Skip to contentSkip to conversation

mac80211: move mesh sync beacon handler into neighbour_update

PATCH Superseded Marco Porsch 3 participants 15 messages Feb 15, 2013 – Feb 18, 2013

Subsystems (2)

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Move the beacon handler into mesh_neighbour_update where the STA pointer is already available. This avoids additional overhead and simplifies the handler.
The repositioning will also benefit mesh PS which uses the TSF and the just updated T_offset value.

Rename the handler to better reflect its purpose.

---

changes since RFC:
- rebased on todays mac80211-next/master

 net/mac80211/ieee80211_i.h |   10 ++++-----
 net/mac80211/mesh.c        |    7 +------
 net/mac80211/mesh.h        |    5 +++--
 net/mac80211/mesh_plink.c  |   16 ++++++++++++---
 net/mac80211/mesh_sync.c   |   48 ++++++++++++++++----------------------------
 5 files changed, 39 insertions(+), 47 deletions(-)
net/mac80211/ieee80211_i.h +5 -5
index 388580a..62eeddf 100644
@@ -529,11 +529,11 @@ struct ieee80211_if_ibss {
  */
 struct ieee802_11_elems;
 struct ieee80211_mesh_sync_ops {
-	void (*rx_bcn_presp)(struct ieee80211_sub_if_data *sdata,
-			     u16 stype,
-			     struct ieee80211_mgmt *mgmt,
-			     struct ieee802_11_elems *elems,
-			     struct ieee80211_rx_status *rx_status);
+	void (*rx_bcn)(struct sta_info *sta,
+		       struct ieee80211_mgmt *mgmt,
+		       struct ieee802_11_elems *elems,
+		       struct ieee80211_rx_status *rx_status,
+		       u64 tsf);
 	void (*adjust_tbtt)(struct ieee80211_sub_if_data *sdata);
 	/* add other framework functions here */
 };
net/mac80211/mesh.c +1 -6
index a77d40e..4b42237 100644
@@ -932,7 +932,6 @@ static void ieee80211_mesh_rx_bcn_presp(struct ieee80211_sub_if_data *sdata,
 					struct ieee80211_rx_status *rx_status)
 {
 	struct ieee80211_local *local = sdata->local;
-	struct ieee80211_if_mesh *ifmsh = &sdata->u.mesh;
 	struct ieee802_11_elems elems;
 	struct ieee80211_channel *channel;
 	size_t baselen;
@@ -968,11 +967,7 @@ static void ieee80211_mesh_rx_bcn_presp(struct ieee80211_sub_if_data *sdata,
 		return;
 
 	if (mesh_matches_local(sdata, &elems))
-		mesh_neighbour_update(sdata, mgmt->sa, &elems);
-
-	if (ifmsh->sync_ops)
-		ifmsh->sync_ops->rx_bcn_presp(sdata,
-			stype, mgmt, &elems, rx_status);
+		mesh_neighbour_update(sdata, mgmt, &elems, rx_status);
 }
 
 static void ieee80211_mesh_rx_mgmt_action(struct ieee80211_sub_if_data *sdata,
net/mac80211/mesh.h +3 -2
index 1a1da87..3fd4657 100644
@@ -286,8 +286,9 @@ int mesh_path_send_to_gates(struct mesh_path *mpath);
 int mesh_gate_num(struct ieee80211_sub_if_data *sdata);
 /* Mesh plinks */
 void mesh_neighbour_update(struct ieee80211_sub_if_data *sdata,
-			   u8 *hw_addr,
-			   struct ieee802_11_elems *ie);
+			   struct ieee80211_mgmt *mgmt,
+			   struct ieee802_11_elems *ie,
+			   struct ieee80211_rx_status *rx_status);
 bool mesh_peer_accepts_plinks(struct ieee802_11_elems *ie);
 u32 mesh_accept_plinks_update(struct ieee80211_sub_if_data *sdata);
 void mesh_plink_broken(struct sta_info *sta);
net/mac80211/mesh_plink.c +13 -3
index f7526e5..fe6fc22 100644
@@ -12,6 +12,7 @@
 #include "ieee80211_i.h"
 #include "rate.h"
 #include "mesh.h"
+#include "driver-ops.h"
 
 #define PLINK_GET_LLID(p) (p + 2)
 #define PLINK_GET_PLID(p) (p + 4)
@@ -488,13 +489,19 @@ mesh_sta_info_get(struct ieee80211_sub_if_data *sdata,
  * Initiates peering if appropriate.
  */
 void mesh_neighbour_update(struct ieee80211_sub_if_data *sdata,
-			   u8 *hw_addr,
-			   struct ieee802_11_elems *elems)
+			   struct ieee80211_mgmt *mgmt,
+			   struct ieee802_11_elems *elems,
+			   struct ieee80211_rx_status *rx_status)
 {
+	struct ieee80211_if_mesh *ifmsh = &sdata->u.mesh;
 	struct sta_info *sta;
 	u32 changed = 0;
+	u64 tsf;
 
