Repository navigation
device/raven, DEVICES: Bluetooth audio and clean reboots, suspend to testing - #395
Conversation
…testing Bluetooth (BLUETOOTH-PLAN B4): a speaker paired from the phone's own menu and played over A2DP, crackling only while Wi-Fi carried traffic at the same time; the chip came up silent after one of seven clean reboots (derekross#77), the one that followed heavy Wi-Fi and Bluetooth use, and not on the six reboots tried after it with Bluetooth idle or after audio from a local file. It stays partial, with that written down. Suspend: three Power-woken s2idle cycles on today's kernels, without and with linux-omarchy-mobile#188 (UFS clean at spm_lvl 5), and auto-sleep on the daily kernel; raven moves from no to testing, as tegu is, with the modem's sleep block noted. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The first version blamed the busy Wi-Fi and Bluetooth before the one failing shutdown. The same silent chip came after reboots with no Bluetooth audio at all (B1, B3, the 10-27 push), at about one boot in five, so one failure in seven clean reboots says nothing about audio. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Thanks for writing up the day. The B4 reboot table is the clearest record of the silent chip so far, and separating "after audio" from "idle" settles that audio isn't the trigger. Two edits and then this is ready:
Two small ones: B4's "a forced restart recovered it" sits next to earlier runs where a forced restart didn't, so "this time a forced restart recovered it" is safer. And the kept "Earlier status: Bluetooth works, at every boot" could carry "(superseded the same day, B3)". This and #400/#401 all change raven's DEVICES row, so whichever merges second needs that one line redone; I'll handle it at merge if you'd rather not chase it. |
…spend cell's kernel Review on derekross#395: B4's "Next" credited the 100 ms hold to stock's driver, but Google's nitrous driver holds BT_REG_ON low 30 ms, as hci_bcm does; 100 ms is the stock HAL's BigHammerBtRegOnDelay. The "Next" becomes the run's result (0 silent in 15 clean reboots on gcbf4d42277a6), with a link to the test branch. The forced restart that recovered the chip is "this time", since earlier ones didn't, and the kept 10-27 status is marked as superseded the same day (B3). The Suspend cell in DEVICES now says its 3 of 3 ran on an unpublished kernel, and that derekross#188 is in review. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Thanks, both corrections are right. In e213d3b:
Both small ones are in too: "this time a forced restart recovered it", and the 10-27 status is marked "superseded the same day, B3". Thanks for offering to redo the DEVICES line at merge; that would be easiest. |
…ms kernel after 14 h up The 100 ms BT_REG_ON hold came up on all 15 short clean reboots on 2026-11-03, but the next clean reboot, after 14 hours up overnight with Bluetooth in use, left the chip silent on 2026-11-04, and the push's reboot-to after it carried that into the 30 ms kernel. B4 now keeps the tally Derek asked for on derekross#77, boot by boot from the phone's journal. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…th this PR's speakers cell)
|
Thank you, all four edits are in. The new tally is exactly the record I hoped for: boot by boot, with the kernel, the kind of restart and the uptime. Writing down the 14-hour failure the same morning, rather than keeping only the 15 of 15, is what makes the plan trustworthy. One optional edit: the paragraph under the 30/100 table still calls the 15 of 15 "strong evidence", right above a tally that's now 1 in 16. At the old rate of about one in five, one or fewer silent boots in 16 happens about one time in seven, so I'd date that sentence "(on 11-03)" or swap in the new figure. The uptime column you added also shows that all 15 good boots came after about two minutes up, while both failures came after much longer, so the next runs might hold uptime steady between the two kernels. Either can go in a later docs commit. Merged, ahead of #401; I've resolved the raven DEVICES row for #401 by hand (the two PRs touch different cells), as you suggested. |
…testing (derekross#395) Bluetooth (BLUETOOTH-PLAN B4): a speaker paired from the phone's menu and played over A2DP, crackling only while Wi-Fi carried traffic. The chip still comes up silent after some clean reboots (derekross#77), at about the old rate of one boot in five, so B4 keeps a boot-by-boot tally: the 100 ms BT_REG_ON test kernel came up on 15 short clean reboots, then stayed silent after one following 14 hours up. Google's driver holds REG_ON low 30 ms; the 100 ms is the stock HAL's. Bluetooth stays partial. Suspend: three Power-woken s2idle cycles on today's kernels, without and with linux-omarchy-mobile#188 (UFS clean at spm_lvl 5), and auto-sleep on the daily kernel; raven moves from no to testing, on an unpublished kernel, with the modem's sleep block noted.
What and why
Today's raven results, written into the plans and the matrix. Bluetooth now pairs from the phone's own menu and plays over A2DP; it crackles only while Wi-Fi is busy at the same time, and the chip came up silent after one of seven clean reboots (#77), so it stays partial with that written down (BLUETOOTH-PLAN B4). Suspend moves from no to testing, as tegu is: three Power-woken s2idle cycles each on the published kernel without and with linux-omarchy-mobile#188, which makes UFS resume clean, plus auto-sleep on the daily kernel (SUSPEND-PLAN, 2026-11-03).
Device and build
uname -r): 7.3.0-rc6-omamobile-push1-ge1e6adb69bbf (daily); the suspend runs on gb0a658e17f21 and g1094d975a193 (omamobile+ tools/phone-mirror: the panel size from the phone #116, without and with First boot without a clock skips the new release key (no RTC, no network) #188)Checks (no phone needed)
tools/pr-check origin/mainpassestests/allpassestests/packagespasses (or the packages I touched: none)versionbumped, and the new version is in the commit subjectCo-Authored-By:orAssisted-by:trailer, never aSigned-off-by:On the phone
userdata, weakens the storage guard, or reads registers from userspace (AGENTS.md §1)🤖 Generated with Claude Code