krd/linux
1
0
Fork 0
Code Activity

release 6.15.2 (preliminary)

Browse Source
This commit is contained in:
Konstantin Demin
2025-06-18 12:24:58 +03:00
parent 4d2691343a
commit 43dc655d2e
242 changed files with 7729 additions and 32303 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
debian/patches/misc-openwrt/0002-mac80211-avoid-crashing-missing-band.patch vendored Normal file
View File

@@ -0,0 +1,34 @@
From: David Bauer <mail@david-bauer.net>
Date: Thu, 30 Nov 2023 07:32:52 +0100
Subject: [PATCH] mac80211: avoid crashing on invalid band info
Frequent crashes have been observed on MT7916 based platforms. While the
root of these crashes are currently unknown, they happen when decoding
rate information of connected STAs in AP mode. The rate-information is
associated with a band which is not available on the PHY.
Check for this condition in order to avoid crashing the whole system.
This patch should be removed once the roout cause has been found and
fixed.
Link: https://github.com/freifunk-gluon/gluon/issues/2980
Signed-off-by: David Bauer <mail@david-bauer.net>
---
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -2474,6 +2474,13 @@ static void sta_stats_decode_rate(struct
sband = local->hw.wiphy->bands[band];
+ if (!sband) {
+ wiphy_warn(local->hw.wiphy,
+ "Invalid band %d\n",
+ band);
+ break;
+ }
+
if (WARN_ON_ONCE(!sband->bitrates))
break;

38
debian/patches/misc-openwrt/0003-mac80211-sta-randomize-BA-session-dialog-token-alloc.patch vendored Normal file
View File

@@ -0,0 +1,38 @@
From b478e06a16a8baa00c5ecc87c1d636981f2206d5 Mon Sep 17 00:00:00 2001
From: Johannes Berg <johannes.berg@intel.com>
Date: Tue, 29 Oct 2019 10:25:25 +0100
Subject: [PATCH] mac80211: sta: randomize BA session dialog token allocator
We currently always start the dialog token generator at zero,
so the first dialog token we use is always 1. This would be
OK if we had a perfect guarantee that we always do a proper
deauth/re-auth handshake, but in IBSS mode this doesn't always
happen properly.
To make problems with block ack (aggregation) sessions getting
stuck less likely, randomize the dialog token so if we start a
new session but the peer still has old state for us, it can
better detect this.
This is really just a workaround to make things a bit more
robust than they are now - a better fix would be to do a full
authentication handshake in IBSS mode upon having discovered a
new station, and on the receiver resetting the state (removing
and re-adding the station) on receiving the authentication
packet.
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
---
net/mac80211/sta_info.c | 1 +
1 file changed, 1 insertion(+)
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -583,6 +583,7 @@ __sta_info_alloc(struct ieee80211_sub_if
spin_lock_init(&sta->ps_lock);
INIT_WORK(&sta->drv_deliver_wk, sta_deliver_ps_frames);
wiphy_work_init(&sta->ampdu_mlme.work, ieee80211_ba_session_work);
+ sta->ampdu_mlme.dialog_token_allocator = get_random_u32_below(U8_MAX);
#ifdef CONFIG_MAC80211_MESH
if (ieee80211_vif_is_mesh(&sdata->vif)) {
sta->mesh = kzalloc(sizeof(*sta->mesh), gfp);

21
debian/patches/misc-openwrt/0004-mac80211-minstrel_ht-fix-MINSTREL_FRAC-macro.patch vendored Normal file
View File

@@ -0,0 +1,21 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Wed, 28 Apr 2021 21:03:13 +0200
Subject: [PATCH] mac80211: minstrel_ht: fix MINSTREL_FRAC macro
Add missing braces to avoid issues with e.g. using additions in the
div expression
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/net/mac80211/rc80211_minstrel_ht.h
+++ b/net/mac80211/rc80211_minstrel_ht.h
@@ -14,7 +14,7 @@
/* scaled fraction values */
#define MINSTREL_SCALE 12
-#define MINSTREL_FRAC(val, div) (((val) << MINSTREL_SCALE) / div)
+#define MINSTREL_FRAC(val, div) (((val) << MINSTREL_SCALE) / (div))
#define MINSTREL_TRUNC(val) ((val) >> MINSTREL_SCALE)
#define EWMA_LEVEL 96 /* ewma weighting factor [/EWMA_DIV] */

30
debian/patches/misc-openwrt/0005-mac80211-minstrel_ht-reduce-fluctuations-in-rate-pro.patch vendored Normal file
View File

