2022-06-11 20:14:17

by Felix Fietkau

[permalink] [raw]
Subject: [PATCH v2 1/4] mac80211: fix overflow issues in airtime fairness code

The airtime weight calculation overflows with a default weight value of 256
whenever more than 8ms worth of airtime is reported.
Bigger weight values impose even smaller limits on maximum airtime values.
This can mess up airtime based calculations for drivers that don't report
per-PPDU airtime values, but batch up values instead.

Fix this by reordering multiplications/shifts and by reducing unnecessary
intermediate precision (which was lost in a later stage anyway).

The new shift value limits the maximum weight to 4096, which should be more
than enough. Any values bigger than that will be rejected

Signed-off-by: Felix Fietkau <[email protected]>
---
v2: reject weight values >4096
net/mac80211/cfg.c | 3 +++
net/mac80211/ieee80211_i.h | 41 +++++++++++---------------------------
net/mac80211/sta_info.c | 20 +++++++------------
3 files changed, 22 insertions(+), 42 deletions(-)

diff --git a/net/mac80211/cfg.c b/net/mac80211/cfg.c
index f7896f257e1b..fc0e1543bfc8 100644
--- a/net/mac80211/cfg.c
+++ b/net/mac80211/cfg.c
@@ -1619,6 +1619,9 @@ static int sta_apply_parameters(struct ieee80211_local *local,
mask = params->sta_flags_mask;
set = params->sta_flags_set;

+ if (params->airtime_weight > BIT(IEEE80211_RECIPROCAL_SHIFT_STA))
+ return -EINVAL;
+
if (ieee80211_vif_is_mesh(&sdata->vif)) {
/*
* In mesh mode, ASSOCIATED isn't part of the nl80211
diff --git a/net/mac80211/ieee80211_i.h b/net/mac80211/ieee80211_i.h
index 86ef0a46a68c..3bfe0d3e3439 100644
--- a/net/mac80211/ieee80211_i.h
+++ b/net/mac80211/ieee80211_i.h
@@ -1677,50 +1677,33 @@ static inline struct airtime_info *to_airtime_info(struct ieee80211_txq *txq)
/* To avoid divisions in the fast path, we keep pre-computed reciprocals for
* airtime weight calculations. There are two different weights to keep track
* of: The per-station weight and the sum of weights per phy.
- *
- * For the per-station weights (kept in airtime_info below), we use 32-bit
- * reciprocals with a devisor of 2^19. This lets us keep the multiplications and
- * divisions for the station weights as 32-bit operations at the cost of a bit
- * of rounding error for high weights; but the choice of divisor keeps rounding
- * errors <10% for weights <2^15, assuming no more than 8ms of airtime is
- * reported at a time.
- *
- * For the per-phy sum of weights the values can get higher, so we use 64-bit
- * operations for those with a 32-bit divisor, which should avoid any
- * significant rounding errors.
+ * The per-sta shift value supports weight values of 1-4096
*/
-#define IEEE80211_RECIPROCAL_DIVISOR_64 0x100000000ULL
-#define IEEE80211_RECIPROCAL_SHIFT_64 32
-#define IEEE80211_RECIPROCAL_DIVISOR_32 0x80000U
-#define IEEE80211_RECIPROCAL_SHIFT_32 19
+#define IEEE80211_RECIPROCAL_SHIFT_SUM 24
+#define IEEE80211_RECIPROCAL_SHIFT_STA 12
+#define IEEE80211_WEIGHT_SHIFT 8

-static inline void airtime_weight_set(struct airtime_info *air_info, u16 weight)
+static inline void airtime_weight_set(struct airtime_info *air_info, u32 weight)
{
if (air_info->weight == weight)
return;

air_info->weight = weight;
- if (weight) {
- air_info->weight_reciprocal =
- IEEE80211_RECIPROCAL_DIVISOR_32 / weight;
- } else {
- air_info->weight_reciprocal = 0;
- }
+ if (weight)
+ weight = BIT(IEEE80211_RECIPROCAL_SHIFT_STA) / weight;
+ air_info->weight_reciprocal = weight;
}

static inline void airtime_weight_sum_set(struct airtime_sched_info *air_sched,
- int weight_sum)
+ u32 weight_sum)
{
if (air_sched->weight_sum == weight_sum)
return;

air_sched->weight_sum = weight_sum;
- if (air_sched->weight_sum) {
- air_sched->weight_sum_reciprocal = IEEE80211_RECIPROCAL_DIVISOR_64;
- do_div(air_sched->weight_sum_reciprocal, air_sched->weight_sum);
- } else {
- air_sched->weight_sum_reciprocal = 0;
- }
+ if (weight_sum)
+ weight_sum = BIT(IEEE80211_RECIPROCAL_SHIFT_SUM) / weight_sum;
+ air_sched->weight_sum_reciprocal = weight_sum;
}

/* A problem when trying to enforce airtime fairness is that we want to divide
diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c
index e04a0905e941..2f1cf9d4e68d 100644
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -1906,9 +1906,9 @@ void ieee80211_register_airtime(struct ieee80211_txq *txq,
{
struct ieee80211_sub_if_data *sdata = vif_to_sdata(txq->vif);
struct ieee80211_local *local = sdata->local;
- u64 weight_sum, weight_sum_reciprocal;
struct airtime_sched_info *air_sched;
struct airtime_info *air_info;
+ u64 weight_sum_reciprocal;
u32 airtime = 0;

air_sched = &local->airtime[txq->ac];
@@ -1919,27 +1919,21 @@ void ieee80211_register_airtime(struct ieee80211_txq *txq,
if (local->airtime_flags & AIRTIME_USE_RX)
airtime += rx_airtime;

- /* Weights scale so the unit weight is 256 */
- airtime <<= 8;
-
spin_lock_bh(&air_sched->lock);

air_info->tx_airtime += tx_airtime;
air_info->rx_airtime += rx_airtime;

- if (air_sched->weight_sum) {
- weight_sum = air_sched->weight_sum;
+ if (air_sched->weight_sum)
weight_sum_reciprocal = air_sched->weight_sum_reciprocal;
- } else {
- weight_sum = air_info->weight;
+ else
weight_sum_reciprocal = air_info->weight_reciprocal;
- }