-	sta = mesh_sta_info_get(sdata, hw_addr, elems);
+	/* get tsf before entering rcu-read section */
+	tsf = drv_get_tsf(sdata->local, sdata);
+
+	sta = mesh_sta_info_get(sdata, mgmt->sa, elems);
 	if (!sta)
 		goto out;
 
@@ -505,6 +512,9 @@ void mesh_neighbour_update(struct ieee80211_sub_if_data *sdata,
 	    rssi_threshold_check(sta, sdata))
 		changed = mesh_plink_open(sta);
 
+	if (ifmsh->sync_ops)
+		ifmsh->sync_ops->rx_bcn(sta, mgmt, elems, rx_status, tsf);
+
 	ieee80211_mps_frame_release(sta, elems);
 out:
 	rcu_read_unlock();
net/mac80211/mesh_sync.c +17 -31

93 lines not on this page · raw message

Signature · 1 line
1.7.9.5
#1 lore raw

PATCH 2/3 mac80211: mesh power save doze scheduling

Configure the HW for PS mode if the local mesh PS parameters allow so.

Expose a callback ieee80211_mps_init for drivers to register mesh powersave ops:
- hw_doze - put the radio to sleep now, wake up at given TBTT
- hw_wakeup - wake the radio up for frame RX
These ops may be extended in the future to allow drivers/HW to implement mesh PS themselves. (The current design goal was to concentrate most mesh PS routines in mac80211 to keep driver modifications minimal.

Track the beacon timing information of peers we are in PS mode towards. Calculate the next TBTT per STA. Stay awake to receive multicast traffic after DTIM beacons.

When going to doze state, get the most imminent STA TBTT and configure the driver to trigger a wakeup on time to catch that beacon. After successful receipt put the HW to doze again. Set a timeout for the case that the beacon is not received on time. In this case calculate the following TBTT and go to doze again.

For mesh Awake Windows wakeup on SWBA (beacon_get_tim) and start a timer which triggers a doze call on expiry.

---

changes since RFC:
- commit logs
- resume doze after multicast (CAB) receipt
- timer_pending as check if beacons received once
- check TIM pointer before use
- TODO comment for using ieee80211_beacon_get_tim as time reference (Johannes)
- spin_lock_bh instead of rcu_lock in ieee80211_mps_sta_tbtt_timeout

 include/net/mac80211.h     |   29 ++++
 net/mac80211/ieee80211_i.h |    7 +-
 net/mac80211/mesh.c        |   11 ++
 net/mac80211/mesh.h        |   13 ++
 net/mac80211/mesh_plink.c  |    3 +
 net/mac80211/mesh_ps.c     |  313 ++++++++++++++++++++++++++++++++++++++++++++
 net/mac80211/sta_info.h    |   10 ++
 net/mac80211/tx.c          |    2 +
 8 files changed, 387 insertions(+), 1 deletion(-)
include/net/mac80211.h +29 -0
index f7eba13..a27383f 100644
@@ -2801,6 +2801,18 @@ enum ieee80211_tpt_led_trigger_flags {
 	IEEE80211_TPT_LEDTRIG_FL_CONNECTED	= BIT(2),
 };
 
+/**
+ * struct ieee80211_mps_ops - callbacks from mac80211 to the driver for mesh PS
+ *
+ * @hw_doze: put the radio to doze state to conserve power, schedule wakeup
+ *	at given TSF value to receive peer beacon
+ * @hw_wakeup: wake up the radio to receive frames again
+ */
+struct ieee80211_mps_ops {
+	void (*hw_doze)(struct ieee80211_hw *hw, u64 nexttbtt);
+	void (*hw_wakeup)(struct ieee80211_hw *hw);
+};
+
 #ifdef CONFIG_MAC80211_LEDS
 extern char *__ieee80211_get_tx_led_name(struct ieee80211_hw *hw);
 extern char *__ieee80211_get_rx_led_name(struct ieee80211_hw *hw);
@@ -4058,6 +4070,23 @@ void ieee80211_stop_rx_ba_session(struct ieee80211_vif *vif, u16 ba_rx_bitmap,
  */
 void ieee80211_send_bar(struct ieee80211_vif *vif, u8 *ra, u16 tid, u16 ssn);
 