@@ -0,0 +1,30 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Sat, 6 Feb 2021 16:08:01 +0100
Subject: [PATCH] mac80211: minstrel_ht: reduce fluctuations in rate
probability stats
In some scenarios when there is a lot of fluctuation in packet error rates,
rate switching can be amplified when the statistics get skewed by time slots
with very few tries.
Make the input data to the moving average more smooth by adding the
success/attempts count from the last stats window as well. This has the
advantage of smoothing the data without introducing any extra lag to sampling
rates.
This significantly improves rate stability on a strong test link subjected to
periodic noise bursts generated with a SDR
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/net/mac80211/rc80211_minstrel_ht.c
+++ b/net/mac80211/rc80211_minstrel_ht.c
@@ -769,7 +769,8 @@ minstrel_ht_calc_rate_stats(struct minst
unsigned int cur_prob;
if (unlikely(mrs->attempts > 0)) {
- cur_prob = MINSTREL_FRAC(mrs->success, mrs->attempts);
+ cur_prob = MINSTREL_FRAC(mrs->success + mrs->last_success,
+ mrs->attempts + mrs->last_attempts);
minstrel_filter_avg_add(&mrs->prob_avg,
&mrs->prob_avg_1, cur_prob);
mrs->att_hist += mrs->attempts;

151
debian/patches/misc-openwrt/0006-mac80211-minstrel_ht-rework-rate-downgrade-code-and-.patch vendored Normal file
View File

@@ -0,0 +1,151 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Sat, 6 Feb 2021 16:33:14 +0100
Subject: [PATCH] mac80211: minstrel_ht: rework rate downgrade code and
max_prob rate selection
The current fallback code for fast rate switching on potentially failing rates
is triggering too often if there is some strong noise on the channel. This can
lead to wild fluctuations in the rate selection.
Additionally, switching down to max_prob_rate can create a significant gap down
in throughput, especially when using only 2 spatial streams, because max_prob_rate
is limited to using fewer streams than the max_tp rates.
In order to improve throughput without reducing reliability too much, use the
rate downgrade code for the max_prob_rate only, and allow the non-downgraded
max_prob_rate to use as many spatial streams as the max_tp rates
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/net/mac80211/rc80211_minstrel_ht.c
+++ b/net/mac80211/rc80211_minstrel_ht.c
@@ -580,6 +580,14 @@ minstrel_ht_set_best_prob_rate(struct mi
int cur_tp_avg, cur_group, cur_idx;
int max_gpr_group, max_gpr_idx;
int max_gpr_tp_avg, max_gpr_prob;
+ int min_dur;
+
+ min_dur = max(minstrel_get_duration(mi->max_tp_rate[0]),
+ minstrel_get_duration(mi->max_tp_rate[1]));
+
+ /* make the rate at least 18% slower than max tp rates */
+ if (minstrel_get_duration(index) <= min_dur * 19 / 16)
+ return;
cur_group = MI_RATE_GROUP(index);
cur_idx = MI_RATE_IDX(index);
@@ -601,11 +609,6 @@ minstrel_ht_set_best_prob_rate(struct mi
!minstrel_ht_is_legacy_group(max_tp_group))
return;
- /* skip rates faster than max tp rate with lower prob */
- if (minstrel_get_duration(mi->max_tp_rate[0]) > minstrel_get_duration(index) &&
- mrs->prob_avg < max_tp_prob)
- return;
-
max_gpr_group = MI_RATE_GROUP(mg->max_group_prob_rate);
max_gpr_idx = MI_RATE_IDX(mg->max_group_prob_rate);
max_gpr_prob = mi->groups[max_gpr_group].rates[max_gpr_idx].prob_avg;
@@ -663,40 +666,6 @@ minstrel_ht_assign_best_tp_rates(struct
}
-/*
- * Try to increase robustness of max_prob rate by decrease number of
- * streams if possible.
- */
-static inline void
-minstrel_ht_prob_rate_reduce_streams(struct minstrel_ht_sta *mi)
-{
- struct minstrel_mcs_group_data *mg;
- int tmp_max_streams, group, tmp_idx, tmp_prob;
- int tmp_tp = 0;
-
- if (!mi->sta->deflink.ht_cap.ht_supported)
- return;
-
- group = MI_RATE_GROUP(mi->max_tp_rate[0]);
- tmp_max_streams = minstrel_mcs_groups[group].streams;
- for (group = 0; group < ARRAY_SIZE(minstrel_mcs_groups); group++) {
- mg = &mi->groups[group];
- if (!mi->supported[group] || group == MINSTREL_CCK_GROUP)
- continue;
-
- tmp_idx = MI_RATE_IDX(mg->max_group_prob_rate);
- tmp_prob = mi->groups[group].rates[tmp_idx].prob_avg;
-
- if (tmp_tp < minstrel_ht_get_tp_avg(mi, group, tmp_idx, tmp_prob) &&
- (minstrel_mcs_groups[group].streams < tmp_max_streams)) {
- mi->max_prob_rate = mg->max_group_prob_rate;
- tmp_tp = minstrel_ht_get_tp_avg(mi, group,
- tmp_idx,
- tmp_prob);
- }
- }
-}
-
static u16
__minstrel_ht_get_sample_rate(struct minstrel_ht_sta *mi,
enum minstrel_sample_type type)
@@ -1176,8 +1145,6 @@ minstrel_ht_update_stats(struct minstrel
mi->max_prob_rate = tmp_max_prob_rate;
- /* Try to increase robustness of max_prob_rate*/
- minstrel_ht_prob_rate_reduce_streams(mi);
minstrel_ht_refill_sample_rates(mi);
#ifdef CONFIG_MAC80211_DEBUGFS
@@ -1256,7 +1223,7 @@ minstrel_ht_ri_txstat_valid(struct minst
}
static void
-minstrel_downgrade_rate(struct minstrel_ht_sta *mi, u16 *idx, bool primary)
+minstrel_downgrade_prob_rate(struct minstrel_ht_sta *mi, u16 *idx)
{
int group, orig_group;
@@ -1271,11 +1238,7 @@ minstrel_downgrade_rate(struct minstrel_
minstrel_mcs_groups[orig_group].streams)
continue;
- if (primary)
- *idx = mi->groups[group].max_group_tp_rate[0];
- else
- *idx = mi->groups[group].max_group_tp_rate[1];
- break;
+ *idx = mi->groups[group].max_group_prob_rate;
}
}
@@ -1286,7 +1249,7 @@ minstrel_ht_tx_status(void *priv, struct
struct ieee80211_tx_info *info = st->info;
struct minstrel_ht_sta *mi = priv_sta;
struct ieee80211_tx_rate *ar = info->status.rates;
- struct minstrel_rate_stats *rate, *rate2;
+ struct minstrel_rate_stats *rate;
struct minstrel_priv *mp = priv;
u32 update_interval = mp->update_interval;
bool last, update = false;
@@ -1354,18 +1317,13 @@ minstrel_ht_tx_status(void *priv, struct
/*
* check for sudden death of spatial multiplexing,
* downgrade to a lower number of streams if necessary.
+ * only do this for the max_prob_rate to prevent spurious
+ * rate fluctuations when the link changes suddenly
*/
- rate = minstrel_get_ratestats(mi, mi->max_tp_rate[0]);
+ rate = minstrel_get_ratestats(mi, mi->max_prob_rate);
if (rate->attempts > 30 &&
rate->success < rate->attempts / 4) {
- minstrel_downgrade_rate(mi, &mi->max_tp_rate[0], true);
- update = true;
- }
-
- rate2 = minstrel_get_ratestats(mi, mi->max_tp_rate[1]);
- if (rate2->attempts > 30 &&
- rate2->success < rate2->attempts / 4) {
- minstrel_downgrade_rate(mi, &mi->max_tp_rate[1], false);
+ minstrel_downgrade_prob_rate(mi, &mi->max_prob_rate);
update = true;
}
}