/* Round the calculation of global vt */
- air_sched->v_t += (u64)((airtime + (weight_sum >> 1)) *
- weight_sum_reciprocal) >> IEEE80211_RECIPROCAL_SHIFT_64;
- air_info->v_t += (u32)((airtime + (air_info->weight >> 1)) *
- air_info->weight_reciprocal) >> IEEE80211_RECIPROCAL_SHIFT_32;
+ air_sched->v_t += ((u64)airtime * weight_sum_reciprocal) >>
+ (IEEE80211_RECIPROCAL_SHIFT_SUM - IEEE80211_WEIGHT_SHIFT);
+ air_info->v_t += (airtime * air_info->weight_reciprocal) >>
+ (IEEE80211_RECIPROCAL_SHIFT_STA - IEEE80211_WEIGHT_SHIFT);
ieee80211_resort_txq(&local->hw, txq);

spin_unlock_bh(&air_sched->lock);
--
2.36.1


2022-06-11 20:49:34

by Felix Fietkau

[permalink] [raw]
Subject: [PATCH v2 3/4] mac80211: fix ieee80211_txq_may_transmit regression

After switching to the virtual time based airtime scheduler, there were reports
that ath10k with tx queueing in push-pull mode was experiencing significant
latency for some stations.
The reason for it is the fact that queues from which the ath10k firmware wants
to pull are getting starved by airtime fairness constraints.
Theoretically the same issue should have been there before the switch to virtual
time, however it seems that in the old round-robin implementation it was simply
looping until the requested txq was considered eligible, which led to it pretty
much ignoring fairness constraints anyway.

In order to fix the immediate regression, let's make bypassing airtime fairness
explicit for now.
Also update the documentation for ieee80211_txq_may_transmit, which was still
referring to implementation details of the old round-robin scheduler

Fixes: 2433647bc8d9 ("mac80211: Switch to a virtual time-based airtime scheduler")
Signed-off-by: Felix Fietkau <[email protected]>
---
include/net/mac80211.h | 21 +++++----------------
net/mac80211/tx.c | 17 +----------------
2 files changed, 6 insertions(+), 32 deletions(-)

diff --git a/include/net/mac80211.h b/include/net/mac80211.h
index d8c4d49bcc49..cb65abab8f36 100644
--- a/include/net/mac80211.h
+++ b/include/net/mac80211.h
@@ -6821,22 +6821,11 @@ void ieee80211_return_txq(struct ieee80211_hw *hw, struct ieee80211_txq *txq,
/**
* ieee80211_txq_may_transmit - check whether TXQ is allowed to transmit
*
- * This function is used to check whether given txq is allowed to transmit by
- * the airtime scheduler, and can be used by drivers to access the airtime
- * fairness accounting without going using the scheduling order enfored by
- * next_txq().
- *
- * Returns %true if the airtime scheduler thinks the TXQ should be allowed to
- * transmit, and %false if it should be throttled. This function can also have
- * the side effect of rotating the TXQ in the scheduler rotation, which will
- * eventually bring the deficit to positive and allow the station to transmit
- * again.
- *
- * The API ieee80211_txq_may_transmit() also ensures that TXQ list will be
- * aligned against driver's own round-robin scheduler list. i.e it rotates
- * the TXQ list till it makes the requested node becomes the first entry
- * in TXQ list. Thus both the TXQ list and driver's list are in sync. If this
- * function returns %true, the driver is expected to schedule packets
+ * Returns %true if there is remaining AQL budget for the tx queue and %false
+ * if it should be throttled. It will also mark the queue as active for the
+ * airtime scheduler.
+ *
+ * If this function returns %true, the driver is expected to schedule packets
* for transmission, and then return the TXQ through ieee80211_return_txq().
*
* @hw: pointer as obtained from ieee80211_alloc_hw()
diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index ecb3975680bc..f844e9ed4c14 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -4099,15 +4099,13 @@ EXPORT_SYMBOL(ieee80211_txq_airtime_check);
bool ieee80211_txq_may_transmit(struct ieee80211_hw *hw,
struct ieee80211_txq *txq)
{
- struct txq_info *first_txqi = NULL, *txqi = to_txq_info(txq);
struct ieee80211_local *local = hw_to_local(hw);
+ struct txq_info *txqi = to_txq_info(txq);
struct airtime_sched_info *air_sched;
struct airtime_info *air_info;
- struct rb_node *node = NULL;
bool ret = false;
u64 now;

-
if (!ieee80211_txq_airtime_check(hw, txq))
return false;

@@ -4119,19 +4117,6 @@ bool ieee80211_txq_may_transmit(struct ieee80211_hw *hw,

now = ktime_get_coarse_boottime_ns();

- /* Like in ieee80211_next_txq(), make sure the first station in the
- * scheduling order is eligible for transmission to avoid starvation.
- */
- node = rb_first_cached(&air_sched->active_txqs);
- if (node) {
- first_txqi = container_of(node, struct txq_info,
- schedule_order);
- air_info = to_airtime_info(&first_txqi->txq);
-
- if (air_sched->v_t < air_info->v_t)
- airtime_catchup_v_t(air_sched, air_info->v_t, now);
- }
-
air_info = to_airtime_info(&txqi->txq);
if (air_info->v_t <= air_sched->v_t) {
air_sched->last_schedule_activity = now;
--
2.36.1

2022-06-11 20:54:34

by Felix Fietkau

[permalink] [raw]
Subject: [PATCH v2 4/4] mac80211: rework the airtime fairness implementation

The current ATF implementation has a number of issues which have shown up
during testing. Since it does not take into account the AQL budget of
pending packets, the implementation might queue up large amounts of packets
for a single txq until airtime gets reported after tx completion.
The same then happens to the next txq afterwards. While the end result could
still be considered fair, the bursty behavior introduces a large amount of
latency.
The current code also tries to avoid frequent re-sorting of txq entries in
order to avoid having to re-balance the rbtree often.

