notes: Bug #5 RX-degradation campaign — Phase 0 plan #13

2026-05-07T21:23:32Z

claude-noether commented

2026-05-07 21:23:32 +00:00

New campaign, independent of Patch C v3 (which closed via PR #5). Post-13-min RX collapse is firmware-side, not relay/race.

TL;DR

Research question locked, rig armed at fresh boot (23:18:58 CEST, T0=1778188738), 35-min capture in flight with 5 in-window stress probes at T+5/10/15/20/25 min.

Reasoning

Mechanism candidates §6 enumerated; Phase 4 plan branches off whichever Phase 3 data points to. The likely top-3:

Firmware resource exhaustion (scan failure -22 = EINVAL suggests refused request, not lock-up)
NM scan-fail recovery competing with data
AP-side rate limiting after sustained traffic

Status while review happens

Long capture is RUNNING. Probes scheduled. Post-capture analysis populates Phase 3. Result snapshot will be added as a follow-up note (bug5-rx-degradation-phase3-...) once the rig finishes (~T+30 min from now).

What's NOT in this PR

No code change
No Phase 4 plan yet (gated on Phase 3 data)
No mitigation suggestions yet — strictly observation per CLAUDE.md anti-theatre rule

New campaign, independent of Patch C v3 (which closed via PR #5). Post-13-min RX collapse is firmware-side, not relay/race. ## TL;DR Research question locked, rig armed at fresh boot (23:18:58 CEST, T0=1778188738), 35-min capture in flight with 5 in-window stress probes at T+5/10/15/20/25 min. ## Reasoning Mechanism candidates §6 enumerated; Phase 4 plan branches off whichever Phase 3 data points to. The likely top-3: 1. Firmware resource exhaustion (scan failure -22 = EINVAL suggests refused request, not lock-up) 2. NM scan-fail recovery competing with data 3. AP-side rate limiting after sustained traffic ## Status while review happens Long capture is RUNNING. Probes scheduled. Post-capture analysis populates Phase 3. Result snapshot will be added as a follow-up note (`bug5-rx-degradation-phase3-...`) once the rig finishes (~T+30 min from now). ## What's NOT in this PR - No code change - No Phase 4 plan yet (gated on Phase 3 data) - No mitigation suggestions yet — strictly observation per CLAUDE.md anti-theatre rule

claude-noether added 1 commit 2026-05-07 21:23:33 +00:00

notes: Bug #5 RX-degradation campaign — Phase 0 plan + research question 6bae531917

After Patch C v3 closed (PR #5 merged, Phase 7 N=3 verified at +73%
throughput vs Patch B baseline), the post-13-min RX-degradation
pattern remains.  Reproduces on Patch B, F, and v3 alike — independent
of the relay/race issues v3 addressed.  Side-effect that was masked
by the throughput floor while v2's race was the dominant variable.

Research question (locked):

  Why does the bes2600 RX path collapse from ~2 MB/s sustained @
  fresh-chip uptime to ~180 B/s @ ~28-min uptime, with periodic
  wsm_generic_confirm failed for request 0x0007 + ieee80211 phy0:
  [SCAN] Scan failed (-22) every 300 s in the intervening window?

Phase 0 protocol:

  - long-capture rig armed on ohm at uptime 0 (fresh boot 23:13 CEST)
  - ftrace events: workqueue, mac80211, cfg80211, mmc, sdhci, power
  - iw event (cfg80211 reason codes), dmesg follow, per-30s netdev
    counter snap, 5 stress probes at T+5/10/15/20/25 min

Phase 0 will:

  - re-anchor the predecessor data via the long capture (in-session
    N=1; re-run if anomalous)
  - characterize state transitions (first scan-fail, first throughput
    drop) via cfg80211/mac80211 ftrace + iw event correlation
  - feed Phase 1 metric formulation

Mechanism candidates (Phase 4 will discriminate):

  1. Firmware-side resource exhaustion (per-scan accumulator)
  2. NetworkManager scan-fail recovery loop competing with data
  3. AP-side rate limiting / fairness probation
  4. PSM state machine deadlock (c7 latch stale)
  5. SDIO bus retune interaction
  6. Power-management busy-event accumulator leak

Out of scope: Patch C2/D/E, higher-rate ramp, reproducing on different
APs.  Independent campaign from Patch C closure.

marfrit merged commit fc88ff41c3 into main

2026-05-07 21:52:14 +00:00

marfrit referenced this issue from a commit

2026-05-07 21:52:15 +00:00

notes: Bug #5 RX-degradation campaign — Phase 0 plan (#13)

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: marfrit/besser#13