53
debian/patches/misc-openwrt/0007-mac80211-increase-quantum-for-airtime-scheduler.patch vendored Normal file
View File

@@ -0,0 +1,53 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Sun, 26 Jun 2022 11:43:25 +0200
Subject: [PATCH] mac80211: increase quantum for airtime scheduler
Given the typical AQL budget and queue length, a quantum of 256 with the
default station weight often requires iterating over all queues frequently,
until one of them becomes eligible.
Improve performance by using 8 times station weight as scheduler quantum
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/net/mac80211/ieee80211_i.h
+++ b/net/mac80211/ieee80211_i.h
@@ -103,6 +103,8 @@ ieee80211_sta_keep_active(struct sta_inf
return time_before_eq(jiffies, sta->airtime[ac].last_active + HZ / 10);
}
+#define AIRTIME_QUANTUM_SHIFT 3
+
struct ieee80211_bss {
u32 device_ts_beacon, device_ts_presp;
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -4084,7 +4084,7 @@ struct ieee80211_txq *ieee80211_next_txq
if (deficit < 0)
sta->airtime[txqi->txq.ac].deficit +=
- sta->airtime_weight;
+ sta->airtime_weight << AIRTIME_QUANTUM_SHIFT;
if (deficit < 0 || !aql_check) {
list_move_tail(&txqi->schedule_order,
@@ -4227,7 +4227,8 @@ bool ieee80211_txq_may_transmit(struct i
}
sta = container_of(iter->txq.sta, struct sta_info, sta);
if (ieee80211_sta_deficit(sta, ac) < 0)
- sta->airtime[ac].deficit += sta->airtime_weight;
+ sta->airtime[ac].deficit += sta->airtime_weight <<
+ AIRTIME_QUANTUM_SHIFT;
list_move_tail(&iter->schedule_order, &local->active_txqs[ac]);
}
@@ -4235,7 +4236,7 @@ bool ieee80211_txq_may_transmit(struct i
if (sta->airtime[ac].deficit >= 0)
goto out;
- sta->airtime[ac].deficit += sta->airtime_weight;
+ sta->airtime[ac].deficit += sta->airtime_weight << AIRTIME_QUANTUM_SHIFT;
list_move_tail(&txqi->schedule_order, &local->active_txqs[ac]);
spin_unlock_bh(&local->active_txq_lock[ac]);

293
debian/patches/misc-openwrt/0008-mac80211-add-AQL-support-for-broadcast-packets.patch vendored Normal file
View File