In order to fix these issues, introduce skip lists as a data structure, which
offer similar lookup/insert/delete times as rbtree, but avoids the need for
rebalacing by being probabilistic.
Use this to keep tx entries sorted by virtual time + pending AQL budget and
re-sort after each ieee80211_return_txq call.

Since multiple txqs share a single air_time struct with a virtual time value,
switch the active_txqs list to queue up air_time structs instead of queues.
This helps avoid imbalance between shared txqs by servicing them round robin.

ieee80211_next_txq now only dequeues the first element of active_txqs. To
make that work for non-AQL or non-ATF drivers as well, add estimated tx
airtime directly to air_info virtual time if either AQL or ATF is not
supported.

Signed-off-by: Felix Fietkau <[email protected]>
---
v2:
- code cleanup
- rework ieee80211_txq_may_transmit
- fix airtime catchup

include/linux/skiplist.h | 250 +++++++++++++++++++++++++++++++++++++
net/mac80211/cfg.c | 5 -
net/mac80211/ieee80211_i.h | 27 +++-
net/mac80211/main.c | 2 +-
net/mac80211/sta_info.c | 3 +-
net/mac80211/sta_info.h | 6 +-
net/mac80211/tx.c | 226 ++++++++++++++++-----------------
7 files changed, 386 insertions(+), 133 deletions(-)
create mode 100644 include/linux/skiplist.h

diff --git a/include/linux/skiplist.h b/include/linux/skiplist.h
new file mode 100644
index 000000000000..2312ed87cfe9
--- /dev/null
+++ b/include/linux/skiplist.h
@@ -0,0 +1,250 @@
+/* SPDX-License-Identifier: GPL-2.0-or-later */
+/*
+ * A skip list is a probabilistic alternative to balanced trees. Unlike the
+ * red-black tree, it does not require rebalancing.
+ *
+ * This implementation uses only unidirectional next pointers and is optimized
+ * for use in a priority queue where elements are mostly deleted from the front
+ * of the queue.
+ *
+ * When storing up to 2^n elements in a n-level skiplist. lookup and deletion
+ * for the first element happens in O(1) time, other than that, insertion and
+ * deletion takes O(log n) time, assuming that the number of elements for an
+ * n-level list does not exceed 2^n.
+ *
+ * Usage:
+ * DECLARE_SKIPLIST_TYPE(foo, 5) will define the data types for a 5-level list:
+ * struct foo_list: the list data type
+ * struct foo_node: the node data for an element in the list
+ *
+ * DECLARE_SKIPLIST_IMPL(foo, foo_cmp_fn)
+ *
+ * Adds the skip list implementation. It depends on a provided function:
+ * int foo_cmp_fn(struct foo_list *list, struct foo_node *n1, struct foo_node *n2)
+ * This compares two elements given by their node pointers, returning values <0
+ * if n1 is less than n2, =0 and >0 for equal or bigger than respectively.
+ *
+ * This macro implements the following functions:
+ *
+ * void foo_list_init(struct foo_list *list)
+ * initializes the skip list
+ *
+ * void foo_node_init(struct foo_node *node)
+ * initializes a node. must be called before adding the node to the list
+ *
+ * struct foo_node *foo_node_next(struct foo_node *node)
+ * gets the node directly after the provided node, or NULL if it was the last
+ * element in the list.
+ *
+ * bool foo_is_queued(struct foo_node *node)
+ * returns true if the node is on a list
+ *
+ * struct foo_node *foo_dequeue(struct foo_list *list)
+ * deletes and returns the first element of the list (or returns NULL if empty)
+ *
+ * struct foo_node *foo_peek(struct foo_list *list)
+ * returns the first element of the list
+ *
+ * void foo_insert(struct foo_list *list, struct foo_node *node)
+ * inserts the node into the list. the node must be initialized and not on a
+ * list already.
+ *
+ * void foo_delete(struct foo_list *list, struct foo_node *node)
+ * deletes the node from the list, or does nothing if it's not on the list
+ */
+#ifndef __SKIPLIST_H
+#define __SKIPLIST_H
+
+#include <linux/bits.h>
+#include <linux/minmax.h>
+#include <linux/bug.h>
+#include <linux/prandom.h>
+
+#define SKIPLIST_POISON ((void *)1)
+
+#define DECLARE_SKIPLIST_TYPE(name, levels) \
+struct name##_node { \
+ struct name##_node *next[levels]; \
+}; \
+struct name##_list { \
+ struct name##_node head; \
+ unsigned int max_level; \
+ unsigned int count; \
+};
+
+#define DECLARE_SKIPLIST_IMPL(name, cmp_fn) \
+static inline void \
+name##_list_init(struct name##_list *list) \
+{ \
+ memset(list, 0, sizeof(*list)); \
+} \
+static inline void \
+name##_node_init(struct name##_node *node) \
+{ \
+ node->next[0] = SKIPLIST_POISON; \
+} \
+static inline struct name##_node * \
+name##_node_next(struct name##_node *node) \
+{ \
+ return node->next[0]; \
+} \
+static inline bool \
+name##_is_queued(struct name##_node *node) \
+{ \
+ return node->next[0] != SKIPLIST_POISON; \
+} \
+static inline int \
+__skiplist_##name##_cmp_impl(void *head, void *n1, void *n2) \
+{ \
+ return cmp_fn(head, n1, n2); \
+} \
+static inline void \
+__##name##_delete(struct name##_list *list) \
+{ \
+ list->count--; \
+ while (list->max_level && \
+ !list->head.next[list->max_level]) \
+ list->max_level--; \
+} \
+static inline struct name##_node * \
+name##_dequeue(struct name##_list *list) \
+{ \
+ struct name##_node *ret; \
+ unsigned int max_level = ARRAY_SIZE(list->head.next) - 1; \
+ ret = (void *)__skiplist_dequeue((void **)&list->head, \
+ max_level); \
+ if (!ret) \
+ return NULL; \
+ __##name##_delete(list); \
+ return ret; \
+} \
+static inline struct name##_node * \
+name##_peek(struct name##_list *list) \
+{ \
+ return list->head.next[0]; \
+} \
+static inline void \
+name##_insert(struct name##_list *list, struct name##_node *node) \
+{ \
+ int level = __skiplist_level(ARRAY_SIZE(list->head.next) - 1, \
+ list->count, prandom_u32()); \
+ level = min_t(int, level, list->max_level + 1); \
+ __skiplist_insert((void *)&list->head, (void *)node, level, \
+ __skiplist_##name##_cmp_impl); \
+ if (level > list->max_level) \
+ list->max_level = level; \
+ list->count++; \
+} \
+static inline void \
+name##_delete(struct name##_list *list, struct name##_node *node) \
+{ \
+ if (node->next[0] == SKIPLIST_POISON) \
+ return; \
+ __skiplist_delete((void *)&list->head, (void *)node, \
+ ARRAY_SIZE(list->head.next) - 1, \
+ __skiplist_##name##_cmp_impl); \
+ __##name##_delete(list); \
+}
+
+
+typedef int (*__skiplist_cmp_t)(void *head, void *n1, void *n2);
+
+#define __skiplist_cmp(cmp, head, cur, node) \
+ ({ \
+ int cmp_val = cmp(head, cur, node); \
+ if (!cmp_val) \
+ cmp_val = (unsigned long)(cur) - \
+ (unsigned long)(node); \
+ cmp_val; \
+ })
+
+static inline void *
+__skiplist_dequeue(void **list, int max_level)
+{
+ void **node = list[0];
+ unsigned int i;
+
+ if (!node)
+ return NULL;
+
+ list[0] = node[0];
+ for (i = 1; i <= max_level; i++) {
+ if (list[i] != node)
+ break;
+
+ list[i] = node[i];
+ }
+ node[0] = SKIPLIST_POISON;
+
+ return node;
+}
+
+static inline void
+__skiplist_insert(void **list, void **node, int level, __skiplist_cmp_t cmp)
+{
+ void **head = list;
+
+ if (WARN(node[0] != SKIPLIST_POISON, "Insert on already inserted or uninitialized node"))
+ return;
+ for (; level >= 0; level--) {
+ while (list[level] &&
+ __skiplist_cmp(cmp, head, list[level], node) < 0)
+ list = list[level];
+
+ node[level] = list[level];
+ list[level] = node;
+ }
+}
+
+
+static inline void
+__skiplist_delete(void **list, void **node, int max_level, __skiplist_cmp_t cmp)
+{
+ void *head = list;
+ int i;
+
+ for (i = max_level; i >= 0; i--) {
+ while (list[i] && list[i] != node &&
+ __skiplist_cmp(cmp, head, list[i], node) <= 0)
+ list = list[i];
+
+ if (list[i] != node)
+ continue;
+
+ list[i] = node[i];
+ }
+ node[0] = SKIPLIST_POISON;
+}
+
+static inline unsigned int
+__skiplist_level(unsigned int max_level, unsigned int count, unsigned int seed)
+{
+ unsigned int level = 0;
+
+ if (max_level >= 16 && !(seed & GENMASK(15, 0))) {
+ level += 16;
+ seed >>= 16;
+ }
+
+ if (max_level >= 8 && !(seed & GENMASK(7, 0))) {
+ level += 8;
+ seed >>= 8;
+ }
+
+ if (max_level >= 4 && !(seed & GENMASK(3, 0))) {
+ level += 4;
+ seed >>= 4;
+ }
+
+ if (!(seed & GENMASK(1, 0))) {
+ level += 2;
+ seed >>= 2;
+ }
+
+ if (!(seed & BIT(0)))
+ level++;
+
+ return min(level, max_level);
+}
+
+#endif
diff --git a/net/mac80211/cfg.c b/net/mac80211/cfg.c
index fc0e1543bfc8..3ab76f30159a 100644
--- a/net/mac80211/cfg.c
+++ b/net/mac80211/cfg.c
@@ -1580,7 +1580,6 @@ static void sta_apply_airtime_params(struct ieee80211_local *local,
for (ac = 0; ac < IEEE80211_NUM_ACS; ac++) {
struct airtime_sched_info *air_sched = &local->airtime[ac];
struct airtime_info *air_info = &sta->airtime[ac];
- struct txq_info *txqi;
u8 tid;

