| Message ID | 1360259677-19278-1-git-send-email-seth.forshee@canonical.com (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Headers | show |
On Thu, 2013-02-07 at 11:54 -0600, Seth Forshee wrote: > Scans currently work by stopping the netdev tx queues but leaving the > mac80211 queues active. This stops the flow of incoming packets while > still allowing mac80211 to transmit nullfunc and probe request frames to > facilitate scanning. However, the driver may try to wake the mac80211 > queues while in this state, which will also wake the netdev queues. > > To prevent this, add a new queue stop reason, > IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL, to be used when stopping the tx > queues for off-channel operation. This prevents the netdev queues from > waking when a driver wakes the mac80211 queues. > > This also stops all frames from being transmitted, even those required > for scanning. To get around this, add a new offchan_tx_ok argument to > most of the tx interfaces. This flag can be set for frames which need to > be transmitted during off-channel operation, allowing off-channel frames > to be passed down to the driver if the queues have only been stopped for > off-channel. Add ieee80211_tx_skb_offchannel() for transmitting > off-channel frames with this flag set. I started wondering -- is there a reason to modify the entire TX path? Could we maybe bypass it instead and call the driver's TX op almost directly? The frames in question don't really need much TX handling, the only thing that might be relevant _could_ be rate control but even that I'd argue isn't really needed, just using rate_control_send_low() should be ok (by setting IEEE80211_TX_CTL_USE_MINRATE it will always return true). For the null data packets the sta pointer is also obvious, the AP station (BSSID) ... we don't need any of the extra monitor/whatever handling either. That might be simpler overall? johannes -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Feb 08, 2013 at 10:03:32AM +0100, Johannes Berg wrote: > On Thu, 2013-02-07 at 11:54 -0600, Seth Forshee wrote: > > Scans currently work by stopping the netdev tx queues but leaving the > > mac80211 queues active. This stops the flow of incoming packets while > > still allowing mac80211 to transmit nullfunc and probe request frames to > > facilitate scanning. However, the driver may try to wake the mac80211 > > queues while in this state, which will also wake the netdev queues. > > > > To prevent this, add a new queue stop reason, > > IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL, to be used when stopping the tx > > queues for off-channel operation. This prevents the netdev queues from > > waking when a driver wakes the mac80211 queues. > > > > This also stops all frames from being transmitted, even those required > > for scanning. To get around this, add a new offchan_tx_ok argument to > > most of the tx interfaces. This flag can be set for frames which need to > > be transmitted during off-channel operation, allowing off-channel frames > > to be passed down to the driver if the queues have only been stopped for > > off-channel. Add ieee80211_tx_skb_offchannel() for transmitting > > off-channel frames with this flag set. > > I started wondering -- is there a reason to modify the entire TX path? > Could we maybe bypass it instead and call the driver's TX op almost > directly? The frames in question don't really need much TX handling, the > only thing that might be relevant _could_ be rate control but even that > I'd argue isn't really needed, just using rate_control_send_low() should > be ok (by setting IEEE80211_TX_CTL_USE_MINRATE