@@ -0,0 +1,293 @@
From: Felix Fietkau <nbd@nbd.name>
Date: Fri, 9 Feb 2024 19:43:40 +0100
Subject: [PATCH] mac80211: add AQL support for broadcast packets
Excessive broadcast traffic with little competing unicast traffic can easily
flood hardware queues, leading to throughput issues. Additionally, filling
the hardware queues with too many packets breaks FQ for broadcast data.
Fix this by enabling AQL for broadcast packets.
Signed-off-by: Felix Fietkau <nbd@nbd.name>
---
--- a/include/net/cfg80211.h
+++ b/include/net/cfg80211.h
@@ -3464,6 +3464,7 @@ enum wiphy_params_flags {
/* The per TXQ device queue limit in airtime */
#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_L 5000
#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_H 12000
+#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_BC 50000
/* The per interface airtime threshold to switch to lower queue limit */
#define IEEE80211_AQL_THRESHOLD 24000
--- a/net/mac80211/debugfs.c
+++ b/net/mac80211/debugfs.c
@@ -212,11 +212,13 @@ static ssize_t aql_pending_read(struct f
"VI %u us\n"
"BE %u us\n"
"BK %u us\n"
+ "BC/MC %u us\n"
"total %u us\n",
atomic_read(&local->aql_ac_pending_airtime[IEEE80211_AC_VO]),
atomic_read(&local->aql_ac_pending_airtime[IEEE80211_AC_VI]),
atomic_read(&local->aql_ac_pending_airtime[IEEE80211_AC_BE]),
atomic_read(&local->aql_ac_pending_airtime[IEEE80211_AC_BK]),
+ atomic_read(&local->aql_bc_pending_airtime),
atomic_read(&local->aql_total_pending_airtime));
return simple_read_from_buffer(user_buf, count, ppos,
buf, len);
@@ -241,7 +243,8 @@ static ssize_t aql_txq_limit_read(struct
"VO %u %u\n"
"VI %u %u\n"
"BE %u %u\n"
- "BK %u %u\n",
+ "BK %u %u\n"
+ "BC/MC %u\n",
local->aql_txq_limit_low[IEEE80211_AC_VO],
local->aql_txq_limit_high[IEEE80211_AC_VO],
local->aql_txq_limit_low[IEEE80211_AC_VI],
@@ -249,7 +252,8 @@ static ssize_t aql_txq_limit_read(struct
local->aql_txq_limit_low[IEEE80211_AC_BE],
local->aql_txq_limit_high[IEEE80211_AC_BE],
local->aql_txq_limit_low[IEEE80211_AC_BK],
- local->aql_txq_limit_high[IEEE80211_AC_BK]);
+ local->aql_txq_limit_high[IEEE80211_AC_BK],
+ local->aql_txq_limit_bc);
return simple_read_from_buffer(user_buf, count, ppos,
buf, len);
}
@@ -275,6 +279,11 @@ static ssize_t aql_txq_limit_write(struc
else
buf[count] = '\0';
+ if (sscanf(buf, "mcast %u", &q_limit_low) == 1) {
+ local->aql_txq_limit_bc = q_limit_low;
+ return count;
+ }
+
if (sscanf(buf, "%u %u %u", &ac, &q_limit_low, &q_limit_high) != 3)
return -EINVAL;
--- a/net/mac80211/ieee80211_i.h
+++ b/net/mac80211/ieee80211_i.h
@@ -1368,10 +1368,12 @@ struct ieee80211_local {
spinlock_t handle_wake_tx_queue_lock;
u16 airtime_flags;
+ u32 aql_txq_limit_bc;
u32 aql_txq_limit_low[IEEE80211_NUM_ACS];
u32 aql_txq_limit_high[IEEE80211_NUM_ACS];
u32 aql_threshold;
atomic_t aql_total_pending_airtime;
+ atomic_t aql_bc_pending_airtime;
atomic_t aql_ac_pending_airtime[IEEE80211_NUM_ACS];
const struct ieee80211_ops *ops;
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c
@@ -959,6 +959,7 @@ struct ieee80211_hw *ieee80211_alloc_hw_
spin_lock_init(&local->rx_path_lock);
spin_lock_init(&local->queue_stop_reason_lock);
+ local->aql_txq_limit_bc = IEEE80211_DEFAULT_AQL_TXQ_LIMIT_BC;
for (i = 0; i < IEEE80211_NUM_ACS; i++) {
INIT_LIST_HEAD(&local->active_txqs[i]);
spin_lock_init(&local->active_txq_lock[i]);
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -2388,13 +2388,28 @@ EXPORT_SYMBOL(ieee80211_sta_recalc_aggre
void ieee80211_sta_update_pending_airtime(struct ieee80211_local *local,
struct sta_info *sta, u8 ac,
- u16 tx_airtime, bool tx_completed)
+ u16 tx_airtime, bool tx_completed,
+ bool mcast)
{
int tx_pending;
if (!wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL))
return;
+ if (mcast) {
+ if (!tx_completed) {
+ atomic_add(tx_airtime, &local->aql_bc_pending_airtime);
+ return;
+ }
+
+ tx_pending = atomic_sub_return(tx_airtime,
+ &local->aql_bc_pending_airtime);
+ if (tx_pending < 0)
+ atomic_cmpxchg(&local->aql_bc_pending_airtime,
+ tx_pending, 0);
+ return;
+ }
+
if (!tx_completed) {
if (sta)
atomic_add(tx_airtime,