spin_lock_bh(&air_sched->lock);
@@ -1592,10 +1591,6 @@ static void sta_apply_airtime_params(struct ieee80211_local *local,

airtime_weight_set(air_info, params->airtime_weight);

- txqi = to_txq_info(sta->sta.txq[tid]);
- if (RB_EMPTY_NODE(&txqi->schedule_order))
- continue;
-
ieee80211_update_airtime_weight(local, air_sched,
0, true);
}
diff --git a/net/mac80211/ieee80211_i.h b/net/mac80211/ieee80211_i.h
index 3bfe0d3e3439..2bc34b14dd6e 100644
--- a/net/mac80211/ieee80211_i.h
+++ b/net/mac80211/ieee80211_i.h
@@ -25,7 +25,8 @@
#include <linux/leds.h>
#include <linux/idr.h>
#include <linux/rhashtable.h>
-#include <linux/rbtree.h>
+#include <linux/prandom.h>
+#include <linux/skiplist.h>
#include <net/ieee80211_radiotap.h>
#include <net/cfg80211.h>
#include <net/mac80211.h>
@@ -855,6 +856,7 @@ enum txq_info_flags {
IEEE80211_TXQ_AMPDU,
IEEE80211_TXQ_NO_AMSDU,
IEEE80211_TXQ_STOP_NETIF_TX,
+ IEEE80211_TXQ_FORCE_ACTIVE,
};