it will always return > true). For the null data packets the sta pointer is also obvious, the AP > station (BSSID) ... we don't need any of the extra monitor/whatever > handling either. > > That might be simpler overall? Okay, I'll take a look at this. Another option that might simplify things a bit would be to use a ieee80211_tx_data flag. If I added another interface into tx.c for offchannel frames then the offchan argument would only be needed for ieee80211_xmit() and ieee80211_tx(). Though it would be nice to avoid adding an argument to ieee80211_xmit(). Seth -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 2013-02-08 at 11:11 -0600, Seth Forshee wrote: > > I started wondering -- is there a reason to modify the entire TX path? > > Could we maybe bypass it instead and call the driver's TX op almost > > directly? The frames in question don't really need much TX handling, the > > only thing that might be relevant _could_ be rate control but even that > > I'd argue isn't really needed, just using rate_control_send_low() should > > be ok (by setting IEEE80211_TX_CTL_USE_MINRATE it will always return > > true). For the null data packets the sta pointer is also obvious, the AP > > station (BSSID) ... we don't need any of the extra monitor/whatever > > handling either. > > > > That might be simpler overall? > > Okay, I'll take a look at this. > > Another option that might simplify things a bit would be to use a > ieee80211_tx_data flag. If I added another interface into tx.c for > offchannel frames then the offchan argument would only be needed for > ieee80211_xmit() and ieee80211_tx(). Though it would be nice to avoid > adding an argument to ieee80211_xmit(). Oh, I forgot all about ROC and off-channel frames there, but those are public action frames (only?) so the same applies with min-rate etc. Overall I'm not sure. On the one hand that might make the code changes simpler, on the other it might make the code more complex? johannes -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Feb 08, 2013 at 06:17:32PM +0100, Johannes Berg wrote: > On Fri, 2013-02-08 at 11:11 -0600, Seth Forshee wrote: > > > > I started wondering -- is there a reason to modify the entire TX path? > > > Could we maybe bypass it instead and call the driver's TX op almost > > > directly? The frames in question don't really need much TX handling, the > > > only thing that might be relevant _could_ be rate control but even that > > > I'd argue isn't really needed, just using rate_control_send_low() should > > > be ok (by setting IEEE80211_TX_CTL_USE_MINRATE it will always return > > > true). For the null data packets the sta pointer is also obvious, the AP > > > station (BSSID) ... we don't need any of the extra monitor/whatever > > > handling either. > > > > > > That might be simpler overall? > > > > Okay, I'll take a look at this. > > > > Another option that might simplify things a bit would be to use a > > ieee80211_tx_data flag. If I added another interface into tx.c for > > offchannel frames then the offchan argument would only be needed for > > ieee80211_xmit() and ieee80211_tx(). Though it would be nice to avoid > > adding an argument to ieee80211_xmit(). This actually doesn't work out. ieee80211_tx_data doesn't propogate down to ieee80211_tx_frags() and isn't even there in the txpending case. So without more substantial changes it just turns back into an argument after a couple of levels. > Oh, I forgot all about ROC and off-channel frames there, but those are > public action frames (only?) so the same applies with min-rate etc. I thought of that and was going to check and see what kind of ROC frames are