Yibo Zhao <[email protected]> writes: > Global airtime weight sum is updated only when txq is added/removed > from rbtree. If upper layer configures sta weight during high load, > airtime weight sum will not be updated since txq is most likely on the > tree. It could a little late for upper layer to reconfigure sta weight > when txq is already in the rbtree. And thus, incorrect airtime weight sum > will lead to incorrect global virtual time calculation as well as global > airtime weight sum overflow of airtime weight sum during txq removed. > > Hence, need to update airtime weight sum upon receiving event for > configuring sta weight once sta's txq is on the rbtree. > > Besides, if airtime weight sum of ACs and sta weight is synced under the > same per AC lock protection, there can be a very short window causing > incorrct airtime weight sum calculation as below: > > active_txq_lock_VO . > VO weight sum is syncd . > sta airtime weight sum is synced . > active_txq_unlock_VO . > . . > active_txq_lock_VI . > VI weight sum is syncd . > sta airtime weight sum active_txq_lock_BE > active_txq_unlock_VI Remove txq and thus sum > . is calculated with synced > . sta airtime weight > . active_txq_unlock_BE > > So introduce a per ac synced station airtime weight synced with per > AC synced weight sum together. And the per-AC station airtime weight > is used to calculate weight sum. > > Signed-off-by: Yibo Zhao <[email protected]> > --- > net/mac80211/cfg.c | 27 +++++++++++++++++++++++++-- > net/mac80211/sta_info.c | 6 ++++-- > net/mac80211/sta_info.h | 3 +++ > net/mac80211/tx.c | 4 ++-- > 4 files changed, 34 insertions(+), 6 deletions(-) > > diff --git a/net/mac80211/cfg.c b/net/mac80211/cfg.c > index d65aa01..4b420bb 100644 > --- a/net/mac80211/cfg.c > +++ b/net/mac80211/cfg.c > @@ -1284,7 +1284,8 @@ static int sta_apply_parameters(struct ieee80211_local > *local, > int ret = 0; > struct ieee80211_supported_band *sband; > struct ieee80211_sub_if_data *sdata = sta->sdata; > - u32 mask, set; > + u32 mask, set, tid, ac; > + struct txq_info *txqi; > > sband = ieee80211_get_sband(sdata); > if (!sband) > @@ -1452,8 +1453,30 @@ static int sta_apply_parameters(struct ieee80211_local > *local, > if (ieee80211_vif_is_mesh(&sdata->vif)) > sta_apply_mesh_params(local, sta, params); > > - if (params->airtime_weight) > + if (params->airtime_weight && > + params->airtime_weight != sta->airtime_weight) { > + for (ac = 0; ac < IEEE80211_NUM_ACS; ac++) { > + spin_lock_bh(&local->active_txq_lock[ac]); > + for (tid = 0; tid < IEEE80211_NUM_TIDS + 1; tid++) { > + if (!sta->sta.txq[tid] || > + ac != ieee80211_ac_from_tid(tid)) > + continue; > + > + sta->airtime_weight_synced[ac] = > + params->airtime_weight; > + > + txqi = to_txq_info(sta->sta.txq[tid]); > + if (RB_EMPTY_NODE(&txqi->schedule_order)) > + continue; > + > + local->airtime_weight_sum[ac] = > local->airtime_weight_sum[ac] + > + > params->airtime_weight - > + > sta->airtime_weight; > + } > + spin_unlock_bh(&local->active_txq_lock[ac]); > + } > sta->airtime_weight = params->airtime_weight;
With this, airtime_weight is basically only used to return to and from userspace, right? I.e., after the above loop has run, it will match the contents of airtime_weight_synced; so why not just turn airtime_weight into a per-ac array? You could just use airtime_weight[0] as the value to return to userspace... -Toke