/**
@@ -871,7 +873,6 @@ struct txq_info {
struct fq_tin tin;
struct codel_vars def_cvars;
struct codel_stats cstats;
- struct rb_node schedule_order;

struct sk_buff_head frags;
unsigned long flags;
@@ -1191,8 +1192,7 @@ enum mac80211_scan_state {
*
* @lock: spinlock that protects all the fields in this struct
* @active_txqs: rbtree of currently backlogged queues, sorted by virtual time
- * @schedule_pos: the current position maintained while a driver walks the tree
- * with ieee80211_next_txq()
+ * @schedule_pos: last used airtime_info node while a driver walks the tree
* @active_list: list of struct airtime_info structs that were active within
* the last AIRTIME_ACTIVE_DURATION (100 ms), used to compute
* weight_sum
@@ -1213,8 +1213,8 @@ enum mac80211_scan_state {
*/
struct airtime_sched_info {
spinlock_t lock;
- struct rb_root_cached active_txqs;
- struct rb_node *schedule_pos;
+ struct airtime_sched_list active_txqs;
+ struct airtime_sched_node *schedule_pos;
struct list_head active_list;
u64 last_weight_update;
u64 last_schedule_activity;
@@ -1674,6 +1674,20 @@ static inline struct airtime_info *to_airtime_info(struct ieee80211_txq *txq)
return &sdata->airtime[txq->ac];
}

+static inline int
+airtime_sched_cmp(struct airtime_sched_list *list,
+ struct airtime_sched_node *n1, struct airtime_sched_node *n2)
+{
+ struct airtime_info *a1, *a2;
+
+ a1 = container_of(n1, struct airtime_info, schedule_order);
+ a2 = container_of(n2, struct airtime_info, schedule_order);
+
+ return a1->v_t_cur - a2->v_t_cur;
+}
+
+DECLARE_SKIPLIST_IMPL(airtime_sched, airtime_sched_cmp);
+
/* To avoid divisions in the fast path, we keep pre-computed reciprocals for
* airtime weight calculations. There are two different weights to keep track
* of: The per-station weight and the sum of weights per phy.
@@ -1760,6 +1774,7 @@ static inline void init_airtime_info(struct airtime_info *air_info,
air_info->aql_limit_high = air_sched->aql_txq_limit_high;
airtime_weight_set(air_info, IEEE80211_DEFAULT_AIRTIME_WEIGHT);
INIT_LIST_HEAD(&air_info->list);
+ airtime_sched_node_init(&air_info->schedule_order);
}

static inline int ieee80211_bssid_match(const u8 *raddr, const u8 *addr)
diff --git a/net/mac80211/main.c b/net/mac80211/main.c
index 5a385d4146b9..7d855167cafb 100644
--- a/net/mac80211/main.c
+++ b/net/mac80211/main.c
@@ -709,7 +709,7 @@ struct ieee80211_hw *ieee80211_alloc_hw_nm(size_t priv_data_len,
for (i = 0; i < IEEE80211_NUM_ACS; i++) {
struct airtime_sched_info *air_sched = &local->airtime[i];

- air_sched->active_txqs = RB_ROOT_CACHED;
+ airtime_sched_list_init(&air_sched->active_txqs);
INIT_LIST_HEAD(&air_sched->active_list);
spin_lock_init(&air_sched->lock);
air_sched->aql_txq_limit_low = IEEE80211_DEFAULT_AQL_TXQ_LIMIT_L;
diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c
index 2f1cf9d4e68d..9481a6ab1405 100644
--- a/net/mac80211/sta_info.c
+++ b/net/mac80211/sta_info.c
@@ -1914,8 +1914,7 @@ void ieee80211_register_airtime(struct ieee80211_txq *txq,
air_sched = &local->airtime[txq->ac];
air_info = to_airtime_info(txq);

- if (local->airtime_flags & AIRTIME_USE_TX)
- airtime += tx_airtime;
+ airtime += tx_airtime;
if (local->airtime_flags & AIRTIME_USE_RX)
airtime += rx_airtime;

diff --git a/net/mac80211/sta_info.h b/net/mac80211/sta_info.h
index 35c390bedfba..9b8774e889f5 100644
--- a/net/mac80211/sta_info.h
+++ b/net/mac80211/sta_info.h
@@ -135,11 +135,14 @@ enum ieee80211_agg_stop_reason {
#define AIRTIME_USE_TX BIT(0)
#define AIRTIME_USE_RX BIT(1)

+DECLARE_SKIPLIST_TYPE(airtime_sched, 5);

struct airtime_info {
+ struct airtime_sched_node schedule_order;
+ struct ieee80211_txq *txq[3];
u64 rx_airtime;
u64 tx_airtime;
- u64 v_t;
+ u64 v_t, v_t_cur;
u64 last_scheduled;
struct list_head list;
atomic_t aql_tx_pending; /* Estimated airtime for frames pending */
@@ -147,6 +150,7 @@ struct airtime_info {
u32 aql_limit_high;
u32 weight_reciprocal;
u16 weight;
+ u8 txq_idx;
};

void ieee80211_sta_update_pending_airtime(struct ieee80211_local *local,
diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index f844e9ed4c14..6bce89db9275 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -19,6 +19,7 @@
#include <linux/rcupdate.h>
#include <linux/export.h>
#include <linux/timekeeping.h>
+#include <linux/prandom.h>
#include <net/net_namespace.h>
#include <net/ieee80211_radiotap.h>
#include <net/cfg80211.h>
@@ -1477,11 +1478,12 @@ void ieee80211_txq_init(struct ieee80211_sub_if_data *sdata,
struct sta_info *sta,
struct txq_info *txqi, int tid)
{
+ struct airtime_info *air_info;
+
fq_tin_init(&txqi->tin);
codel_vars_init(&txqi->def_cvars);
codel_stats_init(&txqi->cstats);
__skb_queue_head_init(&txqi->frags);
- RB_CLEAR_NODE(&txqi->schedule_order);

txqi->txq.vif = &sdata->vif;

@@ -1490,7 +1492,7 @@ void ieee80211_txq_init(struct ieee80211_sub_if_data *sdata,
txqi->txq.tid = 0;
txqi->txq.ac = IEEE80211_AC_BE;

