…12, and the watcher that read them The results as derekross#378 had them, brought here with the acknowledgement fix so the two land together: the echo after the first seconds in every call, and what the AoC's own log showed on 11-12, that those first seconds are a silent uplink and not cancellation. bringup/call-audio-watch.py is the read-only watcher and call-levels.py reads its output.
|
Thank you. Taking your points in order; the branch is updated.
Your question on the six files: I've seen all six on one factory build only, the palmtop's own (BP3A.250905.014). That is no basis for a stop, so a missing one is a warning now: the set is left out whole, since the loader needs a route's three files together, the warning names the file, and everything else is extracted and checksummed as before. On #626: yes. I'll read it, and its author and I can compare the two on the palmtop, levels and switch included. One thing I can say already from this morning: with the tuning on, the echo comes back louder than with it off (a talking clock as the far end, the uplink's samples in the AoC's log: about 130 against a downlink of 2,000 with the tuning off, 4,000 to 8,000 with it on), so until the cancellation works an earpiece route is the part of this that helps a caller, and a live switch matters more than I gave it. It stays a draft. What I'm trying next for the echo is in "Next for the echo"; a call to a talking clock shows the echo in the AoC's log with nobody listening, so the tries don't need a far end.
|
…th the speakers at 0 and with the reference recorded Four more results from 2026-11-12. Without the tuning the uplink stays low through the clock's tone; with it the uplink is quiet most of the time and breaks through in bursts of several thousand; with the speakers at 0 the bursts are gone, so the echo is acoustic; and the processor's reference capture carries the downlink. The next steps are brought up to date.
|
Thank you. This was a lot of work in a day, and the result that matters most is a negative one found honestly: the two seconds without echo that five calls agreed on were a silent uplink after the answer, not cancellation. You found that with a watcher that only reads, checked it with a voice beside the palmtop, and then found a way to see the echo in the AoC's own log with nobody listening. Owning "my misreading, five times over" in the plan is exactly how I'd want it written. On the points from last time:
Thank you for the plain answer on the six files (one factory build), and for turning the stop into a warning that leaves the set out whole. It stays a draft, as you say. What I'd like before it leaves draft:
One thing is mine to answer and I haven't: on 11-05 I said I'd come back to you on the protocol description. The section still names the stock HAL's internal functions in two places and describes what one of them does. I'll tell you in a separate comment what can stay and what should go; please don't spend time on it until then. The version is mine to set at merge. |
tests/all keeps main's install-guided lines with the call tuning test after them; tools/extract-vendor-firmware keeps the loop that leaves an absent optional set out and sorts with LC_ALL=C, as main now does.
…ning stands, and the route key checked only on the 9a The plan's status and the shipped comment said the trial was still to run. It has: the AoC takes the tuning and the echo isn't cancelled; with it the echo comes through in louder bursts. Both say so now, and that tegu_aec stays off. The plan no longer says a missing tuning file fails extraction (the set is optional and left out whole), and it has the rules for the tests that are left: an AoC capture new to a call is a first probe, one per boot, attended, with the kernel log streamed first; only uplink tuning files are edited, and the owner's file goes back and is checked. VoiceAudio checked tegu_route on every phone, so a typo in that key would have left a Pixel 7 Pro's call without audio. It is checked on the 9a's route only. The two bring-up scripts get their SPDX lines.
|
Thank you. Two commits: the merge of main (58adf61) and e01f78e.
Your questions on the two captures, from the record:
Still open, and it stays a draft:
|
# Conflicts: # packages/omamobile-modem-sit/version # tests/all
|
Thank you. Everything I asked for on the 13th that was yours is in: the plan and the shipped comment now say plainly that the tuning doesn't cancel the echo and that Where this stands now:
And one thing is still mine: the comment on the protocol description, which I've now promised twice. It comes next; please don't spend time on that section until then. |
0ce0e5c to
821d2b6
Compare
omamobile-modem-sit is 0.3.46 here, after main's 0.3.44 and derekross#837's 0.3.45; the audio defaults keep this branch's tegu_route and tegu_aec beside main's control follow_screen.
On today's main; #371 and #372, which it was stacked on, have merged. This PR's own change is one commit (
add opt-in tegu call tuning and earpiece audio); the branch below it holds the earlier forms of what merged, and a merge of main makes the diff that commit alone.The palmtop's calls work on the speakerphone today, but without echo cancellation the far end hears itself, and there's no way to hold it to the ear. This adds both, off by default.
With
[audio] tegu_aec = yesin/etc/omamobile/modem-sit.conf, sitd uploads the owner's own AoC voice tuning (the uplink AEC on/off and downlink protobufs from the phone's factory image) through the AoC's output tuning services before the call's audio starts, waits for each acknowledgement, and turns AEC off again on cleanup. An unknown outcome leaves a marker in/run/omamobile-sit/and later attempts refuse until a reboot.tools/extract-vendor-firmware -d tegunow copies the six files (speaker and handset, unchanged, checksummed), and extraction fails if any is missing.tegu_route = earpieceplays through the top speaker at -30 dB with the bottom one at its minimum gain, and needstegu_aec = yes. Speaker stays the default.This head by itself has not run on the palmtop. Calls with
tegu_aec = yesran on 2026-11-02 on the heads of #378, which sits on this one: the uploads were confirmed by the AoC and the far end still heard the echo. They are recorded in #378's copy ofdevice/tegu/CALL-AUDIO-TUNING.md("Results on the palmtop, 2026-11-02"); this branch's copy predates them and still says no trial. The earpiece route hasn't been tried. The plan has the attended runbook (speaker first, the earpiece away from the ear first).Tests:
tests/packages omamobile-modem-sit,tests/extract-tegu-call-tuning,tools/pr-check.