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -2556,7 +2556,7 @@ static u16 ieee80211_store_ack_skb(struc
spin_lock_irqsave(&local->ack_status_lock, flags);
id = idr_alloc(&local->ack_status_frames, ack_skb,
- 1, 0x2000, GFP_ATOMIC);
+ 1, 0x1000, GFP_ATOMIC);
spin_unlock_irqrestore(&local->ack_status_lock, flags);
if (id >= 0) {
@@ -3985,20 +3985,20 @@ begin:
encap_out:
info->control.vif = vif;
- if (tx.sta &&
- wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL)) {
- bool ampdu = txq->ac != IEEE80211_AC_VO;
+ if (wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL)) {
+ bool ampdu = txq->sta && txq->ac != IEEE80211_AC_VO;
u32 airtime;
airtime = ieee80211_calc_expected_tx_airtime(hw, vif, txq->sta,
skb->len, ampdu);
- if (airtime) {
- airtime = ieee80211_info_set_tx_time_est(info, airtime);
- ieee80211_sta_update_pending_airtime(local, tx.sta,
- txq->ac,
- airtime,
- false);
- }
+ if (!airtime)
+ return skb;
+
+ airtime = ieee80211_info_set_tx_time_est(info, airtime);
+ info->tx_time_mc = !tx.sta;
+ ieee80211_sta_update_pending_airtime(local, tx.sta, txq->ac,
+ airtime, false,
+ info->tx_time_mc);
}
return skb;
@@ -4050,6 +4050,7 @@ struct ieee80211_txq *ieee80211_next_txq
struct ieee80211_txq *ret = NULL;
struct txq_info *txqi = NULL, *head = NULL;
bool found_eligible_txq = false;
+ bool aql_check;
spin_lock_bh(&local->active_txq_lock[ac]);
@@ -4073,26 +4074,26 @@ struct ieee80211_txq *ieee80211_next_txq
if (!head)
head = txqi;
+ aql_check = ieee80211_txq_airtime_check(hw, &txqi->txq);
+ if (aql_check)
+ found_eligible_txq = true;
+
if (txqi->txq.sta) {
struct sta_info *sta = container_of(txqi->txq.sta,
struct sta_info, sta);
- bool aql_check = ieee80211_txq_airtime_check(hw, &txqi->txq);
- s32 deficit = ieee80211_sta_deficit(sta, txqi->txq.ac);
-
- if (aql_check)
- found_eligible_txq = true;
-
- if (deficit < 0)
+ if (ieee80211_sta_deficit(sta, txqi->txq.ac) < 0) {
sta->airtime[txqi->txq.ac].deficit +=
sta->airtime_weight << AIRTIME_QUANTUM_SHIFT;
-
- if (deficit < 0 || !aql_check) {
- list_move_tail(&txqi->schedule_order,
- &local->active_txqs[txqi->txq.ac]);
- goto begin;
+ aql_check = false;
}
}
+ if (!aql_check) {
+ list_move_tail(&txqi->schedule_order,
+ &local->active_txqs[txqi->txq.ac]);
+ goto begin;
+ }
+
if (txqi->schedule_round == local->schedule_round[ac])
goto out;
@@ -4157,7 +4158,8 @@ bool ieee80211_txq_airtime_check(struct
return true;
if (!txq->sta)
- return true;
+ return atomic_read(&local->aql_bc_pending_airtime) <
+ local->aql_txq_limit_bc;
if (unlikely(txq->tid == IEEE80211_NUM_TIDS))
return true;
@@ -4206,15 +4208,15 @@ bool ieee80211_txq_may_transmit(struct i
spin_lock_bh(&local->active_txq_lock[ac]);
- if (!txqi->txq.sta)
- goto out;
-
if (list_empty(&txqi->schedule_order))
goto out;
if (!ieee80211_txq_schedule_airtime_check(local, ac))
goto out;
+ if (!txqi->txq.sta)
+ goto out;
+
list_for_each_entry_safe(iter, tmp, &local->active_txqs[ac],
schedule_order) {
if (iter == txqi)
--- a/include/net/mac80211.h
+++ b/include/net/mac80211.h
@@ -1238,8 +1238,8 @@ struct ieee80211_tx_info {
status_data_idr:1,
status_data:13,
hw_queue:4,
+ tx_time_mc:1,
tx_time_est:10;
- /* 1 free bit */
union {
struct {
--- a/net/mac80211/sta_info.h
+++ b/net/mac80211/sta_info.h
@@ -147,7 +147,8 @@ struct airtime_info {
void ieee80211_sta_update_pending_airtime(struct ieee80211_local *local,
struct sta_info *sta, u8 ac,
- u16 tx_airtime, bool tx_completed);
+ u16 tx_airtime, bool tx_completed,
+ bool mcast);
struct sta_info;
--- a/net/mac80211/status.c
+++ b/net/mac80211/status.c
@@ -734,7 +734,7 @@ static void ieee80211_report_used_skb(st
ieee80211_sta_update_pending_airtime(local, sta,
skb_get_queue_mapping(skb),
tx_time_est,
- true);
+ true, info->tx_time_mc);
rcu_read_unlock();
}
@@ -1143,10 +1143,11 @@ void ieee80211_tx_status_ext(struct ieee
/* Do this here to avoid the expensive lookup of the sta
* in ieee80211_report_used_skb().
*/
+ bool mcast = IEEE80211_SKB_CB(skb)->tx_time_mc;
ieee80211_sta_update_pending_airtime(local, sta,
skb_get_queue_mapping(skb),
tx_time_est,
- true);
+ true, mcast);
ieee80211_info_set_tx_time_est(IEEE80211_SKB_CB(skb), 0);
}