- return;
+ goto out;
}

if (tid == IEEE80211_NUM_TIDS) {
@@ -1512,6 +1514,12 @@ void ieee80211_txq_init(struct ieee80211_sub_if_data *sdata,
txqi->txq.sta = &sta->sta;
txqi->txq.tid = tid;
sta->sta.txq[tid] = &txqi->txq;
+
+out:
+ air_info = to_airtime_info(&txqi->txq);
+ air_info->txq[air_info->txq_idx++] = &txqi->txq;
+ if (air_info->txq_idx == ARRAY_SIZE(air_info->txq))
+ air_info->txq_idx--;
}

void ieee80211_txq_purge(struct ieee80211_local *local,
@@ -3632,6 +3640,8 @@ struct sk_buff *ieee80211_tx_dequeue(struct ieee80211_hw *hw,
struct ieee80211_tx_data tx;
ieee80211_tx_result r;
struct ieee80211_vif *vif = txq->vif;
+ u32 airtime;
+ bool ampdu;

WARN_ON_ONCE(softirq_count() == 0);

@@ -3790,20 +3800,35 @@ struct sk_buff *ieee80211_tx_dequeue(struct ieee80211_hw *hw,
encap_out:
IEEE80211_SKB_CB(skb)->control.vif = vif;

- if (vif &&
- wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL)) {
- bool ampdu = txq->ac != IEEE80211_AC_VO;
- u32 airtime;
+ if (!vif)
+ return skb;

- airtime = ieee80211_calc_expected_tx_airtime(hw, vif, txq->sta,
- skb->len, ampdu);
- if (!airtime)
- airtime = IEEE80211_TX_TIME_EST_UNIT;
+ ampdu = txq->ac != IEEE80211_AC_VO;
+ airtime = ieee80211_calc_expected_tx_airtime(hw, vif, txq->sta,
+ skb->len, ampdu);
+ if (!airtime)
+ airtime = IEEE80211_TX_TIME_EST_UNIT;

- airtime = ieee80211_info_set_tx_time_est(info, airtime);
- ieee80211_sta_update_pending_airtime(local, tx.sta, txq->ac,
- airtime, false);
- }
+ /*
+ * Tx queue scheduling always happens in airtime order and queues are
+ * sorted by virtual time + pending AQL budget.
+ * If AQL is not supported, pending AQL budget is always zero.
+ * If airtime fairness is not supported, virtual time won't be directly
+ * increased by driver tx completion.
+ * Because of that, we register estimated tx time as airtime if either
+ * AQL or ATF support is missing.
+ */
+ if (!wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL) ||
+ !wiphy_ext_feature_isset(local->hw.wiphy,
+ NL80211_EXT_FEATURE_AIRTIME_FAIRNESS))
+ ieee80211_register_airtime(txq, airtime, 0);
+
+ if (!wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL))
+ return skb;
+
+ airtime = ieee80211_info_set_tx_time_est(info, airtime);
+ ieee80211_sta_update_pending_airtime(local, tx.sta, txq->ac,
+ airtime, false);

return skb;

@@ -3814,85 +3839,92 @@ struct sk_buff *ieee80211_tx_dequeue(struct ieee80211_hw *hw,
}
EXPORT_SYMBOL(ieee80211_tx_dequeue);

+static struct ieee80211_txq *
+airtime_info_next_txq_idx(struct airtime_info *air_info)
+{
+ air_info->txq_idx++;
+ if (air_info->txq_idx >= ARRAY_SIZE(air_info->txq) ||
+ !air_info->txq[air_info->txq_idx])
+ air_info->txq_idx = 0;
+ return air_info->txq[air_info->txq_idx];
+}
+
struct ieee80211_txq *ieee80211_next_txq(struct ieee80211_hw *hw, u8 ac)
{
struct ieee80211_local *local = hw_to_local(hw);
struct airtime_sched_info *air_sched;
u64 now = ktime_get_coarse_boottime_ns();
- struct ieee80211_txq *ret = NULL;
+ struct airtime_sched_node *node = NULL;
+ struct ieee80211_txq *txq;
struct airtime_info *air_info;
struct txq_info *txqi = NULL;
- struct rb_node *node;
- bool first = false;
+ u8 txq_idx;

air_sched = &local->airtime[ac];
spin_lock_bh(&air_sched->lock);

- node = air_sched->schedule_pos;
-
begin:
- if (!node) {
- node = rb_first_cached(&air_sched->active_txqs);
- first = true;
- } else {
- node = rb_next(node);
- }
+ txq = NULL;
+ if (airtime_sched_peek(&air_sched->active_txqs) ==
+ air_sched->schedule_pos)
+ goto out;

+ node = airtime_sched_dequeue(&air_sched->active_txqs);
if (!node)
goto out;

- txqi = container_of(node, struct txq_info, schedule_order);
- air_info = to_airtime_info(&txqi->txq);
-
- if (air_info->v_t > air_sched->v_t &&
- (!first || !airtime_catchup_v_t(air_sched, air_info->v_t, now)))
- goto out;
+ air_info = container_of(node, struct airtime_info, schedule_order);

- if (!ieee80211_txq_airtime_check(hw, &txqi->txq)) {
- first = false;
+ txq = airtime_info_next_txq_idx(air_info);
+ txq_idx = air_info->txq_idx;
+ if (!ieee80211_txq_airtime_check(hw, txq))
goto begin;
+
+ while (1) {
+ txqi = to_txq_info(txq);
+ if (test_and_clear_bit(IEEE80211_TXQ_FORCE_ACTIVE, &txqi->flags))
+ break;
+
+ if (txq_has_queue(txq))
+ break;
+
+ txq = airtime_info_next_txq_idx(air_info);
+ if (txq_idx == air_info->txq_idx)
+ goto begin;
}

+ if (air_info->v_t_cur > air_sched->v_t)
+ airtime_catchup_v_t(air_sched, air_info->v_t_cur, now);
+
air_sched->schedule_pos = node;
air_sched->last_schedule_activity = now;
- ret = &txqi->txq;
out:
spin_unlock_bh(&air_sched->lock);
- return ret;
+ return txq;
}
EXPORT_SYMBOL(ieee80211_next_txq);