+/**
+ * ieee80211_mps_init - register callbacks for mesh powersave mode
+ *
+ * @hw: the hardware
+ * @ops: callbacks for this device
+ *
+ * called by driver on mesh interface add/remove
+ */
+#ifdef CONFIG_MAC80211_MESH
+void ieee80211_mps_init(struct ieee80211_hw *hw,
+			const struct ieee80211_mps_ops *ops);
+#else
+static inline void ieee80211_mps_init(struct ieee80211_hw *hw,
+				      const struct ieee80211_mps_ops *ops)
+{ return; }
+#endif
+
 /* Rate control API */
 
 /**
net/mac80211/ieee80211_i.h +6 -1
index 62eeddf..fb55391 100644
@@ -600,6 +600,7 @@ struct ieee80211_if_mesh {
 	int ps_peers_light_sleep;
 	int ps_peers_deep_sleep;
 	struct ps_data ps;
+	struct timer_list awake_window_end_timer;
 };
 
 #ifdef CONFIG_MAC80211_MESH
@@ -1093,7 +1094,7 @@ struct ieee80211_local {
 	bool pspolling;
 	bool offchannel_ps_enabled;
 	/*
-	 * PS can only be enabled when we have exactly one managed
+	 * managed mode PS can only be enabled when we have exactly one managed
 	 * interface (and monitors) in PS, this then points there.
 	 */
 	struct ieee80211_sub_if_data *ps_sdata;
@@ -1112,6 +1113,10 @@ struct ieee80211_local {
 
 	int user_power_level; /* in dBm, for all interfaces */
 
+	/* mesh power save can be enabled for multiple (but only mesh) vif */
+	bool mps_enabled;
+	const struct ieee80211_mps_ops *mps_ops;
+
 	enum ieee80211_smps_mode smps_mode;
 
 	struct work_struct restart_work;
net/mac80211/mesh.c +11 -0
index 4b42237..f4af64e 100644
@@ -168,6 +168,7 @@ void mesh_sta_cleanup(struct sta_info *sta)
 	if (sdata->u.mesh.security == IEEE80211_MESH_SEC_NONE) {
 		changed |= mesh_plink_deactivate(sta);
 		del_timer_sync(&sta->plink_timer);
+		del_timer_sync(&sta->nexttbtt_timer);
 	}
 
 	if (changed)
@@ -1020,6 +1021,7 @@ void ieee80211_mesh_rx_queued_mgmt(struct ieee80211_sub_if_data *sdata,
 
 void ieee80211_mesh_work(struct ieee80211_sub_if_data *sdata)
 {
+	struct ieee80211_local *local = sdata->local;
 	struct ieee80211_if_mesh *ifmsh = &sdata->u.mesh;
 
 	if (ifmsh->preq_queue_len &&
@@ -1041,6 +1043,12 @@ void ieee80211_mesh_work(struct ieee80211_sub_if_data *sdata)
 
 	if (test_and_clear_bit(MESH_WORK_DRIFT_ADJUST, &ifmsh->wrkq_flags))
 		mesh_sync_adjust_tbtt(sdata);
+
+	if (test_and_clear_bit(MESH_WORK_PS_HW_CONF, &ifmsh->wrkq_flags))
+		ieee80211_mps_hw_conf(local);
+
+	if (test_and_clear_bit(MESH_WORK_PS_DOZE, &ifmsh->wrkq_flags))
+		ieee80211_mps_doze(local);
 }
 
 void ieee80211_mesh_notify_scan_completed(struct ieee80211_local *local)
@@ -1080,6 +1088,9 @@ void ieee80211_mesh_init_sdata(struct ieee80211_sub_if_data *sdata)
 	setup_timer(&ifmsh->mesh_path_root_timer,
 		    ieee80211_mesh_path_root_timer,
 		    (unsigned long) sdata);
+	setup_timer(&ifmsh->awake_window_end_timer,
+		    ieee80211_mps_awake_window_end,
+		    (unsigned long) sdata);
 	INIT_LIST_HEAD(&ifmsh->preq_queue.list);
 	skb_queue_head_init(&ifmsh->ps.bc_buf);
 	spin_lock_init(&ifmsh->mesh_preq_queue_lock);
net/mac80211/mesh.h +13 -0
index 3fd4657..87774e5 100644
@@ -58,6 +58,8 @@ enum mesh_path_flags {
  * @MESH_WORK_ROOT: the mesh root station needs to send a frame
  * @MESH_WORK_DRIFT_ADJUST: time to compensate for clock drift relative to other
  * mesh nodes
+ * @MESH_WORK_PS_HW_CONF: perform checks for PS mode and configure hardware
+ * @MESH_WORK_PS_DOZE: perform checks for doze state and put hardware to doze
  */
 enum mesh_deferred_task_flags {
 	MESH_WORK_HOUSEKEEPING,
@@ -65,6 +67,8 @@ enum mesh_deferred_task_flags {
 	MESH_WORK_GROW_MPP_TABLE,
 	MESH_WORK_ROOT,
 	MESH_WORK_DRIFT_ADJUST,
+	MESH_WORK_PS_HW_CONF,
+	MESH_WORK_PS_DOZE,
 };
 