sent. Also probe request frames are obviously sent off-channel, but I'd guess the same applies to them as well? > Overall I'm not sure. On the one hand that might make the code changes > simpler, on the other it might make the code more complex? Well, I think the way that will be simplest with the fewest code changes would be to use a tx control flag. Of course then we've gobbled up one of the last available flags. Seth -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 2013-02-08 at 12:10 -0600, Seth Forshee wrote: > Well, I think the way that will be simplest with the fewest code changes > would be to use a tx control flag. Of course then we've gobbled up one > of the last available flags. Let's do that anyway then. I still think we need to do the PS/scan/offchannel thing I described in another mail anyway, so that'd be a better interim step than changing all the prototypes... johannes -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Feb 08, 2013 at 08:05:59PM +0100, Johannes Berg wrote: > On Fri, 2013-02-08 at 12:10 -0600, Seth Forshee wrote: > > > Well, I think the way that will be simplest with the fewest code changes > > would be to use a tx control flag. Of course then we've gobbled up one > > of the last available flags. > > Let's do that anyway then. I still think we need to do the > PS/scan/offchannel thing I described in another mail anyway, so that'd > be a better interim step than changing all the prototypes... When I originally looked at using a tx control flag I didn't think that using IEEE80211_TX_CTL_TX_OFFCHAN would work, but now I'm not sure why. Is there any reason not to do this? Seth -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 2013-02-08 at 14:46 -0600, Seth Forshee wrote: > > > Well, I think the way that will be simplest with the fewest code changes > > > would be to use a tx control flag. Of course then we've gobbled up one > > > of the last available flags. > > > > Let's do that anyway then. I still think we need to do the > > PS/scan/offchannel thing I described in another mail anyway, so that'd > > be a better interim step than changing all the prototypes... > > When I originally looked at using a tx control flag I didn't think that > using IEEE80211_TX_CTL_TX_OFFCHAN would work, but now I'm not sure why. > Is there any reason not to do this? I guess you'd have to pre-assign the queue, since the code in ieee80211_tx() might skip that part, but that seems easy enough. Other than that, a driver that checks IEEE80211_TX_CTL_TX_OFFCHAN might treat that frame specially, but OTOH the only drivers using it right now are ours and TI's and they both have HW scan/roc, so your code never executes. What a future driver might do is anyone's guess... johannes -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, Feb 08, 2013 at 09:53:12PM +0100, Johannes Berg wrote: > On Fri, 2013-02-08 at 14:46 -0600, Seth Forshee wrote: > > > > > Well, I think the way that will be simplest with the fewest code changes > > > > would be to use a tx control flag. Of course then we've gobbled up one > > > > of the last available flags. > > > > > > Let's do that anyway then. I still think we need to do the > > > PS/scan/offchannel thing I described in another mail anyway, so that'd > > > be a better interim step than changing all the prototypes... > > > > When I originally looked at using a tx control flag I didn't think that > > using IEEE80211_TX_CTL_TX_OFFCHAN would work, but now I'm not sure why. > > Is there any reason not to do this? > > I guess you'd have to pre-assign the queue, since the code in > ieee80211_tx() might skip that part, but that seems easy