-static void __ieee80211_insert_txq(struct rb_root_cached *root,
+static void __ieee80211_insert_txq(struct ieee80211_local *local,
+ struct airtime_sched_info *air_sched,
struct txq_info *txqi)
{
- struct rb_node **new = &root->rb_root.rb_node;
- struct airtime_info *old_air, *new_air;
- struct rb_node *parent = NULL;
- struct txq_info *__txqi;
- bool leftmost = true;
-
- while (*new) {
- parent = *new;
- __txqi = rb_entry(parent, struct txq_info, schedule_order);
- old_air = to_airtime_info(&__txqi->txq);
- new_air = to_airtime_info(&txqi->txq);
-
- if (new_air->v_t <= old_air->v_t) {
- new = &parent->rb_left;
- } else {
- new = &parent->rb_right;
- leftmost = false;
- }
+ struct airtime_info *air_info = to_airtime_info(&txqi->txq);
+ u32 aql_time = 0;
+
+ if (wiphy_ext_feature_isset(local->hw.wiphy, NL80211_EXT_FEATURE_AQL)) {
+ aql_time = atomic_read(&air_info->aql_tx_pending);
+ aql_time *= air_info->weight_reciprocal;
+ aql_time >>= IEEE80211_RECIPROCAL_SHIFT_STA - IEEE80211_WEIGHT_SHIFT;
}

- rb_link_node(&txqi->schedule_order, parent, new);
- rb_insert_color_cached(&txqi->schedule_order, root, leftmost);
+ airtime_sched_delete(&air_sched->active_txqs, &air_info->schedule_order);
+ air_info->v_t_cur = air_info->v_t + aql_time;
+ airtime_sched_insert(&air_sched->active_txqs, &air_info->schedule_order);
}

void ieee80211_resort_txq(struct ieee80211_hw *hw,
struct ieee80211_txq *txq)
{
- struct airtime_info *air_info = to_airtime_info(txq);
struct ieee80211_local *local = hw_to_local(hw);
struct txq_info *txqi = to_txq_info(txq);
struct airtime_sched_info *air_sched;
@@ -3900,41 +3932,7 @@ void ieee80211_resort_txq(struct ieee80211_hw *hw,
air_sched = &local->airtime[txq->ac];

lockdep_assert_held(&air_sched->lock);
-
- if (!RB_EMPTY_NODE(&txqi->schedule_order)) {
- struct airtime_info *a_prev = NULL, *a_next = NULL;
- struct txq_info *t_prev, *t_next;
- struct rb_node *n_prev, *n_next;
-
- /* Erasing a node can cause an expensive rebalancing operation,
- * so we check the previous and next nodes first and only remove
- * and re-insert if the current node is not already in the
- * correct position.
- */
- if ((n_prev = rb_prev(&txqi->schedule_order)) != NULL) {
- t_prev = container_of(n_prev, struct txq_info,
- schedule_order);
- a_prev = to_airtime_info(&t_prev->txq);
- }
-
- if ((n_next = rb_next(&txqi->schedule_order)) != NULL) {
- t_next = container_of(n_next, struct txq_info,
- schedule_order);
- a_next = to_airtime_info(&t_next->txq);
- }
-
- if ((!a_prev || a_prev->v_t <= air_info->v_t) &&
- (!a_next || a_next->v_t > air_info->v_t))
- return;
-
- if (air_sched->schedule_pos == &txqi->schedule_order)
- air_sched->schedule_pos = n_prev;
-
- rb_erase_cached(&txqi->schedule_order,
- &air_sched->active_txqs);
- RB_CLEAR_NODE(&txqi->schedule_order);
- __ieee80211_insert_txq(&air_sched->active_txqs, txqi);
- }
+ __ieee80211_insert_txq(local, air_sched, txqi);
}

void ieee80211_update_airtime_weight(struct ieee80211_local *local,
@@ -3983,7 +3981,7 @@ void ieee80211_schedule_txq(struct ieee80211_hw *hw,
was_active = airtime_is_active(air_info, now);
airtime_set_active(air_sched, air_info, now);

- if (!RB_EMPTY_NODE(&txqi->schedule_order))
+ if (airtime_sched_is_queued(&air_info->schedule_order))
goto out;

/* If the station has been inactive for a while, catch up its v_t so it
@@ -3995,7 +3993,7 @@ void ieee80211_schedule_txq(struct ieee80211_hw *hw,
air_info->v_t = air_sched->v_t;

ieee80211_update_airtime_weight(local, air_sched, now, !was_active);
- __ieee80211_insert_txq(&air_sched->active_txqs, txqi);
+ __ieee80211_insert_txq(local, air_sched, txqi);

out:
spin_unlock_bh(&air_sched->lock);
@@ -4016,24 +4014,14 @@ static void __ieee80211_unschedule_txq(struct ieee80211_hw *hw,

lockdep_assert_held(&air_sched->lock);

+ airtime_sched_delete(&air_sched->active_txqs, &air_info->schedule_order);
if (purge) {
list_del_init(&air_info->list);
ieee80211_update_airtime_weight(local, air_sched, 0, true);
- }
-
- if (RB_EMPTY_NODE(&txqi->schedule_order))
- return;
-
- if (air_sched->schedule_pos == &txqi->schedule_order)
- air_sched->schedule_pos = rb_prev(&txqi->schedule_order);
-
- if (!purge)
+ } else {
airtime_set_active(air_sched, air_info,
ktime_get_coarse_boottime_ns());
-
- rb_erase_cached(&txqi->schedule_order,
- &air_sched->active_txqs);
- RB_CLEAR_NODE(&txqi->schedule_order);
+ }
}

void ieee80211_unschedule_txq(struct ieee80211_hw *hw,
@@ -4053,14 +4041,22 @@ void ieee80211_return_txq(struct ieee80211_hw *hw,
{
struct ieee80211_local *local = hw_to_local(hw);
struct txq_info *txqi = to_txq_info(txq);
+ struct airtime_sched_info *air_sched;
+ struct airtime_info *air_info;

- spin_lock_bh(&local->airtime[txq->ac].lock);
+ air_sched = &local->airtime[txq->ac];
+ air_info = to_airtime_info(&txqi->txq);