 /**
@@ -261,6 +265,15 @@ void ieee80211_mpsp_trigger_process(u8 *qc, struct sta_info *sta,
 				    bool tx, bool acked);
 void ieee80211_mps_frame_release(struct sta_info *sta,
 				 struct ieee802_11_elems *elems);
+void ieee80211_mps_hw_conf(struct ieee80211_local *local);
+void ieee80211_mps_sta_tbtt_update(struct sta_info *sta,
+				   struct ieee80211_mgmt *mgmt,
+				   const struct ieee80211_tim_ie *tim,
+				   u64 tsf);
+void ieee80211_mps_sta_tbtt_timeout(unsigned long data);
+void ieee80211_mps_awake_window_start(struct ieee80211_sub_if_data *sdata);
+void ieee80211_mps_awake_window_end(unsigned long data);
+void ieee80211_mps_doze(struct ieee80211_local *local);
 
 /* Mesh paths */
 int mesh_nexthop_lookup(struct sk_buff *skb,
net/mac80211/mesh_plink.c +3 -0
index fe6fc22..e1fe2a1 100644
@@ -413,6 +413,8 @@ __mesh_sta_info_alloc(struct ieee80211_sub_if_data *sdata, u8 *hw_addr)
 
 	sta->plink_state = NL80211_PLINK_LISTEN;
 	init_timer(&sta->plink_timer);
+	setup_timer(&sta->nexttbtt_timer, ieee80211_mps_sta_tbtt_timeout,
+		    (unsigned long) sta);
 
 	sta_info_pre_move_state(sta, IEEE80211_STA_AUTH);
 	sta_info_pre_move_state(sta, IEEE80211_STA_ASSOC);
@@ -516,6 +518,7 @@ void mesh_neighbour_update(struct ieee80211_sub_if_data *sdata,
 		ifmsh->sync_ops->rx_bcn(sta, mgmt, elems, rx_status, tsf);
 
 	ieee80211_mps_frame_release(sta, elems);
+	ieee80211_mps_sta_tbtt_update(sta, mgmt, elems->tim, tsf);
 out:
 	rcu_read_unlock();
 	ieee80211_mbss_info_change_notify(sdata, changed);
net/mac80211/mesh_ps.c +313 -0

352 lines not on this page · raw message

net/mac80211/sta_info.h +10 -0
index 63dfdb5..80866b5 100644
@@ -58,6 +58,7 @@
  * @WLAN_STA_TOFFSET_KNOWN: toffset calculated for this station is valid.
  * @WLAN_STA_MPSP_OWNER: local STA is owner of a mesh Peer Service Period.
  * @WLAN_STA_MPSP_RECIPIENT: local STA is recipient of a MPSP.
+ * @WLAN_STA_MPS_WAIT_FOR_CAB: multicast frames from this STA are imminent.
  */
 enum ieee80211_sta_info_flags {
 	WLAN_STA_AUTH,
@@ -82,6 +83,7 @@ enum ieee80211_sta_info_flags {
 	WLAN_STA_TOFFSET_KNOWN,
 	WLAN_STA_MPSP_OWNER,
 	WLAN_STA_MPSP_RECIPIENT,
+	WLAN_STA_MPS_WAIT_FOR_CAB,
 };
 
 #define ADDBA_RESP_INTERVAL HZ
@@ -289,6 +291,10 @@ struct sta_ampdu_mlme {
  * @local_pm: local link-specific power save mode
  * @peer_pm: peer-specific power save mode towards local STA
  * @nonpeer_pm: STA power save mode towards non-peer neighbors
+ * @beacon_interval: beacon interval of neighbor STA (in us)
+ * @nexttbtt_tsf: next TBTT in local TSF units
+ * @nexttbtt_jiffies: next TBTT in jiffies units
+ * @nexttbtt_timer: timeout for missed beacons
  * @debugfs: debug filesystem info
  * @dead: set to true when sta is unlinked
  * @uploaded: set to true when sta is uploaded to the driver
@@ -391,6 +397,10 @@ struct sta_info {
 	enum nl80211_mesh_power_mode local_pm;
 	enum nl80211_mesh_power_mode peer_pm;
 	enum nl80211_mesh_power_mode nonpeer_pm;
+	u32 beacon_interval;
+	u64 nexttbtt_tsf;
+	unsigned long nexttbtt_jiffies;
+	struct timer_list nexttbtt_timer;
 #endif
 
 #ifdef CONFIG_MAC80211_DEBUGFS
net/mac80211/tx.c +2 -0
index fe644f9..f7a4c32 100644
@@ -2464,6 +2464,8 @@ struct sk_buff *ieee80211_beacon_get_tim(struct ieee80211_hw *hw,
 		memcpy(skb_put(skb, bcn->head_len), bcn->head, bcn->head_len);
 		ieee80211_beacon_add_tim(sdata, &ifmsh->ps, skb);
 		memcpy(skb_put(skb, bcn->tail_len), bcn->tail, bcn->tail_len);
+
+		ieee80211_mps_awake_window_start(sdata);
 	} else {
 		WARN_ON(1);
 		goto out;
Signature · 1 line
1.7.9.5
#2 lore raw

PATCH 3/3 ath9k: mesh powersave support

In reply to Marco Porsch

Register mesh PS ops on interface add and de-register on removal.

