Adrian Chadd 1a85141ad4 Pull out the if_transmit() work and revert back to ath_start(). ... My changed had some rather significant behavioural changes to throughput. The two issues I noticed: * With if_start and the ifnet mbuf queue, any temporary latency would get eaten up by some mbufs being queued. With ath_transmit() queuing things to ath_buf's, I'd only get 512 TX buffers before I couldn't queue any further frames. * There's also some non-zero latency involved with TX being pushed into a taskqueue via direct dispatch. Any time the scheduler didn't immediately schedule the ath TX task would cause extra latency. Various 1ge/10ge drivers implement both direct dispatch (if the TX lock can be acquired) and deferred task transmission (if the TX lock can't be acquired), with frames being pushed into a drbd queue. I'll have to do this at some point, but until I figure out how to deal with 802.11 fragments, I'll have to wait a while longer. So what I saw: * lots of extra latency, specially under load - if the taskqueue wasn't immediately scheduled, things went pear shaped; * any extra latency would result in TX ath_buf's taking their sweet time being replenished, so any further calls to ath_transmit() would drop mbufs. * .. yes, there's no explicit backpressure here - things are just dropped. Eek. With this, the general performance has gone up, but those subtle if_start() related race conditions are back. For some reason, this is doubly-obvious with the AR5416 NIC and I don't quite understand why yet. There's an unrelated issue with AR5416 performance in STA mode (it's fine in AP mode when bridging frames, weirdly..) that requires a little further investigation. Specifically - it works fine on a Lenovo T40 (single core CPU) running a March 2012 9-STABLE kernel, but a Lenovo T60 (dual core) running an early November 2012 kernel behaves very poorly. The same hardware with an AR9160 or AR9280 behaves perfectly.		2013-02-13 05:32:19 +00:00
..
ath_dfs/null	Add a method to explicitly disable radar reporting if required.	2013-01-02 01:36:10 +00:00
ath_hal	The encryption type field needs to be preserved for each descriptor	2013-02-09 02:42:01 +00:00
ath_rate	I've had some feedback that CCK rates are more reliable than MCS 0	2012-10-31 06:35:50 +00:00
ah_osdep.c	Fix the HAL debugging to only use one bit to mark a message as unmaskable.	2012-07-01 02:34:32 +00:00
ah_osdep.h	Migrate the ath_hal_filltxdesc() API to take a list of buffer/seglen values.	2012-08-05 10:12:27 +00:00
if_ath.c	Pull out the if_transmit() work and revert back to ath_start().	2013-02-13 05:32:19 +00:00
if_ath_ahb.c	Create a new TX lock specifically for queuing frames.	2013-02-07 07:50:16 +00:00
if_ath_alq.c	Extend the timestamp to be a timeval, rather than ticks.	2013-02-11 02:48:49 +00:00
if_ath_alq.h	Extend the timestamp to be a timeval, rather than ticks.	2013-02-11 02:48:49 +00:00
if_ath_beacon.c	Add a quick work-around if ath_beacon_config() to not die if it's called	2013-01-17 16:26:40 +00:00
if_ath_beacon.h	Extend the beacon code slightly to support AP mode beaconing for the	2012-08-11 23:26:19 +00:00
if_ath_debug.c	TX EDMA debugging fixes:	2012-11-05 07:08:45 +00:00
if_ath_debug.h	Add a debug method to dump the EDMA TX status descriptor contents out.	2012-11-03 04:53:44 +00:00
if_ath_keycache.c	Don't allocate or program a key for the AR5210.	2012-11-19 23:54:05 +00:00
if_ath_keycache.h	Add a 'vap' to ath_keyset().	2011-11-08 19:25:52 +00:00
if_ath_led.c	Bring over the initial 802.11n bluetooth coexistence support code.	2012-06-26 22:16:53 +00:00
if_ath_led.h	Refactor out the software LED config code into a common function, called	2011-12-26 05:46:22 +00:00
if_ath_misc.h	Pull out the if_transmit() work and revert back to ath_start().	2013-02-13 05:32:19 +00:00
if_ath_pci.c	Create a new TX lock specifically for queuing frames.	2013-02-07 07:50:16 +00:00
if_ath_rx.c	Work around some rather unfortunate race conditions inside net80211.	2013-01-31 00:14:25 +00:00
if_ath_rx.h	Convert sc_rxpending to a per-EDMA queue, and use that for the legacy code.	2012-07-10 00:02:19 +00:00
if_ath_rx_edma.c	Mechanically substitute flags from historic mbuf allocator with	2012-12-04 09:32:43 +00:00
if_ath_rx_edma.h	Begin abstracting out the RX path in preparation for RX EDMA support.	2012-07-03 06:59:12 +00:00
if_ath_spectral.c	Fix format size.	2013-01-08 22:42:15 +00:00
if_ath_spectral.h	Add a new (skeleton) spectral mode manager module.	2013-01-02 03:59:02 +00:00
if_ath_sysctl.c	Implement frame (data) transmission using if_transmit(), rather than	2013-01-15 18:01:23 +00:00
if_ath_sysctl.h	Break out most of the HAL related tweaks into a per-HAL instance,	2011-06-23 02:38:36 +00:00
if_ath_tdma.c	Pull out the debugging code from the critical path and make sure it	2012-11-28 01:55:22 +00:00
if_ath_tdma.h	Migrate the TDMA management functions out of if_ath.c into if_ath_tdma.c.	2012-05-20 02:49:42 +00:00
if_ath_tsf.h	Migrate the bulk of the RX routines out from if_ath.c to if_ath_rx.[ch].	2012-05-20 02:05:10 +00:00
if_ath_tx.c	Pull out the if_transmit() work and revert back to ath_start().	2013-02-13 05:32:19 +00:00
if_ath_tx.h	ALQ logging enhancements:	2012-11-16 19:57:16 +00:00
if_ath_tx_edma.c	Methodize the process of adding the software TX queue to the taskqueue.	2013-02-07 02:15:25 +00:00
if_ath_tx_edma.h	Begin separating out the TX DMA setup in preparation for TX EDMA support.	2012-07-23 03:52:18 +00:00
if_ath_tx_ht.c	Fix a corner case that I noticed with the AR5416 (and it's currently	2013-02-08 09:07:03 +00:00
if_ath_tx_ht.h	Introduce TX aggregation and software TX queue management	2011-11-08 22:43:13 +00:00
if_athdfs.h	Add a method to explicitly disable radar reporting if required.	2013-01-02 01:36:10 +00:00
if_athioctl.h	Add support for triggering spectral scan upon a channel reset/change.	2013-01-08 22:15:13 +00:00
if_athrate.h	Create an ioctl API for fetching the current rate control information.	2012-07-20 01:27:20 +00:00
if_athvar.h	Pull out the if_transmit() work and revert back to ath_start().	2013-02-13 05:32:19 +00:00