- if (!RB_EMPTY_NODE(&txqi->schedule_order) && !force &&
- !txq_has_queue(txq))
- __ieee80211_unschedule_txq(hw, txq, false);
+ if (force)
+ set_bit(IEEE80211_TXQ_FORCE_ACTIVE, &txqi->flags);

- spin_unlock_bh(&local->airtime[txq->ac].lock);
+ spin_lock_bh(&air_sched->lock);
+ if (force || (txq_has_queue(txq) &&
+ ieee80211_txq_airtime_check(hw, &txqi->txq)))
+ __ieee80211_insert_txq(local, air_sched, txqi);
+ else
+ __ieee80211_unschedule_txq(hw, txq, false);
+ spin_unlock_bh(&air_sched->lock);
}
EXPORT_SYMBOL(ieee80211_return_txq);

@@ -4112,9 +4108,6 @@ bool ieee80211_txq_may_transmit(struct ieee80211_hw *hw,
air_sched = &local->airtime[txq->ac];
spin_lock_bh(&air_sched->lock);

- if (RB_EMPTY_NODE(&txqi->schedule_order))
- goto out;
-
now = ktime_get_coarse_boottime_ns();

air_info = to_airtime_info(&txqi->txq);
@@ -4123,7 +4116,6 @@ bool ieee80211_txq_may_transmit(struct ieee80211_hw *hw,
ret = true;
}

-out:
spin_unlock_bh(&air_sched->lock);
return ret;
}
@@ -4134,9 +4126,7 @@ void ieee80211_txq_schedule_start(struct ieee80211_hw *hw, u8 ac)
struct ieee80211_local *local = hw_to_local(hw);
struct airtime_sched_info *air_sched = &local->airtime[ac];

- spin_lock_bh(&air_sched->lock);
air_sched->schedule_pos = NULL;
- spin_unlock_bh(&air_sched->lock);
}
EXPORT_SYMBOL(ieee80211_txq_schedule_start);

--
2.36.1

2022-06-12 00:32:51

by Felix Fietkau

[permalink] [raw]
Subject: [PATCH v2 2/4] mac80211: improve AQL tx time estimation

If airtime cannot be calculated because of missing or unsupported rate info,
use the smallest possible non-zero value for estimated tx time.
This improves handling of these cases by preventing queueing of as many packets
as the driver/hardware queue can hold for these stations.
Also slightly improve limiting queueing by explicitly rounding up small values.

Signed-off-by: Felix Fietkau <[email protected]>
---
include/net/mac80211.h | 18 +++++++++++-------
net/mac80211/status.c | 2 ++
net/mac80211/tx.c | 13 ++++++-------
3 files changed, 19 insertions(+), 14 deletions(-)

diff --git a/include/net/mac80211.h b/include/net/mac80211.h
index ebadb2103968..d8c4d49bcc49 100644
--- a/include/net/mac80211.h
+++ b/include/net/mac80211.h
@@ -1127,20 +1127,24 @@ struct ieee80211_tx_info {
};
};

+#define IEEE80211_TX_TIME_EST_UNIT 4
+
+static inline u16
+ieee80211_info_get_tx_time_est(struct ieee80211_tx_info *info)
+{
+ return info->tx_time_est * IEEE80211_TX_TIME_EST_UNIT;
+}
+
static inline u16
ieee80211_info_set_tx_time_est(struct ieee80211_tx_info *info, u16 tx_time_est)
{
/* We only have 10 bits in tx_time_est, so store airtime
* in increments of 4us and clamp the maximum to 2**12-1
*/
- info->tx_time_est = min_t(u16, tx_time_est, 4095) >> 2;
- return info->tx_time_est << 2;
-}
+ tx_time_est = DIV_ROUND_UP(tx_time_est, IEEE80211_TX_TIME_EST_UNIT);
+ info->tx_time_est = min_t(u16, tx_time_est, BIT(10) - 1);

-static inline u16
-ieee80211_info_get_tx_time_est(struct ieee80211_tx_info *info)
-{
- return info->tx_time_est << 2;
+ return ieee80211_info_get_tx_time_est(info);
}

/***
diff --git a/net/mac80211/status.c b/net/mac80211/status.c
index e69272139437..cea053d89dbb 100644
--- a/net/mac80211/status.c
+++ b/net/mac80211/status.c
@@ -1014,6 +1014,8 @@ static void __ieee80211_tx_status(struct ieee80211_hw *hw,
NULL,
skb->len,
false);
+ if (!airtime)
+ airtime = IEEE80211_TX_TIME_EST_UNIT;

ieee80211_register_airtime(txq, airtime, 0);
}
diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index 0e4efc08c762..ecb3975680bc 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -3797,13 +3797,12 @@ struct sk_buff *ieee80211_tx_dequeue(struct ieee80211_hw *hw,

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)
+ airtime = IEEE80211_TX_TIME_EST_UNIT;
+
+ airtime = ieee80211_info_set_tx_time_est(info, airtime);
+ ieee80211_sta_update_pending_airtime(local, tx.sta, txq->ac,
+ airtime, false);
}

return skb;
--
2.36.1

2022-06-18 09:04:12

by Felix Fietkau

[permalink] [raw]
Subject: Re: [PATCH v2 1/4] mac80211: fix overflow issues in airtime fairness code

On 11.06.22 22:01, Felix Fietkau wrote:
> The airtime weight calculation overflows with a default weight value of 256
> whenever more than 8ms worth of airtime is reported.
> Bigger weight values impose even smaller limits on maximum airtime values.
> This can mess up airtime based calculations for drivers that don't report
> per-PPDU airtime values, but batch up values instead.
>
> Fix this by reordering multiplications/shifts and by reducing unnecessary
> intermediate precision (which was lost in a later stage anyway).
>
> The new shift value limits the maximum weight to 4096, which should be more
> than enough. Any values bigger than that will be rejected
>
> Signed-off-by: Felix Fietkau <[email protected]>
Please don't merge this series yet. I'm still investigating reports of
regressions caused by it.

- Felix