React to doze/wakeup calls issued by mac80211. Add a PS status flag PS_MAC80211_CTL to store last mesh PS command from mac80211.

Initialize HW beacon wakeup registers.
On doze call configure the HW to wakeup on the given TSF value.

---
 drivers/net/wireless/ath/ath9k/ath9k.h  |    1 +
 drivers/net/wireless/ath/ath9k/beacon.c |   21 ++++++++++-
 drivers/net/wireless/ath/ath9k/hw.c     |   17 +++++++--
 drivers/net/wireless/ath/ath9k/main.c   |   63 ++++++++++++++++++++++++++++++-
 4 files changed, 95 insertions(+), 7 deletions(-)
drivers/net/wireless/ath/ath9k/ath9k.h +1 -0
index 97c90b2..b82727b 100644
@@ -652,6 +652,7 @@ enum sc_op_flags {
 #define PS_WAIT_FOR_TX_ACK        BIT(3)
 #define PS_BEACON_SYNC            BIT(4)
 #define PS_WAIT_FOR_ANI           BIT(5)
+#define PS_MAC80211_CTL           BIT(6)
 
 struct ath_rate_table;
 
drivers/net/wireless/ath/ath9k/beacon.c +20 -1
index dd37719..7ef698b 100644
@@ -599,6 +599,23 @@ static void ath9k_beacon_config_adhoc(struct ath_softc *sc,
 	ath9k_beacon_init(sc, nexttbtt, intval);
 }
 
+static void ath9k_beacon_config_mesh(struct ath_softc *sc,
+				     struct ath_beacon_config *conf)
+{
+	struct ath9k_beacon_state bs;
+
+	/*
+	 * when PS is enabled, ath9k_hw_setrxabort is set.
+	 * to wake up again to receive peers' beacons, we set an
+	 * arbitrary initial value for sleepduration here
+	 */
+	memset(&bs, 0, sizeof(bs));
+	bs.bs_sleepduration = IEEE80211_MS_TO_TU(100);
+	ath9k_hw_set_sta_beacon_timers(sc->sc_ah, &bs);
+
+	ath9k_beacon_config_adhoc(sc, conf);
+}
+
 bool ath9k_allow_beacon_config(struct ath_softc *sc, struct ieee80211_vif *vif)
 {
 	struct ath_common *common = ath9k_hw_common(sc->sc_ah);
@@ -707,9 +724,11 @@ void ath9k_set_beacon(struct ath_softc *sc)
 		ath9k_beacon_config_ap(sc, cur_conf);
 		break;
 	case NL80211_IFTYPE_ADHOC:
-	case NL80211_IFTYPE_MESH_POINT:
 		ath9k_beacon_config_adhoc(sc, cur_conf);
 		break;
+	case NL80211_IFTYPE_MESH_POINT:
+		ath9k_beacon_config_mesh(sc, cur_conf);
+		break;
 	case NL80211_IFTYPE_STATION:
 		ath9k_beacon_config_sta(sc, cur_conf);
 		break;
drivers/net/wireless/ath/ath9k/hw.c +13 -4
index 42cf3c7..0e0fe03 100644
@@ -2254,16 +2254,24 @@ void ath9k_hw_beaconinit(struct ath_hw *ah, u32 next_beacon, u32 beacon_period)
 }
 EXPORT_SYMBOL(ath9k_hw_beaconinit);
 
+/**
+ * ath9k_hw_set_sta_beacon_timers
+ *
+ * in mesh mode overwriting AR_NEXT_TBTT_TIMER and setting AR_TBTT_TIMER_EN
+ * would shift the own TBTT
+ */
 void ath9k_hw_set_sta_beacon_timers(struct ath_hw *ah,
 				    const struct ath9k_beacon_state *bs)
 {
 	u32 nextTbtt, beaconintval, dtimperiod, beacontimeout;
+	u32 ar_timer_mode = AR_DTIM_TIMER_EN | AR_TIM_TIMER_EN;
 	struct ath9k_hw_capabilities *pCap = &ah->caps;
 	struct ath_common *common = ath9k_hw_common(ah);
 