enough. Other > than that, a driver that checks IEEE80211_TX_CTL_TX_OFFCHAN might treat > that frame specially, but OTOH the only drivers using it right now are > ours and TI's and they both have HW scan/roc, so your code never > executes. What a future driver might do is anyone's guess... All right, I decided to go ahead with the new flag since what I need it to do is somewhat incongruous with what IEEE80211_TX_CTL_TX_OFFCHAN indicates. New patches will follow. Seth -- To unsubscribe from this list: send the line "unsubscribe linux-wireless" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/net/mac80211/agg-tx.c b/net/mac80211/agg-tx.c index 13b7683..f0840b3 100644 --- a/net/mac80211/agg-tx.c +++ b/net/mac80211/agg-tx.c @@ -107,7 +107,7 @@ static void ieee80211_send_addba_request(struct ieee80211_sub_if_data *sdata, mgmt->u.action.u.addba_req.start_seq_num = cpu_to_le16(start_seq_num << 4); - ieee80211_tx_skb_tid(sdata, skb, tid); + ieee80211_tx_skb_tid(sdata, skb, tid, false); } void ieee80211_send_bar(struct ieee80211_vif *vif, u8 *ra, u16 tid, u16 ssn) @@ -137,7 +137,7 @@ void ieee80211_send_bar(struct ieee80211_vif *vif, u8 *ra, u16 tid, u16 ssn) IEEE80211_SKB_CB(skb)->flags |= IEEE80211_TX_INTFL_DONT_ENCRYPT | IEEE80211_TX_CTL_REQ_TX_STATUS; - ieee80211_tx_skb_tid(sdata, skb, tid); + ieee80211_tx_skb_tid(sdata, skb, tid, false); } EXPORT_SYMBOL(ieee80211_send_bar); diff --git a/net/mac80211/cfg.c b/net/mac80211/cfg.c index 661b878..fd38c37 100644 --- a/net/mac80211/cfg.c +++ b/net/mac80211/cfg.c @@ -3211,7 +3211,7 @@ static int ieee80211_probe_client(struct wiphy *wiphy, struct net_device *dev, nullfunc->qos_ctrl = cpu_to_le16(7); local_bh_disable(); - ieee80211_xmit(sdata, skb, band); + ieee80211_xmit(sdata, skb, band, false); local_bh_enable(); rcu_read_unlock(); diff --git a/net/mac80211/ht.c b/net/mac80211/ht.c index 61ac7c4..da61c41 100644 --- a/net/mac80211/ht.c +++ b/net/mac80211/ht.c @@ -311,7 +311,7 @@ void ieee80211_send_delba(struct ieee80211_sub_if_data *sdata, mgmt->u.action.u.delba.params = cpu_to_le16(params); mgmt->u.action.u.delba.reason_code = cpu_to_le16(reason_code); - ieee80211_tx_skb_tid(sdata, skb, tid); + ieee80211_tx_skb_tid(sdata, skb, tid, false); } void ieee80211_process_delba(struct ieee80211_sub_if_data *sdata, diff --git a/net/mac80211/ieee80211_i.h b/net/mac80211/ieee80211_i.h index 5fba867..600ffd2 100644 --- a/net/mac80211/ieee80211_i.h +++ b/net/mac80211/ieee80211_i.h @@ -840,6 +840,7 @@ enum queue_stop_reason { IEEE80211_QUEUE_STOP_REASON_AGGREGATION, IEEE80211_QUEUE_STOP_REASON_SUSPEND, IEEE80211_QUEUE_STOP_REASON_SKB_ADD, + IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL, }; #ifdef CONFIG_MAC80211_LEDS @@ -1495,24 +1496,25 @@ void mac80211_ev_michael_mic_failure(struct ieee80211_sub_if_data *sdata, int ke void ieee80211_set_wmm_default(struct ieee80211_sub_if_data *sdata, bool bss_notify); void ieee80211_xmit(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb, - enum ieee80211_band band); + enum ieee80211_band band, bool offchan_tx_ok); void __ieee80211_tx_skb_tid_band(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb, int tid, - enum ieee80211_band band); + enum ieee80211_band band, bool offchan_tx_ok); static inline void ieee80211_tx_skb_tid_band(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb, int tid, - enum ieee80211_band band) + enum ieee80211_band band, bool offchan_tx_ok) { rcu_read_lock(); - __ieee80211_tx_skb_tid_band(sdata, skb, tid, band); + __ieee80211_tx_skb_tid_band(sdata, skb, tid, band, offchan_tx_ok); rcu_read_unlock(); } static inline void ieee80211_tx_skb_tid(struct ieee80211_sub_if_data *sdata, - struct