# bes2600/join-confirm-reset-danctnix

Danctnix-flavor patch closing besser#25 (wsm_join_confirm failure cascade).

## What it does

When firmware returns status 1 on a JOIN command (`wsm_join_confirm ret 1`),
add a direct `wsm_reset(...)` call so the firmware returns to a clean IDLE
state, plus `queue_work(workqueue, &priv->unjoin_work)` for serialisation of
the next association attempt.

## Why it's a fork-divergence fix

`cw1200_join_work()` (cw1200 ancestor, `drivers/net/wireless/st/cw1200/sta.c:1339-1344`)
queues `unjoin_work` on join failure: `cw1200_do_unjoin()` calls `wsm_reset`
when `join_status == STA`.

bes2600's `bes2600_unjoin_work()` gates the same `wsm_reset` on
`join_status != PASSIVE`. After a failed JOIN, `join_status` stays PASSIVE
(only set to STA on success) — queuing `unjoin_work` alone is insufficient
on bes2600. The danctnix variant carries a direct `wsm_reset` in the
failure path *and* the queue_work serialisation.

## Observable effects (pkgrel=6 soak)

Beyond closing the cascade (besser#25 acceptance), this patch also
collapsed the periodic ~600 ms latency jitter on ohm:

| | pkgrel=5 | pkgrel=6 |
|---|---|---|
| max RTT | 612 ms | 13.9 ms |
| mdev | 103.5 ms | 1.55 ms |

The bgscan-driven roam-attempt to a 5 GHz BSSID followed by `wsm_join`
reject was briefly stalling TX every minute even when the cascade did
not fire.

## Upstream

- besser issue: marfrit/besser#25
- bes2600-dkms branch (Mobian flavor): bes2600/wsm-join-confirm-reset
  (PR #12 against `cleanups`)
- bes2600-dkms branch (danctnix flavor): bes2600/join-confirm-failure-reset
  (top commit `3d833f8`)
- shipped as patch 0022 in danctnix-besser-pkgbuild kernel/ (pkgrel=6,
  srcversion 0E16463FA8D85F4704DE93F)