221
debian/patches/misc-openwrt/0009-mac80211-revert-dynamically-set-codel-parameters-per-station.patch vendored Normal file
View File

@@ -0,0 +1,221 @@
This reverts commit 484a54c2e597dbc4ace79c1687022282905afba0. The CoDel
parameter change essentially disables CoDel on slow stations, with some
questionable assumptions, as Dave pointed out in [0]. Quoting from
there:
But here are my pithy comments as to why this part of mac80211 is so
wrong...
static void sta_update_codel_params(struct sta_info *sta, u32 thr)
{
- if (thr && thr < STA_SLOW_THRESHOLD * sta->local->num_sta) {
1) sta->local->num_sta is the number of associated, rather than
active, stations. "Active" stations in the last 50ms or so, might have
been a better thing to use, but as most people have far more than that
associated, we end up with really lousy codel parameters, all the
time. Mistake numero uno!
2) The STA_SLOW_THRESHOLD was completely arbitrary in 2016.
- sta->cparams.target = MS2TIME(50);
This, by itself, was probably not too bad. 30ms might have been
better, at the time, when we were battling powersave etc, but 20ms was
enough, really, to cover most scenarios, even where we had low rate
2Ghz multicast to cope with. Even then, codel has a hard time finding
any sane drop rate at all, with a target this high.
- sta->cparams.interval = MS2TIME(300);
But this was horrible, a total mistake, that is leading to codel being
completely ineffective in almost any scenario on clients or APS.
100ms, even 80ms, here, would be vastly better than this insanity. I'm
seeing 5+seconds of delay accumulated in a bunch of otherwise happily
fq-ing APs....
100ms of observed jitter during a flow is enough. Certainly (in 2016)
there were interactions with powersave that I did not understand, and
still don't, but if you are transmitting in the first place, powersave
shouldn't be a problemmmm.....
- sta->cparams.ecn = false;
At the time we were pretty nervous about ecn, I'm kind of sanguine
about it now, and reliably indicating ecn seems better than turning it
off for any reason.
[...]
In production, on p2p wireless, I've had 8ms and 80ms for target and
interval for years now, and it works great.
I think Dave's arguments above are basically sound on the face of it,
and various experimentation with tighter CoDel parameters in the OpenWrt
community have show promising results[1]. So I don't think there's any
reason to keep this parameter fiddling; hence this revert.
[0] https://lore.kernel.org/linux-wireless/CAA93jw6NJ2cmLmMauz0xAgC2MGbBq6n0ZiZzAdkK0u4b+O2yXg@mail.gmail.com/
[1] https://forum.openwrt.org/t/reducing-multiplexing-latencies-still-further-in-wifi/133605/130
Suggested-By: Dave Taht <dave.taht@gmail.com>
In-memory-of: Dave Taht <dave.taht@gmail.com>
Signed-off-by: Toke Høiland-Jørgensen <toke@toke.dk>
--- a/include/net/mac80211.h
+++ b/include/net/mac80211.h
@@ -5347,22 +5347,6 @@ void ieee80211_get_tx_rates(struct ieee8
int max_rates);
/**
- * ieee80211_sta_set_expected_throughput - set the expected tpt for a station
- *
- * Call this function to notify mac80211 about a change in expected throughput
- * to a station. A driver for a device that does rate control in firmware can
- * call this function when the expected throughput estimate towards a station
- * changes. The information is used to tune the CoDel AQM applied to traffic
- * going towards that station (which can otherwise be too aggressive and cause
- * slow stations to starve).
- *
- * @pubsta: the station to set throughput for.
- * @thr: the current expected throughput in kbps.
- */
-void ieee80211_sta_set_expected_throughput(struct ieee80211_sta *pubsta,
- u32 thr);
-
-/**
* ieee80211_tx_rate_update - transmit rate update callback
*
* Drivers should call this functions with a non-NULL pub sta
--- a/net/mac80211/debugfs_sta.c
+++ b/net/mac80211/debugfs_sta.c