sk_buff *skb, int tid) + struct sk_buff *skb, int tid, + bool offchan_tx_ok) { struct ieee80211_chanctx_conf *chanctx_conf; @@ -1525,7 +1527,8 @@ static inline void ieee80211_tx_skb_tid(struct ieee80211_sub_if_data *sdata, } __ieee80211_tx_skb_tid_band(sdata, skb, tid, - chanctx_conf->def.chan->band); + chanctx_conf->def.chan->band, + offchan_tx_ok); rcu_read_unlock(); } @@ -1533,7 +1536,15 @@ static inline void ieee80211_tx_skb(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb) { /* Send all internal mgmt frames on VO. Accordingly set TID to 7. */ - ieee80211_tx_skb_tid(sdata, skb, 7); + ieee80211_tx_skb_tid(sdata, skb, 7, false); +} + +static inline void +ieee80211_tx_skb_offchannel(struct ieee80211_sub_if_data *sdata, + struct sk_buff *skb) +{ + /* Send all internal mgmt frames on VO. Accordingly set TID to 7. */ + ieee80211_tx_skb_tid(sdata, skb, 7, true); } void ieee802_11_parse_elems(u8 *start, size_t len, diff --git a/net/mac80211/mlme.c b/net/mac80211/mlme.c index 5913fb9..d5a3cf7 100644 --- a/net/mac80211/mlme.c +++ b/net/mac80211/mlme.c @@ -685,7 +685,7 @@ void ieee80211_send_nullfunc(struct ieee80211_local *local, IEEE80211_STA_CONNECTION_POLL)) IEEE80211_SKB_CB(skb)->flags |= IEEE80211_TX_CTL_USE_MINRATE; - ieee80211_tx_skb(sdata, skb); + ieee80211_tx_skb_offchannel(sdata, skb); } static void ieee80211_send_4addr_nullfunc(struct ieee80211_local *local, diff --git a/net/mac80211/offchannel.c b/net/mac80211/offchannel.c index 82baf5b..5b9b3b8 100644 --- a/net/mac80211/offchannel.c +++ b/net/mac80211/offchannel.c @@ -113,6 +113,10 @@ void ieee80211_offchannel_stop_vifs(struct ieee80211_local *local) * notify the AP about us leaving the channel and stop all * STA interfaces. */ + + ieee80211_stop_queues_by_reason(&local->hw, + IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL); + mutex_lock(&local->iflist_mtx); list_for_each_entry(sdata, &local->interfaces, list) { if (!ieee80211_sdata_running(sdata)) @@ -133,12 +137,9 @@ void ieee80211_offchannel_stop_vifs(struct ieee80211_local *local) sdata, BSS_CHANGED_BEACON_ENABLED); } - if (sdata->vif.type != NL80211_IFTYPE_MONITOR) { - netif_tx_stop_all_queues(sdata->dev); - if (sdata->vif.type == NL80211_IFTYPE_STATION && - sdata->u.mgd.associated) - ieee80211_offchannel_ps_enable(sdata); - } + if (sdata->vif.type == NL80211_IFTYPE_STATION && + sdata->u.mgd.associated) + ieee80211_offchannel_ps_enable(sdata); } mutex_unlock(&local->iflist_mtx); } @@ -166,20 +167,6 @@ void ieee80211_offchannel_return(struct ieee80211_local *local) sdata->u.mgd.associated) ieee80211_offchannel_ps_disable(sdata); - if (sdata->vif.type != NL80211_IFTYPE_MONITOR) { - /* - * This may wake up queues even though the driver - * currently has them stopped. This is not very - * likely, since the driver won't have gotten any - * (or hardly any) new packets while we weren't - * on the right channel, and even if it happens - * it will at most lead to queueing up one more - * packet per queue in mac80211 rather than on - * the interface qdisc. - */ - netif_tx_wake_all_queues(sdata->dev); - } - if (test_and_clear_bit(SDATA_STATE_OFFCHANNEL_BEACON_STOPPED, &sdata->state)) { sdata->vif.bss_conf.enable_beacon = true; @@ -188,6 +175,9 @@ void ieee80211_offchannel_return(struct ieee80211_local *local) } } mutex_unlock(&local->iflist_mtx); + + ieee80211_wake_queues_by_reason(&local->hw, + IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL); } void ieee80211_handle_roc_started(struct ieee80211_roc_work *roc) @@ -198,7 +188,7 @@ void ieee80211_handle_roc_started(struct ieee80211_roc_work *roc) if (roc->mgmt_tx_cookie) { if (!WARN_ON(!roc->frame)) { ieee80211_tx_skb_tid_band(roc->sdata, roc->frame, 7, - roc->chan->band); + roc->chan->band, true); roc->frame = NULL; } } else { diff --git a/net/mac80211/sta_info.c b/net/mac80211/sta_info.c index 227233c..f57247c 100644 --- a/net/mac80211/sta_info.c +++ b/net/mac80211/sta_info.c @@ -1117,7 +1117,7 @@ static void ieee80211_send_null_response(struct ieee80211_sub_if_data *sdata, return; } - ieee80211_xmit(sdata, skb, chanctx_conf->def.chan->band); + ieee80211_xmit(sdata, skb, chanctx_conf->def.chan->band, false); rcu_read_unlock(); } diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c index a2cb6a3..f477858 100644 --- a/net/mac80211/tx.c +++ b/net/mac80211/tx.c @@ -1204,7 +1204,7 @@ static bool ieee80211_tx_frags(struct ieee80211_local *local, struct ieee80211_vif *vif, struct ieee80211_sta *sta, struct sk_buff_head *skbs, - bool txpending) + bool txpending, bool offchan_tx_ok) { struct ieee80211_tx_control control; struct sk_buff *skb, *tmp; @@ -1223,8 +1223,21 @@ static bool ieee80211_tx_frags(struct ieee80211_local *local, #endif spin_lock_irqsave(&local->queue_stop_reason_lock, flags); - if (local->queue_stop_reasons[q] || - (!txpending && !skb_queue_empty(&local->pending[q]))) { + if (unlikely(offchan_tx_ok)) { + /* + * Drop off-channel frames if queues are stopped for + * any reason other than off-channel operation. Never + * queue them. + */ + if (local->queue_stop_reasons[q] & + ~BIT(IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL)) { + spin_unlock_irqrestore(&local->queue_stop_reason_lock, + flags); + ieee80211_purge_tx_queue(&local->hw, skbs); + return true; + } + } else if (local->queue_stop_reasons[q] || + (!txpending && !skb_queue_empty(&local->pending[q]))) { /* * Since queue is stopped, queue up frames for later * transmission from the tx-pending tasklet when the @@ -1257,7 +1270,8 @@ static bool ieee80211_tx_frags(struct ieee80211_local *local, */ static bool __ieee80211_tx(struct ieee80211_local *local, struct sk_buff_head *skbs, int led_len, - struct sta_info *sta, bool txpending) + struct sta_info *sta, bool txpending, + bool offchan_tx_ok) { struct ieee80211_tx_info *info; struct ieee80211_sub_if_data *sdata; @@ -1305,7 +1319,7 @@ static bool __ieee80211_tx(struct ieee80211_local *local, } result = ieee80211_tx_frags(local, vif, pubsta, skbs, - txpending); + txpending, offchan_tx_ok); ieee80211_tpt_led_trig_tx(local, fc, led_len); ieee80211_led_tx(local, 1); @@ -1376,7 +1390,7 @@ static int invoke_tx_handlers(struct ieee80211_tx_data *tx) */ static bool ieee80211_tx(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb, bool txpending, - enum ieee80211_band band) + enum ieee80211_band band, bool offchan_tx_ok) { struct ieee80211_local *local = sdata->local; struct ieee80211_tx_data tx; @@ -1411,7 +1425,7 @@ static bool ieee80211_tx(struct ieee80211_sub_if_data *sdata, if (!invoke_tx_handlers(&tx)) result = __ieee80211_tx(local, &tx.skbs, led_len, - tx.sta, txpending); + tx.sta, txpending, offchan_tx_ok); return result; } @@ -1448,7 +1462,7 @@ static int ieee80211_skb_resize(struct ieee80211_sub_if_data *sdata, } void ieee80211_xmit(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb, - enum ieee80211_band band) + enum ieee80211_band band, bool offchan_tx_ok) { struct ieee80211_local *local = sdata->local; struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb); @@ -1481,7 +1495,7 @@ void ieee80211_xmit(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb, } ieee80211_set_qos_hdr(sdata, skb); - ieee80211_tx(sdata, skb, false, band); + ieee80211_tx(sdata, skb, false, band, offchan_tx_ok); } static bool ieee80211_parse_tx_radiotap(struct sk_buff *skb) @@ -1701,7 +1715,7 @@ netdev_tx_t ieee80211_monitor_start_xmit(struct sk_buff *skb, IEEE80211_CHAN_PASSIVE_SCAN))) goto fail_rcu; - ieee80211_xmit(sdata, skb, chan->band); + ieee80211_xmit(sdata, skb, chan->band, false); rcu_read_unlock(); return NETDEV_TX_OK; @@ -2145,7 +2159,7 @@ netdev_tx_t ieee80211_subif_start_xmit(struct sk_buff *skb, info->flags = info_flags; info->ack_frame_id = info_id; - ieee80211_xmit(sdata, skb, band); + ieee80211_xmit(sdata, skb, band, false); rcu_read_unlock(); return NETDEV_TX_OK; @@ -2197,7 +2211,7 @@ static bool ieee80211_tx_pending_skb(struct ieee80211_local *local, return true; } result = ieee80211_tx(sdata, skb, true, - chanctx_conf->def.chan->band); + chanctx_conf->def.chan->band, false); } else { struct sk_buff_head skbs; @@ -2207,7 +2221,8 @@ static bool ieee80211_tx_pending_skb(struct ieee80211_local *local, hdr = (struct ieee80211_hdr *)skb->data; sta = sta_info_get(sdata, hdr->addr1); - result = __ieee80211_tx(local, &skbs, skb->len, sta, true); + result = __ieee80211_tx(local, &skbs, skb->len, sta, true, + false); } return result; @@ -2777,7 +2792,7 @@ EXPORT_SYMBOL(ieee80211_get_buffered_bc); void __ieee80211_tx_skb_tid_band(struct ieee80211_sub_if_data *sdata, struct sk_buff *skb, int tid, - enum ieee80211_band band) + enum ieee80211_band band, bool offchan_tx_ok) { int ac = ieee802_1d_to_ac[tid & 7]; @@ -2794,6 +2809,6 @@ void __ieee80211_tx_skb_tid_band(struct ieee80211_sub_if_data *sdata, * requirements are that we do not come into tx with bhs on. */ local_bh_disable(); - ieee80211_xmit(sdata, skb, band); + ieee80211_xmit(sdata, skb, band, offchan_tx_ok); local_bh_enable(); } diff --git a/net/mac80211/util.c b/net/mac80211/util.c index 7519018..5259557 100644 --- a/net/mac80211/util.c +++ b/net/mac80211/util.c @@ -1290,7 +1290,8 @@ void ieee80211_send_probe_req(struct ieee80211_sub_if_data *sdata, u8 *dst, IEEE80211_SKB_CB(skb)->flags |= IEEE80211_TX_CTL_NO_CCK_RATE; if (scan) - ieee80211_tx_skb_tid_band(sdata, skb, 7, channel->band); + ieee80211_tx_skb_tid_band(sdata, skb, 7, + channel->band, true); else ieee80211_tx_skb(sdata, skb); }
Scans currently work by stopping the netdev tx queues but leaving the mac80211 queues active. This stops the flow of incoming packets while still allowing mac80211 to transmit nullfunc and probe request frames to facilitate scanning. However, the driver may try to wake the mac80211 queues while in this state, which will also wake the netdev queues. To prevent this, add a new queue stop reason, IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL, to be used when stopping the tx queues for off-channel operation. This prevents the netdev queues from waking when a driver wakes the mac80211 queues. This also stops all frames from being transmitted, even those required for scanning. To get around this, add a new offchan_tx_ok argument to most of the tx interfaces. This flag can be set for frames which need to be transmitted during off-channel operation, allowing off-channel frames to be passed down to the driver if the queues have only been stopped for off-channel. Add ieee80211_tx_skb_offchannel() for transmitting off-channel frames with this flag set. Signed-off-by: Seth Forshee <seth.forshee@canonical.com> --- net/mac80211/agg-tx.c | 4 ++-- net/mac80211/cfg.c | 2 +- net/mac80211/ht.c | 2 +- net/mac80211/ieee80211_i.h | 25 +++++++++++++++++------- net/mac80211/mlme.c | 2 +- net/mac80211/offchannel.c | 32 +++++++++++-------------------- net/mac80211/sta_info.c | 2 +- net/mac80211/tx.c | 45 +++++++++++++++++++++++++++++--------------- net/mac80211/util.c | 3 ++- 9 files changed, 67 insertions(+), 50 deletions(-)