summaryrefslogtreecommitdiff
path: root/drivers/net/wireless/ath/ath9k/xmit.c
AgeCommit message (Collapse)AuthorFilesLines
2013-09-26ath9k: add txq locking for ath_tx_aggr_startFelix Fietkau1-0/+6
Prevents race conditions when un-aggregated frames are pending in the driver. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-09-26ath9k: don't use BAW tracking on PS responses for non-AMPDU packetsFelix Fietkau1-2/+4
When .release_buffered_frames was implemented, only A-MPDU packets were buffered internally. Now that this has changed, the BUF_AMPDU flag needs to be checked before calling ath_tx_addto_baw Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-09-26ath9k: do not link bf_next across multiple A-MPDUsFelix Fietkau1-1/+3
This might trip up tx completion processing, although the condition that triggers this should not (yet) occur in practice. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-09-26ath9k: fix stale flag handling on buffer cloneFelix Fietkau1-0/+1
Fixes a regression from commit "ath9k: shrink a few data structures by reordering fields" When cloning a buffer, the stale flag (part of bf_state now) needs to be reset after copying the state to prevent tx processing hangs. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-16ath9k: enable CSA functionality in ath9kSimon Wunderlich1-0/+2
CSA is only enabled for one interface, but the same limitation applies for mac80211 too. It checks whether the beacon has been sent (different approaches for non-EDMA-enabled and EDMA-enabled devices), and completes the channel switch after that. Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de> Signed-off-by: Mathias Kretschmer <mathias.kretschmer@fokus.fraunhofer.de> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-16ath9k: reset buffer stale flag in ath_tx_get_tid_subframeFelix Fietkau1-0/+2
If that flag stays set for a buffer that already ran through the tx path once, it might cause issues in tx completion processing. Better clear it early to ensure that this does not happen Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-16ath9k: simplify ath_tid_drainFelix Fietkau1-12/+0
ath_tid_drain is only called when a station entry is being removed, so there is no point in still tracking BAW state. Remove some unnecessary code and a bogus TODO comment related to this. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-16ath9k: shrink a few data structures by reordering fieldsFelix Fietkau1-8/+8
Also reduce the size of a few fields where possible Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: use software queueing for multicast trafficFelix Fietkau1-2/+10
Create a per-vif dummy node entry for keeping the multicast software queues. This helps in setups with a lot of mulitcast traffic that could otherwise potentially drown out unicast traffic to stations. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: improve tx scheduling fairnessFelix Fietkau1-39/+50
Instead of trying to schedule the same TID multiple times in a loop, iterate over other TIDs/stations first. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: use software queues for un-aggregated data packetsFelix Fietkau1-121/+143
This is a first step for improving fairness between legacy and 802.11n traffic, and it should also improve reliability of resets and channel changes by keeping the hardware queue depth very short. When an aggregation session is torn down, all packets in the retry queue will be removed from the BAW and freed. For all subframes that have not been transmitted yet, the A-MPDU flag will be cleared, and a sequence number allocated. This ensures that the next A-MPDU session will get the correct initial sequence number. This happens both on aggregation session start and stop. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: always clear ps filter bit on new assocFelix Fietkau1-0/+1
Otherwise in some cases, EAPOL frames might be filtered during the initial handshake, causing delays and assoc failures. Cc: stable@vger.kernel.org Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: fix clearing expired A-MPDU subframes in tx completionFelix Fietkau1-1/+2
When the tid aggregation state has been marked as inactive, free completed tx packets immediately. When a new aggregation session has not been initialized yet, the BAW checks do not recognize it as expired. Might fix potential stalls in setting up a new aggregation session. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: prepare queueing code for handling unaggregated trafficFelix Fietkau1-6/+14
- Allow ath_tx_get_tid_subframe to return non-AMPDU subframes. - Reset the tid paused state on aggregation stop - Initialize software queues even when HT is not supported Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: fix block ack window tracking checkFelix Fietkau1-6/+9
When a packet has been tracked as part of the BlockAck window and added to the hardware queue, it can end up back in the TID queue again with fi->retries still set to 0 (e.g. if the frame was filtered). Keep an extra bit for the BAW tracking status to fix this corner case. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: simplify ath_tx_form_aggrFelix Fietkau1-21/+12
The check for ATH_AMPDU_SUBFRAME_DEFAULT is unnecessary, since it's set to half the maximum BlockAck Window size, which is already the maximum value that h_baw could possibly have. Also remove unnecessary variables. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: add function for getting the tx tid for a packetFelix Fietkau1-8/+17
Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: split tid retry packets into a separate queueFelix Fietkau1-4/+14
