claude-noether
8b356aa11f
Replaces the 13 broken DKMS-path -danctnix mirrors from PR #17 + adds 9 new series-dirs for the c-stack patches that were never split (Patches A/B/C-v3/F/D/E/C2/G/H) + retires the cumulative-c5x-danctnix single-file interim from fleet/ohm.yaml. Mechanism: cd marfrit/bes2600-dkms-mobian git format-patch fe73571..cleanups --no-merges -o /tmp/cleanups/ git format-patch cleanups..bes2600/bh-c-fossil-cleanup --no-merges -o /tmp/h/ for each commit: route to series-dir, sed-rewrite a/bes2600/foo.c -> a/drivers/staging/bes2600/foo.c The 29 cleanups commits + 1 Patch H commit map to 25 series-dirs (a few series-dirs get multiple commits: lmac-recover gets c5.2 + c5.2.1 as 0001+0002; cw1200-fix-backports gets F3+F2+F1 as 0001-0003; factory-series gets request_firmware + STANDARD_FACTORY_EFUSE_FLAG as 0001+0002). fleet/ohm.yaml apply order matches cleanups commit chronology, which is what produced the working c5x interim. cumulative.patch from ka-promote ohm now has 32 resolved patches (29 cleanups + 1 Patch H + scan-filter-5ghz + xor-neon SCS + besser#18-fix), 276 079 bytes, b2sum 7418db5ddf8fe938b130bc9d0e9f7dc9060f3a13703cd50757835ac43140a13... Apply order in cleanups + bh-c-fossil-cleanup: 1 factory-series (c1 + factory-no-efuse-flag) 3 factory-thread-dev 4 pm-gate-on-handshake 5 remove-chardev-user-interface 6 enable-testmode 7 tx-sdio-dma-oob-danctnix (was 'staging-prep-series') 8 factory-drop-kernel-write-danctnix 9 drop-dpd-file-paths-danctnix 10 drop-orphan-file-io-danctnix 11 pm-timeout-silence-danctnix 12 scan-defer-on-reject-danctnix (c5.1) 13 scan-defer-backoff-tune-danctnix (c5.1.1) 14 lmac-recover-via-mmc-hw-reset-danctnix (c5.2 + c5.2.1) 16 pm-state-resync-danctnix (c6.1) 17 pm-wake-consume-state-danctnix (c6.2) 18 pm-detect-firmware-unsupported-danctnix (c7) 19 decrypt-storm-fast-recover-danctnix (Patch A) 20 connection-loss-fast-recover-danctnix (Patch B) 21 cw1200-fix-backports-danctnix (Patches F3 + F2 + F1) 24 sdio-rx-no-relay-danctnix (Patch C v3) 25 license-spdx-restore-attribution-danctnix (Patch G) 26 ba-lock-atomic-danctnix (Patch D) 27 ps-state-lock-skip-pm-disabled-danctnix (Patch E) 28 rx-list-batch-delivery-danctnix (Patch C2) 29 bh-c-fossil-cleanup-danctnix (Patch H) 30 scan-filter-5ghz-danctnix (besser#1) 31 arch/arm64/xor-neon-... (GCC 15 SCS) 32 queue-pending-record-lock-bh-danctnix (besser#18) Verification: pkgrel=6 build from this manifest in progress; if srcversion == 26B0003FE9F2B05DCE838C4 (pkgrel=5's), source-tree is byte-equivalent to the c5x interim + scan-filter + besser#18 stack that's currently running on ohm. Refs: #17 (the broken mirror), #28 (the interim PR that landed cumulative-c5x), #31 (ka-promote trailer normalisation followup).
54 lines
2.3 KiB
Diff
54 lines
2.3 KiB
Diff
From 88557185112dafbfba0053e9b5229125a97f5154 Mon Sep 17 00:00:00 2001
|
|
From: Markus Fritsche <fritsche.markus@gmail.com>
|
|
Date: Thu, 23 Apr 2026 20:35:17 +0200
|
|
Subject: [PATCH 11/29] bes2600: demote 'wait pm ind timeout' from bes_err to
|
|
bes_devel
|
|
|
|
bes2600_pwr_enter_lp_mode() logs 'wait pm ind timeout' at bes_err
|
|
level every time wait_for_completion_timeout() on the firmware's
|
|
PM-change indication returns 0. The preceding patch ('bes2600:
|
|
gate device LP-mode entry on successful per-VIF firmware
|
|
handshake') already handles this case correctly: the per-VIF
|
|
timeouts counter is incremented, the function returns
|
|
-ETIMEDOUT, and the device-side LP transition is skipped -- the
|
|
cascade into sdio_tx_work splats and [RX] Receive failure
|
|
messages is prevented.
|
|
|
|
The timeout itself is benign steady-state noise on the PineTab2
|
|
(BES2600WM). Firmware occasionally misses the 5 s PM-change
|
|
deadline when mac80211 flips power-save rapidly during
|
|
association or roaming; observed rate on a quiet, associated
|
|
ohm is roughly 3-10 events per 10 min of uptime, with no
|
|
user-visible effect. Keeping it at bes_err() level (== KERN_ERR,
|
|
priority 3) floods dmesg with what is already a handled
|
|
condition and makes real SDIO / PM errors harder to spot.
|
|
|
|
Demote to bes_devel() (== KERN_DEBUG gated on the driver's debug
|
|
flag). The gate in the caller is unchanged, so the downstream
|
|
suppression behaviour introduced by the earlier patch remains.
|
|
Real pathologies -- bes_err("set operation mode fail") on the
|
|
same path, and the timeouts != 0 / -ETIMEDOUT return consumed
|
|
by callers -- still surface at bes_err() / return-value level.
|
|
|
|
Signed-off-by: Markus Fritsche <fritsche.markus@gmail.com>
|
|
---
|
|
bes2600/bes_pwr.c | 2 +-
|
|
1 file changed, 1 insertion(+), 1 deletion(-)
|
|
|
|
diff --git a/drivers/staging/bes2600/bes_pwr.c b/drivers/staging/bes2600/bes_pwr.c
|
|
index f62ae22..474b6f1 100644
|
|
--- a/drivers/staging/bes2600/bes_pwr.c
|
|
+++ b/drivers/staging/bes2600/bes_pwr.c
|
|
@@ -538,7 +538,7 @@ static int bes2600_pwr_enter_lp_mode(struct bes2600_common *hw_priv)
|
|
atomic_set(&hw_priv->bes_power.pm_set_in_process, 0);
|
|
reinit_completion(&hw_priv->bes_power.pm_enter_cmpl);
|
|
if (!status) {
|
|
- bes_err("%s, wait pm ind timeout\n", __func__);
|
|
+ bes_devel("%s, wait pm ind timeout\n", __func__);
|
|
timeouts++;
|
|
}
|
|
} else {
|
|
--
|
|
2.54.0
|
|
|