 	ENABLE_REGWRITE_BUFFER(ah);
 
-	REG_WRITE(ah, AR_NEXT_TBTT_TIMER, TU_TO_USEC(bs->bs_nexttbtt));
+	if (ah->opmode != NL80211_IFTYPE_MESH_POINT)
+		REG_WRITE(ah, AR_NEXT_TBTT_TIMER, TU_TO_USEC(bs->bs_nexttbtt));
 
 	REG_WRITE(ah, AR_BEACON_PERIOD,
 		  TU_TO_USEC(bs->bs_intval));
@@ -2317,9 +2325,10 @@ void ath9k_hw_set_sta_beacon_timers(struct ath_hw *ah,
 
 	REGWRITE_BUFFER_FLUSH(ah);
 
-	REG_SET_BIT(ah, AR_TIMER_MODE,
-		    AR_TBTT_TIMER_EN | AR_TIM_TIMER_EN |
-		    AR_DTIM_TIMER_EN);
+	if (ah->opmode != NL80211_IFTYPE_MESH_POINT)
+		ar_timer_mode |= AR_TBTT_TIMER_EN;
+
+	REG_SET_BIT(ah, AR_TIMER_MODE, ar_timer_mode);
 
 	/* TSF Out of Range Threshold */
 	REG_WRITE(ah, AR_TSFOOR_THRESHOLD, bs->bs_tsfoor_threshold);
drivers/net/wireless/ath/ath9k/main.c +61 -2

87 lines not on this page · raw message

Signature · 1 line
1.7.9.5
maintainer
#3 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

In reply to Marco Porsch

net/mac80211/mesh_sync.c void mesh_sync_adjust_tbtt(struct ieee80211_sub_if_data *sdata) · line 99

Quoted 5 lines · Marco Porsch
net/mac80211/mesh_sync.c +0 -5
-	/* The current tsf is a first approximation for the timestamp
-	 * for the received beacon.  Further down we try to get a
-	 * better value from the rx_status->mactime field if
-	 * available. Also we have to call drv_get_tsf() before
-	 * entering the rcu-read section.*/

I don't think you should drop this comment; also why not just address it? There's a timestamp in rx_status that should be the correct one for ath9k (which is pretty much all this seems to work on anyway :) )

johannes

#4 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Quoted 11 lines · Johannes Berg

net/mac80211/mesh_sync.c void mesh_sync_adjust_tbtt(struct ieee80211_sub_if_data *sdata) · line 99

Quoted 5 lines · Marco Porsch
net/mac80211/mesh_sync.c +0 -5
-	/* The current tsf is a first approximation for the timestamp
-	 * for the received beacon.  Further down we try to get a
-	 * better value from the rx_status->mactime field if
-	 * available. Also we have to call drv_get_tsf() before
-	 * entering the rcu-read section.*/

I don't think you should drop this comment; also why not just address it? There's a timestamp in rx_status that should be the correct one for ath9k (which is pretty much all this seems to work on anyway :) )

The comment is just moved? Mesh already uses rx_status->mactime, if the driver supplies it, which the comment says :)

Signature · 1 line
Bob Copeland %% www.bobcopeland.com
#5 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Quoted 6 lines · Bob Copeland
Quoted 3 lines

I don't think you should drop this comment; also why not just address it? There's a timestamp in rx_status that should be the correct one for ath9k (which is pretty much all this seems to work on anyway :) )

The comment is just moved? Mesh already uses rx_status->mactime, if the driver supplies it, which the comment says :)

Oh I see - you are saying not to drop the comment about rcu locking.

Signature · 1 line
Bob Copeland %% www.bobcopeland.com
#6 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Hi,

net/mac80211/mesh_plink.c mesh_sta_info_get(struct ieee80211_sub_if_data

Quoted 15 lines · Bob Copeland
Quoted 6 lines · Bob Copeland
Quoted 3 lines

I don't think you should drop this comment; also why not just address it? There's a timestamp in rx_status that should be the correct one for ath9k (which is pretty much all this seems to work on anyway :) )

The comment is just moved? Mesh already uses rx_status->mactime, if the driver supplies it, which the comment says :)

Oh I see - you are saying not to drop the comment about rcu locking.

net/mac80211/mesh_plink.c +0 -0
index f7526e5..fe6fc22 100644
@@ -488,13 +489,19 @@ mesh_sta_info_get(struct ieee80211_sub_if_data 

*sdata,

Quoted 6 lines
net/mac80211/mesh_plink.c +1 -2
   * Initiates peering if appropriate.
   */
  void mesh_neighbour_update(struct ieee80211_sub_if_data *sdata,
-			   u8 *hw_addr,
-			   struct ieee802_11_elems *elems)
+			   struct ieee80211_mgmt *mgmt,

net/mac80211/mesh_plink.c mesh_sta_info_get(struct ieee80211_sub_if_data *sdata, · line 501