Improves packet retry order and helps with further tx queueing improvements. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-08-09ath9k: add utility functions for accessing tid queuesFelix Fietkau1-19/+32
Useful for further fixes / cleanups Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-07-26Merge branch 'master' of ↵John W. Linville1-16/+28
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless
2013-07-23ath9k: Fix RTS thresholdSujith Manoharan1-4/+32
Currently, RTS threshold is not handled for HT. Handle user-specified threshold values for both aggregated and unaggregated frames. Use the wiphy's threshold parameter for now, it can be made per-VIF later on. Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-07-17ath9k: fix tx pending frames accounting for dropped packetsFelix Fietkau1-16/+28
When dropping packets that have gone far enough into the tx path, the pending frame counter needs to be decreased. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-06-26Merge branch 'master' of ↵John W. Linville1-1/+5
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless
2013-06-24ath9k: fix queue depth accounting in ath_tx_txqaddbufFelix Fietkau1-3/+7
ath_tx_txqaddbuf assumes that all the linked buffers in the queue passed to it are part of the same A-MPDU or MPDU. The CAB queue rework violates this assumption, which can cause the internal queue depth to go negative. Fix this by increasing the counter for all slots of [bf, bf->bf_lastbf] Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-06-24ath9k: fix an RCU issue in calling ieee80211_get_tx_ratesFelix Fietkau1-1/+5
ath_txq_schedule is called outside of the drv_tx call, so it needs RCU protection. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-06-12ath9k: limit multicast buffer hardware queue depthFelix Fietkau1-15/+100
The CAB (Content after Beacon) queue is used for beacon-triggered transmission of buffered multicast frames. If lots of multicast frames were buffered and this queue fills up, it drowns out all regular traffic. To limit the damage that buffered traffic can do, try to limit the queued data to becaon_interval / 8. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-06-12ath9k: implement support for .release_buffered_frames()Felix Fietkau1-36/+139
This adds support for PS-Poll and U-APSD driver-buffered frames (part of an aggregation session). Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-06-12ath9k: add support for IEEE80211_TX_CTL_PS_RESPONSEFelix Fietkau1-16/+27
Use the UAPSD hardware queue to get PS-Poll responses out as fast as possible and without backoff. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-05-22ath9k: prevent aggregation session deadlocksFelix Fietkau1-99/+39
Waiting for all subframes of an existing aggregation session to drain before allowing mac80211 to start a new one is fragile and deadlocks caused by this behavior have been observed. Since mac80211 has proper synchronization for aggregation session start/stop handling, a better approach to session handling is to simply allow mac80211 to start a new session at any time. This requires changing the code to discard any packets outside of the BlockAck window in the A-MPDU software retry code. This patch implements the above and also simplifies the code. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-05-17ath9k: fix draining aggregation tid buffersFelix Fietkau1-3/+1
After a tx attempt, an A-MPDU subframe can still have fi->retries at 0 (if the retry count wasn't incremented due to powersave). In that case it is still tracked as part of the block ack window, so when draining the tid queue, its sequence number needs to be cleared from the pending frame bitmap. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-05-17ath9k: fix rate handling/reportingFelix Fietkau1-2/+7
This patch fixes some issues introduced in the rate control API rework. When not running aggregation, copy bf->rates into info->control.rates before applying the rate control status to it. In ath_lookup_rate, the rates need to be pulled from bf->rates, not the tx info. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-05-17ath9k: fix aggregation stop/flush handlingFelix Fietkau1-21/+40
When aggregation stop is requested, don't run the mac80211 aggregation stop callback yet, while the session is still blocked. Also, when aggregation flush is requested, don't run the callback at all. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-04-23ath9k: add support for the new rate control APIFelix Fietkau1-7/+16
Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-04-23ath9k: merge ath_tx_start_dma into ath_tx_startFelix Fietkau1-44/+36
The split makes no sense and merging the functions makes further changes easier to implement Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-04-10ath9k: implement buffer holding handling for EDMA FIFOFelix Fietkau1-11/+19
Inside one FIFO slot queue, EDMA chipsets have the same link pointer re-read race condition as older chipsets, so the same buffer holding logic needs to be used in order to avoid use-after-free bugs. Unlike on older chips, it can be skipped for the end of the queue. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-04-10ath9k_common: remove ath9k_cmn_padposFelix Fietkau1-2/+2
It is equivalent to ieee80211_hdrlen Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-02-12ath9k: Fix ATH9K_HW_CAP_HT usageSujith Manoharan1-3/+2
There are a few places where the station's HT capabilities should be checked instead of ATH9K_HW_CAP_HT, which is a global feature for the driver. Fix this. Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-01-23ath9k: use ath_tx_process_buffer instead of open-coding similar codeFelix Fietkau1-37/+32
Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-01-23ath9k: clean up processing of pending tx frames on resetFelix Fietkau1-33/+10
Dropping packets from aggregation sessions is usually not a good idea, as it might upset the synchronization of the BlockAck receive window of the remote node. The use of the retry_tx parameter to reset/tx-drain functions also seemed a bit arbitrary. This patch removes this parameter altogether and ensures that pending tx frames are not dropped for no good reason. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2013-01-08ath9k: use the devres API for allocations/mappingsFelix Fietkau1-32/+5
Signed-off-by: Felix Fietkau <nbd@openwrt.org> Acked-by: Luis R. Rodriguez <mcgrof@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-11-30ath9k: Remove redundant NULL assignmentSujith Manoharan1-1/+0
'bf_next' is cleared using ATH_TXBUF_RESET() in both the callsites of ath_tx_get_buffer(). Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-11-30ath9k: Replace WME_NUM_TID with IEEE80211_NUM_TIDSSujith Manoharan1-4/+4
Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-11-21ath9k: Fix the 'xmit' debugfs fileSujith Manoharan1-0/+2
The 'xmit' debugfs file has become big and unwieldy, fix multiple issues with its usage: * Store TX counters/statistics only for the 4 Access Categories. Use IEEE80211_NUM_ACS instead of ATH9K_NUM_TX_QUEUES. * Move various utility macros to debug.h, they can be reused elsewhere. * Remove tx_complete_poll_work_seen. * Remove code that accesses various internal queue-specific variables without any locking whatsoever. HW/SW queue details will be handled in a subsequent patch. * Do not print internal values like txq_headidx and txq_headidx. They were mostly unused anyway, considering code like: PRX("txq_tailidx: ", txq_headidx); * Handle 'txprocdesc' for EDMA too. Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-11-21ath9k/ath9k_htc: Remove WME macrosSujith Manoharan1-5/+5
Use the macros provided by mac80211 and remove redundant declarations inside the drivers. Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-10-30ath9k: Test for TID only in BlockAcks while checking tx statusSven Eckelmann1-2/+6
The ath9k xmit functions for AMPDUs can send frames as non-aggregate in case only one frame is currently available. The client will then answer using a normal Ack instead of a BlockAck. This acknowledgement has no TID stored and therefore the hardware is not able to provide us the corresponding TID. The TID set by the hardware in the tx status descriptor has to be seen as undefined and not as a valid TID value for normal acknowledgements. Doing otherwise results in a massive amount of retransmissions and stalls of connections. Users may experience low bandwidth and complete connection stalls in environments with transfers using multiple TIDs. This regression was introduced in b11b160defc48e4daa283f785192ea3a23a51f8e ("ath9k: validate the TID in the tx status information"). Signed-off-by: Sven Eckelmann <sven@narfation.org> Signed-off-by: Simon Wunderlich <siwu@hrz.tu-chemnitz.de> Cc: stable@vger.kernel.org Acked-by: Felix Fietkau <nbd@openwrt.org> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-10-29ath9k: fix stale pointers potentially causing access to free'd skbsFelix Fietkau1-0/+2
bf->bf_next is only while buffers are chained as part of an A-MPDU in the tx queue. When a tid queue is flushed (e.g. on tearing down an aggregation session), frames can be enqueued again as normal transmission, without bf_next being cleared. This can lead to the old pointer being dereferenced again later. This patch might fix crashes and "Failed to stop TX DMA!" messages. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Cc: stable@vger.kernel.org Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-10-06ath9k: use ieee80211_free_txskbFelix Fietkau1-24/+29
Using ieee80211_free_txskb for tx frames is required, since mac80211 clones skbs for which socket tx status is requested. Signed-off-by: Felix Fietkau <nbd@openwrt.org> Cc: stable@vger.kernel.org Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-09-28ath9k: Fix chainmask selection for AR9462Sujith Manoharan1-0/+4
Force chain 1 to be used for CCK rates since the target power table stored in EEPROM is too high to transmit with both chains. This is needed to avoid regulatory violation. Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
2012-09-07Merge branch 'master' of ↵John W. Linville1-0/+3
git://git.kernel.org/pub/scm/linux/kernel/git/linville/wireless
2012-09-07ath9k: Fix TX filter usageSujith Manoharan1-1/+1
The TX filter bit for a station would be set by the HW when a frame is not acked. A frame would be completed with ATH9K_TXERR_FILT status only when the corresponding filter bit for the destination station is already set. Currently, un-acknowledged packets are added to the pending queue and retried, but the "clear_dest_mask" bit in the descriptor is set only when the TX status has been ATH9K_TXERR_FILT. This results in packet loss and the log shows: wlan0: dropped TX filtered frame, queue_len=0 PS=0 @4309746071 wlan0: dropped TX filtered frame, queue_len=0 PS=0 @4309746076 wlan0: dropped TX filtered frame, queue_len=0 PS=0 @4309746377 ... ... This issue can be resolved by making sure that the destination mask is cleared when the packet is being retried and the earlier TX status is ATH9K_TXERR_XRETRY. Signed-off-by: Sujith Manoharan <c_manoha@qca.qualcomm.com> Signed-off-by: John W. Linville <linville@tuxdriver.com>
generated by cgit v1.2.3 (git 2.39.0) at 2025-09-03 06:32:25 +0300