@@ -152,12 +152,6 @@ static ssize_t sta_aqm_read(struct file
p += scnprintf(p,
bufsz + buf - p,
- "target %uus interval %uus ecn %s\n",
- codel_time_to_us(sta->cparams.target),
- codel_time_to_us(sta->cparams.interval),
- sta->cparams.ecn ? "yes" : "no");
- p += scnprintf(p,
- bufsz + buf - p,
"tid ac backlog-bytes backlog-packets new-flows drops marks overlimit collisions tx-bytes tx-packets flags\n");
for (i = 0; i < ARRAY_SIZE(sta->sta.txq); i++) {
--- a/net/mac80211/rate.c
+++ b/net/mac80211/rate.c
@@ -990,8 +990,6 @@ int rate_control_set_rates(struct ieee80
if (sta->uploaded)
drv_sta_rate_tbl_update(hw_to_local(hw), sta->sdata, pubsta);
- ieee80211_sta_set_expected_throughput(pubsta, sta_get_expected_throughput(sta));
-
return 0;
}
EXPORT_SYMBOL(rate_control_set_rates);
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -18,7 +18,6 @@
#include <linux/timer.h>
#include <linux/rtnetlink.h>
-#include <net/codel.h>
#include <net/mac80211.h>
#include "ieee80211_i.h"
#include "driver-ops.h"
@@ -702,13 +701,6 @@ __sta_info_alloc(struct ieee80211_sub_if
}
}
- sta->cparams.ce_threshold = CODEL_DISABLED_THRESHOLD;
- sta->cparams.target = MS2TIME(20);
- sta->cparams.interval = MS2TIME(100);
- sta->cparams.ecn = true;
- sta->cparams.ce_threshold_selector = 0;
- sta->cparams.ce_threshold_mask = 0;
-
sta_dbg(sdata, "Allocated STA %pM\n", sta->sta.addr);
return sta;
@@ -2928,27 +2920,6 @@ unsigned long ieee80211_sta_last_active(
return sta->deflink.status_stats.last_ack;
}
-static void sta_update_codel_params(struct sta_info *sta, u32 thr)
-{
- if (thr && thr < STA_SLOW_THRESHOLD * sta->local->num_sta) {
- sta->cparams.target = MS2TIME(50);
- sta->cparams.interval = MS2TIME(300);
- sta->cparams.ecn = false;
- } else {
- sta->cparams.target = MS2TIME(20);
- sta->cparams.interval = MS2TIME(100);
- sta->cparams.ecn = true;
- }
-}
-
-void ieee80211_sta_set_expected_throughput(struct ieee80211_sta *pubsta,
- u32 thr)
-{
- struct sta_info *sta = container_of(pubsta, struct sta_info, sta);
-
- sta_update_codel_params(sta, thr);
-}
-
int ieee80211_sta_allocate_link(struct sta_info *sta, unsigned int link_id)
{
struct ieee80211_sub_if_data *sdata = sta->sdata;
--- a/net/mac80211/sta_info.h
+++ b/net/mac80211/sta_info.h
@@ -467,14 +467,6 @@ struct ieee80211_fragment_cache {
unsigned int next;
};
-/*
- * The bandwidth threshold below which the per-station CoDel parameters will be
- * scaled to be more lenient (to prevent starvation of slow stations). This
- * value will be scaled by the number of active stations when it is being
- * applied.
- */
-#define STA_SLOW_THRESHOLD 6000 /* 6 Mbps */
-
/**
* struct link_sta_info - Link STA information
* All link specific sta info are stored here for reference. This can be
@@ -627,7 +619,6 @@ struct link_sta_info {
* @sta: station information we share with the driver
* @sta_state: duplicates information about station state (for debug)
* @rcu_head: RCU head used for freeing this station struct
- * @cparams: CoDel parameters for this station.
* @reserved_tid: reserved TID (if any, otherwise IEEE80211_TID_UNRESERVED)
* @amsdu_mesh_control: track the mesh A-MSDU format used by the peer:
*
@@ -718,8 +709,6 @@ struct sta_info {
struct dentry *debugfs_dir;
#endif
- struct codel_params cparams;
-
u8 reserved_tid;
s8 amsdu_mesh_control;
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -1402,16 +1402,9 @@ static struct sk_buff *fq_tin_dequeue_fu
local = container_of(fq, struct ieee80211_local, fq);
txqi = container_of(tin, struct txq_info, tin);
+ cparams = &local->cparams;
cstats = &txqi->cstats;
- if (txqi->txq.sta) {
- struct sta_info *sta = container_of(txqi->txq.sta,
- struct sta_info, sta);
- cparams = &sta->cparams;
- } else {
- cparams = &local->cparams;
- }
-
if (flow == &tin->default_flow)
cvars = &txqi->def_cvars;
else