Quoted 10 lines
net/mac80211/mesh_plink.c +5 -1
+			   struct ieee802_11_elems *elems,
+			   struct ieee80211_rx_status *rx_status)
  {
+	struct ieee80211_if_mesh *ifmsh = &sdata->u.mesh;
  	struct sta_info *sta;
  	u32 changed = 0;
+	u64 tsf;
 
-	sta = mesh_sta_info_get(sdata, hw_addr, elems);
+	/* get tsf before entering rcu-read section */

The comment about RCU lock moved here.

net/mac80211/mesh_plink.c mesh_sta_info_get(struct ieee80211_sub_if_data *sdata, · line 507

Quoted 6 lines
net/mac80211/mesh_plink.c +3 -0
+	tsf = drv_get_tsf(sdata->local, sdata);
+
+	sta = mesh_sta_info_get(sdata, mgmt->sa, elems);
  	if (!sta)
  		goto out;
 

--Marco

#7 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

In reply to Johannes Berg

Please check again. The comment is split in two and placed on the respective new positions.

Johannes Berg schrieb am 15.02.13 13:14:

net/mac80211/mesh_sync.c void mesh_sync_adjust_tbtt(struct ieee80211_sub_if_data *sdata) · line 99

Quoted 5 lines · Marco Porsch
net/mac80211/mesh_sync.c +0 -5
-	/* The current tsf is a first approximation for the timestamp
-	 * for the received beacon.  Further down we try to get a
-	 * better value from the rx_status->mactime field if
-	 * available. Also we have to call drv_get_tsf() before
-	 * entering the rcu-read section.*/

I don't think you should drop this comment; also why not just address it? There's a timestamp in rx_status that should be the correct one for ath9k (which is pretty much all this seems to work on anyway :) )

johannes

maintainer
#8 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Quoted 1 line · marco at cozybit.com

Please check again. The comment is split in two and placed on the respective new positions.

Yeah, I see, the API is just total shit. First passing the TSF and then calculating it to override? Why not do the calculation outside the API always?

johannes

#9 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Hi,

Quoted 6 lines · Johannes Berg
Quoted 1 line · marco at cozybit.com

Please check again. The comment is split in two and placed on the respective new positions.

Yeah, I see, the API is just total shit. First passing the TSF and then calculating it to override? Why not do the calculation outside the API always?

The TBTT calculation does intentionally not use the mactime value.

Synchronization uses the time in local TSF units and the exact same time point in peers TSF units at the time of sending/receiving (mactime field). The TBTT calculation uses the time of NOW, i.e. the current TSF after possible delays in firmware/driver/rx-handler which may have outdated the mactime field.

--Marco

maintainer
#10 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Quoted 10 lines · Marco Porsch

Hi,

Quoted 6 lines · Johannes Berg
Quoted 1 line · marco at cozybit.com

Please check again. The comment is split in two and placed on the respective new positions.

Yeah, I see, the API is just total shit. First passing the TSF and then calculating it to override? Why not do the calculation outside the API always?

Quoted 7 lines

The TBTT calculation does intentionally not use the mactime value.

Synchronization uses the time in local TSF units and the exact same time point in peers TSF units at the time of sending/receiving (mactime field). The TBTT calculation uses the time of NOW, i.e. the current TSF after possible delays in firmware/driver/rx-handler which may have outdated the mactime field.

I'm talking about this API:

mesh_neighbour_update:
...

	tsf = drv_get_tsf()
...
	sync_ops->rx_bcn(..., tsf)

mesh_sync_offset_rx_bcn(..., t_r):

	...
	if (have_better_timestamp)
		t_r = get_better_timestamp()

You can hardly claim that's an intuitive API.

johannes

#11 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Quoted 28 lines · Johannes Berg
Quoted 10 lines · Marco Porsch

Hi,

Quoted 6 lines · Johannes Berg
Quoted 1 line · marco at cozybit.com

Please check again. The comment is split in two and placed on the respective new positions.

Yeah, I see, the API is just total shit. First passing the TSF and then calculating it to override? Why not do the calculation outside the API always?

Quoted 7 lines

The TBTT calculation does intentionally not use the mactime value.

Synchronization uses the time in local TSF units and the exact same time point in peers TSF units at the time of sending/receiving (mactime field). The TBTT calculation uses the time of NOW, i.e. the current TSF after possible delays in firmware/driver/rx-handler which may have outdated the mactime field.

I'm talking about this API:

mesh_neighbour_update:
...

	tsf = drv_get_tsf()
...
	sync_ops->rx_bcn(..., tsf)
Quoted 7 lines

mesh_sync_offset_rx_bcn(..., t_r):

	...
	if (have_better_timestamp)
		t_r = get_better_timestamp()

You can hardly claim that's an intuitive API.

