- 13 Apr, 2011 40 commits
-
-
cozybit Inc authored
It is unnecessary to hold the path table resize lock while allocating a new table. Allocate first and take lock later. This resolves a soft-lockup: [ 293.385799] BUG: soft lockup - CPU#0 stuck for 61s! [kworker/u:3:744] (...) [ 293.386049] Call Trace: [ 293.386049] [<c119fd04>] do_raw_read_lock+0x26/0x29 [ 293.386049] [<c14b2982>] _raw_read_lock+0x8/0xa [ 293.386049] [<c148c178>] mesh_path_add+0xb7/0x24e [ 293.386049] [<c148b98d>] ? mesh_path_lookup+0x1b/0xa6 [ 293.386049] [<c148ded5>] hwmp_route_info_get+0x276/0x2fd [ 293.386049] [<c148dfb6>] mesh_rx_path_sel_frame+0x5a/0x5d9 [ 293.386049] [<c102667d>] ? update_curr+0x1cf/0x1d7 [ 293.386049] [<c148b45a>] ieee80211_mesh_rx_queued_mgmt+0x60/0x67 [ 293.386049] [<c147c374>] ieee80211_iface_work+0x1f0/0x258 (...) Signed-off-by:
Javier Cardona <javier@cozybit.com> Signed-off-by:
John W. Linville <linville@tuxdriver.com>
-
Sujith Manoharan authored
Signed-off-by:
Sujith Manoharan <Sujith.Manoharan@atheros.com> Signed-off-by:
-
Sujith Manoharan authored
New debugfs files: <debugfs_root>/ath9k_htc/<phy#>/tgt_int_stats <debugfs_root>/ath9k_htc/<phy#>/tgt_tx_stats <debugfs_root>/ath9k_htc/<phy#>/tgt_rx_stats Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
Beacon transmission needs to involve as little latency as possible after receiving a SWBA event from the target. Since packets are buffered to use TX stream mode, beacon frames sometimes gets queued up and are not sent out immediately. Fix this by decoupling management frame transmission from the normal data path and send them out immediately. Signed-off-by:
-
Sujith Manoharan authored
Occasionally, a WMI event would arrive ahead of the TX URB completion handler. Discarding these events would exhaust the available TX slots, so handle them by running a timer cleaning up such events. Also, timeout packets for which TX completion events have not arrived. Signed-off-by:
-
Sujith Manoharan authored
Location: ath9k_htc/phy#/queue Signed-off-by:
-
Sujith Manoharan authored
Location: ath9k_htc/phy#/slot Signed-off-by:
-
Sujith Manoharan authored
Now that the infrastructure is in place, process WMI TX status events and complete packets. Signed-off-by:
-
Sujith Manoharan authored
When a station entry is removed, there could still be pending packets destined for that station in the HIF layer. Sending these to the target is not necessary, so drain them in the driver itself. Signed-off-by:
-
Sujith Manoharan authored
There is no point in looping over all the endpoints, since the HIF layer uses the start/stop APIs only for the TX pipe. Simplify the API accordingly. Signed-off-by:
-
Sujith Manoharan authored
When doing a channel set or a reset operation the pending frames queued up for transmission have to be flushed and sent to mac80211. Fixing this has to be done in two separate steps: * Flush queued frames and kill the URB TX completion handler. * Complete all the frames that in the TX pending queue. This patch adds proper support for draining and all the callsites namely, channel change/reset/idle/stop are fixed. A separate queue is used for handling failed frames. Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
Maintain a bitmap of slots for transmission and update the cookie field for every packet with the slot value. This value would be used for matching packets when TX completion processing is added. Signed-off-by:
-
Sujith Manoharan authored
There is no need to do endpoint header removal in the ISR. Also, this is needed when TX slot management is added later on. Use a helper function to strip the driver header. Signed-off-by:
-
Sujith Manoharan authored
A new routine that takes an endpoint explicitly is introduced. The normal htc_send() now retrieves the endpoint from the packet's private data. This would be useful in TX completion when the endpoint ID would be required. While at it, use a helper function to map the queue to endpoint. Data/mgmt/beacon packets use htc_send(), while WMI comamnds pass the endpoint to HTC. Signed-off-by:
-
Sujith Manoharan authored
Handle queue start/stop properly by maintaining a counter to check if the pending frame count has exceeded the threshold. Otherwise, packets would be dropped needlessly. While at it, use a simple flag to track queue status and use helper functions too. Signed-off-by:
-
Sujith Manoharan authored
Using a single URB for receiving WMI events is insufficient, increase it to 64 to not lose WMI events in high throughput situations. Signed-off-by:
-
Sujith Manoharan authored
This event will be generated by the target for packet completions. Signed-off-by:
-
Sujith Manoharan authored
Add a new cookie field that would be filled by the host. This can be used to match the TX status WMI event with the appropriate packet. Signed-off-by:
-
Sujith Manoharan authored
The current max queue length of 1024 is quite large and unnecessary. 256 suffices well enough even for high throughput situations. Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
For all packets sent through the USB_WLAN_TX_PIPE endpoint, the private area of the SKB's tx_info can be used to store driver-specific information. For packets sent through USB_REG_OUT_PIPE, this will not make a difference since they are routed through a separate routine that doesn't access the private region. This would help in situations where TX information is required in the URB callback. Signed-off-by:
-
Sujith Manoharan authored
WMI_TGT_TXQ_ENABLE_CMDID WMI_HOST_ATTACH WMI_DEBUG_INFO_CMDID WMI_BEACON_UPDATE_CMDID WMI_RESET_CMDID WMI_RX_LINK_CMDID WMI_STOP_DMA_RECV_CMDID Signed-off-by:
-
Sujith Manoharan authored
The length of the received SKB could be equal to HTC_RX_FRAME_HEADER_SIZE in case of packets with phy/crc errors, in which case they are dropped without being processed. Fix this check so that the error counters are updated correctly. Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
Signed-off-by:
-
Sujith Manoharan authored
Use a queue to handle WMI events and schedule a tasklet to process the events. This fixes the race between the WMI event ISR and the SWBA tasklet when the arrival of WMI events in quick succession could overwrite the SWBA data before the tasklet from a previous iteration could have been scheduled. Also, drain the WMI queue properly. Signed-off-by:
-
Sujith Manoharan authored
Transmission of beacons becomes erratic when TX load is high, since the latency involved in the generation of a SWBA interrupt on the target to the actual sending of a beacon is quite high for USB devices. Fix this by adjusting the beacon response time. Signed-off-by:
-
Sujith Manoharan authored
Use the CAB endpoint to send buffered multicast or broadcast frames after each SWBA event. Signed-off-by:
-
Sujith Manoharan authored
Set operating parameters (cwmin, cwmax) for the beacon queue in AP mode. Signed-off-by:
-
Sujith Manoharan authored
In multi-interface mode, beacons/probe responses that are sent out must have their timestamp field updated. Calculate the TSF adjustment value for each beaconing interface and set it in the frame properly. Signed-off-by:
-
Sujith Manoharan authored
Beacon transmission is now handled through a slot mechanism. This allows multiple beaconing interfaces to be be present. Signed-off-by:
-
Sujith Manoharan authored
Match the beacon header with that of the firmware. Also, the firmware reports the TSF for an SWBA, so store it. Signed-off-by:
-
Sujith Manoharan authored
Also, update the wiphy information and use the correct device pointer when registering. This would fix ethtool. Signed-off-by:
-
Sujith Manoharan authored
Since the new FW requires backward incompatible host driver changes, rename the FW to allow older driver versions to work with the older FW. Signed-off-by:
-