13
debian/patches/misc-openwrt/0010-mac80211-txq-tune.patch vendored Normal file
View File

@@ -0,0 +1,13 @@
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -1599,8 +1599,8 @@ int ieee80211_txq_setup_flows(struct iee
fq->memory_limit = 4 << 20; /* 4 Mbytes */
codel_params_init(&local->cparams);
- local->cparams.interval = MS2TIME(100);
- local->cparams.target = MS2TIME(20);
+ local->cparams.interval = MS2TIME(50);
+ local->cparams.target = MS2TIME(10);
local->cparams.ecn = true;
local->cvars = kvcalloc(fq->flows_cnt, sizeof(local->cvars[0]),

18
debian/patches/misc-openwrt/0011-cfg80211-aql-txq-limit.patch vendored Normal file
View File

@@ -0,0 +1,18 @@
--- a/include/net/cfg80211.h
+++ b/include/net/cfg80211.h
@@ -3462,12 +3462,12 @@ enum wiphy_params_flags {
#define IEEE80211_DEFAULT_AIRTIME_WEIGHT 256
/* The per TXQ device queue limit in airtime */
-#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_L 5000
-#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_H 12000
+#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_L 1500
+#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_H 5000
#define IEEE80211_DEFAULT_AQL_TXQ_LIMIT_BC 50000
/* The per interface airtime threshold to switch to lower queue limit */
-#define IEEE80211_AQL_THRESHOLD 24000
+#define IEEE80211_AQL_THRESHOLD 12000
/**
* struct cfg80211_pmksa - PMK Security Association

72
debian/patches/misc-openwrt/0101-sched-sch_cake-fix-bulk-flow-accounting-logic-for-host.patch vendored Normal file
View File

@@ -0,0 +1,72 @@
From 546ea84d07e3e324644025e2aae2d12ea4c5896e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Toke=20H=C3=B8iland-J=C3=B8rgensen?= <toke@redhat.com>
Date: Tue, 3 Sep 2024 18:08:45 +0200
Subject: [PATCH] sched: sch_cake: fix bulk flow accounting logic for host
fairness
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
In sch_cake, we keep track of the count of active bulk flows per host,
when running in dst/src host fairness mode, which is used as the
round-robin weight when iterating through flows. The count of active
bulk flows is updated whenever a flow changes state.
This has a peculiar interaction with the hash collision handling: when a
hash collision occurs (after the set-associative hashing), the state of
the hash bucket is simply updated to match the new packet that collided,
and if host fairness is enabled, that also means assigning new per-host
state to the flow. For this reason, the bulk flow counters of the
host(s) assigned to the flow are decremented, before new state is
assigned (and the counters, which may not belong to the same host
anymore, are incremented again).
Back when this code was introduced, the host fairness mode was always
enabled, so the decrement was unconditional. When the configuration
flags were introduced the *increment* was made conditional, but
the *decrement* was not. Which of course can lead to a spurious
decrement (and associated wrap-around to U16_MAX).
AFAICT, when host fairness is disabled, the decrement and wrap-around
happens as soon as a hash collision occurs (which is not that common in
itself, due to the set-associative hashing). However, in most cases this
is harmless, as the value is only used when host fairness mode is
enabled. So in order to trigger an array overflow, sch_cake has to first
be configured with host fairness disabled, and while running in this
mode, a hash collision has to occur to cause the overflow. Then, the
qdisc has to be reconfigured to enable host fairness, which leads to the
array out-of-bounds because the wrapped-around value is retained and
used as an array index. It seems that syzbot managed to trigger this,
which is quite impressive in its own right.
This patch fixes the issue by introducing the same conditional check on
decrement as is used on increment.
The original bug predates the upstreaming of cake, but the commit listed
in the Fixes tag touched that code, meaning that this patch won't apply
before that.
Fixes: 712639929912 ("sch_cake: Make the dual modes fairer")
Reported-by: syzbot+7fe7b81d602cc1e6b94d@syzkaller.appspotmail.com
Signed-off-by: Toke Høiland-Jørgensen <toke@redhat.com>
Link: https://patch.msgid.link/20240903160846.20909-1-toke@redhat.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
---
net/sched/sch_cake.c | 11 +++++++----
1 file changed, 7 insertions(+), 4 deletions(-)
--- a/net/sched/sch_cake.c
+++ b/net/sched/sch_cake.c
@@ -833,8 +833,10 @@ skip_hash:
allocate_dst = cake_ddst(flow_mode);
if (q->flows[outer_hash + k].set == CAKE_SET_BULK) {
- cake_dec_srchost_bulk_flow_count(q, &q->flows[outer_hash + k], flow_mode);
- cake_dec_dsthost_bulk_flow_count(q, &q->flows[outer_hash + k], flow_mode);
+ if (allocate_src)
+ q->hosts[q->flows[reduced_hash].srchost].srchost_bulk_flow_count--;
+ if (allocate_dst)
+ q->hosts[q->flows[reduced_hash].dsthost].dsthost_bulk_flow_count--;
}
found:
/* reserve queue for future packets in same flow */

11
debian/patches/misc-openwrt/0201-fq-adjust-memory-size.patch vendored Normal file
View File

@@ -0,0 +1,11 @@
--- a/include/net/fq_impl.h
+++ b/include/net/fq_impl.h
@@ -356,7 +356,7 @@ static int fq_init(struct fq *fq, int fl
fq->flows_cnt = max_t(u32, flows_cnt, 1);
fq->quantum = 300;
fq->limit = 8192;
- fq->memory_limit = 16 << 20; /* 16 MBytes */
+ fq->memory_limit = 32 << 20; /* 32 MBytes */
fq->flows = kvcalloc(fq->flows_cnt, sizeof(fq->flows[0]), GFP_KERNEL);
if (!fq->flows)