Hm, alright. Just saying that ieee80211_mps_sta_tbtt_update still uses the unchanged TSF value. But hey :)

What would be more favourable then?

a) second variable in mesh_sync_offset_rx_bcn

mesh_neighbour_update:
...

	tsf = drv_get_tsf()
...
	sync_ops->rx_bcn(..., tsf)

	ieee80211_mps_sta_tbtt_update(..., tsf);

mesh_sync_offset_rx_bcn(..., tsf):

	...
	if (have_better_timestamp)
		t_r = get_better_timestamp()
	else
		t_r = tsf;

b) second variable in mesh_neighbour_update

mesh_neighbour_update:
...

	tsf = drv_get_tsf()
	if (have_better_timestamp)
		t_r = get_better_timestamp()
	else
		t_r = tsf;
...
	sync_ops->rx_bcn(..., t_r)

	ieee80211_mps_sta_tbtt_update(..., tsf);

--Marco

maintainer
#12 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Quoted 15 lines · Marco Porsch
Quoted 15 lines

I'm talking about this API:

mesh_neighbour_update:
...

	tsf = drv_get_tsf()
...
	sync_ops->rx_bcn(..., tsf)

mesh_sync_offset_rx_bcn(..., t_r):

	...
	if (have_better_timestamp)
		t_r = get_better_timestamp()
Quoted 4 lines
Quoted 1 line

You can hardly claim that's an intuitive API.

Hm, alright. Just saying that ieee80211_mps_sta_tbtt_update still uses the unchanged TSF value. But hey :)

Well, that function doesn't exist in this patch...

Quoted 1 line

What would be more favourable then?

I guess you can tell I'm not in a good mood today. I think any use of get_tsf() for operation is a complete waste of time, there's no way you can get the timings correct. You could be preempted, and suddenly sleep for a few tens or hundreds milliseconds, so none of this makes any sense... To properly do it you have to do calculations in relative times and let the device apply them.

johannes

#13 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Hi,

Quoted 24 lines · Johannes Berg
Quoted 15 lines · Marco Porsch
Quoted 15 lines

I'm talking about this API:

mesh_neighbour_update:
...

	tsf = drv_get_tsf()
...
	sync_ops->rx_bcn(..., tsf)

mesh_sync_offset_rx_bcn(..., t_r):

	...
	if (have_better_timestamp)
		t_r = get_better_timestamp()
Quoted 4 lines
Quoted 1 line

You can hardly claim that's an intuitive API.

Hm, alright. Just saying that ieee80211_mps_sta_tbtt_update still uses the unchanged TSF value. But hey :)

Well, that function doesn't exist in this patch...

Quoted 8 lines
Quoted 1 line

What would be more favourable then?

I guess you can tell I'm not in a good mood today. I think any use of get_tsf() for operation is a complete waste of time, there's no way you can get the timings correct. You could be preempted, and suddenly sleep for a few tens or hundreds milliseconds, so none of this makes any sense... To properly do it you have to do calculations in relative times and let the device apply them.

I don't get your last sentence here. Maybe you can elaborate?

Concerning timestamp vs. TSF usage; wow - I tested it when using the timestamp value for TBTT scheduling. Works fine. Works even better than TSF as it slightly reduces the measured wakeup overhead.

Hm, I was sure the TSF as in *now* would be more appropriate than the TSF at receipt time... But either way, if it works better and is less ugly, it is a win-win =)

I'll send an updated patch with the more intuitive API.

maintainer
#14 lore raw

PATCH 1/3 mac80211: move mesh sync beacon handler into neighbour_update

Quoted 8 lines · Marco Porsch
Quoted 6 lines

I guess you can tell I'm not in a good mood today. I think any use of get_tsf() for operation is a complete waste of time, there's no way you can get the timings correct. You could be preempted, and suddenly sleep for a few tens or hundreds milliseconds, so none of this makes any sense... To properly do it you have to do calculations in relative times and let the device apply them.

I don't get your last sentence here. Maybe you can elaborate?

I'm saying what could happens is this:

tsf = drv_get_tsf()
[be preempted, 100ms later]
do_something_with(tsf)

so I think using drv_get_tsf() is pretty much always wrong.

Quoted 9 lines

Concerning timestamp vs. TSF usage; wow - I tested it when using the timestamp value for TBTT scheduling. Works fine. Works even better than TSF as it slightly reduces the measured wakeup overhead.

Hm, I was sure the TSF as in *now* would be more appropriate than the TSF at receipt time... But either way, if it works better and is less ugly, it is a win-win =)

I'll send an updated patch with the more intuitive API.

:)

johannes