From: Felix Fietkau Date: Thu, 15 Sep 2022 14:38:18 +0000 (+0200) Subject: mac80211: merge pending fixes for tx queueing issues X-Git-Tag: v23.05.0-rc1~2420 X-Git-Url: http://git.cdn.openwrt.org/?a=commitdiff_plain;h=8b06e06832ebe757246582b65306ad2a2537741f;p=openwrt%2Fstaging%2Fpepe2k.git mac80211: merge pending fixes for tx queueing issues Fixes a potential deadlock and a tx queue hang on STA assoc Signed-off-by: Felix Fietkau --- diff --git a/package/kernel/mac80211/patches/subsys/341-mac80211-Fix-deadlock-Don-t-start-TX-while-holding-f.patch b/package/kernel/mac80211/patches/subsys/341-mac80211-Fix-deadlock-Don-t-start-TX-while-holding-f.patch new file mode 100644 index 0000000000..8c56acbf88 --- /dev/null +++ b/package/kernel/mac80211/patches/subsys/341-mac80211-Fix-deadlock-Don-t-start-TX-while-holding-f.patch @@ -0,0 +1,40 @@ +From: Alexander Wetzel +Date: Thu, 15 Sep 2022 14:41:20 +0200 +Subject: [PATCH] mac80211: Fix deadlock: Don't start TX while holding + fq->lock + +ieee80211_txq_purge() calls fq_tin_reset() and +ieee80211_purge_tx_queue(); Both are then calling +ieee80211_free_txskb(). Which can decide to TX the skb again. + +There are at least two ways to get a deadlock: + +1) When we have a TDLS teardown packet queued in either tin or frags + ieee80211_tdls_td_tx_handle() will call ieee80211_subif_start_xmit() + while we still hold fq->lock. ieee80211_txq_enqueue() will thus + deadlock. + +2) A variant of the above happens if aggregation is up and running: + In that case ieee80211_iface_work() will deadlock with the original + task: The original tasks already holds fq->lock and tries to get + sta->lock after kicking off ieee80211_iface_work(). But the worker + can get sta->lock prior to the original task and will then spin for + fq->lock. + +Avoid these deadlocks by not sending out any skbs when called via +ieee80211_free_txskb(). + +Signed-off-by: Alexander Wetzel +--- + +--- a/net/mac80211/status.c ++++ b/net/mac80211/status.c +@@ -698,7 +698,7 @@ static void ieee80211_report_used_skb(st + + if (!sdata) { + skb->dev = NULL; +- } else { ++ } else if (!dropped) { + unsigned int hdr_size = + ieee80211_hdrlen(hdr->frame_control); + diff --git a/package/kernel/mac80211/patches/subsys/342-mac80211-Ensure-vif-queues-are-operational-after-sta.patch b/package/kernel/mac80211/patches/subsys/342-mac80211-Ensure-vif-queues-are-operational-after-sta.patch new file mode 100644 index 0000000000..4310329319 --- /dev/null +++ b/package/kernel/mac80211/patches/subsys/342-mac80211-Ensure-vif-queues-are-operational-after-sta.patch @@ -0,0 +1,47 @@ +From: Alexander Wetzel +Date: Thu, 15 Sep 2022 15:09:46 +0200 +Subject: [PATCH] mac80211: Ensure vif queues are operational after start + +Make sure local->queue_stop_reasons and vif.txqs_stopped stay in sync. + +When a new vif is created the queues may end up in an inconsistent state +and be inoperable: +Communication not using iTXQ will work, allowing to e.g. complete the +association. But the 4-way handshake will time out. The sta will not +send out any skbs queued in iTXQs. + +All normal attempts to start the queues will fail when reaching this +state. +local->queue_stop_reasons will have marked all queues as operational but +vif.txqs_stopped will still be set, creating an inconsistent internal +state. + +In reality this seems to be race between the mac80211 function +ieee80211_do_open() setting SDATA_STATE_RUNNING and the wake_txqs_tasklet: +Depending on the driver and the timing the queues may end up to be +operational or not. + +Cc: stable@vger.kernel.org +Fixes: f856373e2f31 ("wifi: mac80211: do not wake queues on a vif that is being stopped") +Signed-off-by: Alexander Wetzel +--- + +--- a/net/mac80211/util.c ++++ b/net/mac80211/util.c +@@ -301,14 +301,14 @@ static void __ieee80211_wake_txqs(struct + local_bh_disable(); + spin_lock(&fq->lock); + ++ sdata->vif.txqs_stopped[ac] = false; ++ + if (!test_bit(SDATA_STATE_RUNNING, &sdata->state)) + goto out; + + if (sdata->vif.type == NL80211_IFTYPE_AP) + ps = &sdata->bss->ps; + +- sdata->vif.txqs_stopped[ac] = false; +- + list_for_each_entry_rcu(sta, &local->sta_list, list) { + if (sdata != sta->sdata) + continue;