- 14 Jun, 2010 40 commits
-
-
Felix Fietkau authored
ath9k tries to prevent WMM queue tx buffer starvation caused by traffic on different queues by limiting the number of pending frames in a tx queue (tracked in the ath_buf structure). This had a leak issue, because the a skb can be reassigned to a different ath_buf in the tx path, causing the pending frame counter to become inaccurate. To fix this, track the number of frames in an array in the softc, using the mac80211 queue mapping as index. Signed-off-by:
Felix Fietkau <nbd@openwrt.org> Signed-off-by:
John W. Linville <linville@tuxdriver.com>
-
Felix Fietkau authored
AR5416 and all newer chipsets use a 32 bit rx timestamp, so there is no need to keep the 15 bit timestamp extending logic around. This patch removes ath9k_hw_extend_tsf (replaced by a call to ath9k_hw_gettsf64), and reduces the frequency of TSF reads, which can improve performance in some cases. This change also has the side effect of making rx timestamps more accurate. Signed-off-by:
-
Felix Fietkau authored
Signed-off-by:
-
Felix Fietkau authored
Signed-off-by:
-
Felix Fietkau authored
Signed-off-by:
-
Felix Fietkau authored
Signed-off-by:
-
Felix Fietkau authored
ath_get_mac80211_qnum() expects the queue 'subtype' (internal ID for the WMM AC) as argument when looking up the mac80211 queue, however ath_wake_mac80211_queue provides txq->axq_qnum instead, which contains the hardware queue number. Fix this by keeping track of the WMM class ID in the txq data structure. Signed-off-by:
-
Luis R. Rodriguez authored
This updates the initvals for AR9003 to adjust the 5 GHz tx gain tables for femless and high power PA. References: Osprey 2.0 header file ver 72 Osprey 2.2 header file ver 20 Checksums: $ ./initvals -f ar9003-2p0 0x00000000c2bfa7d5 ar9300_2p0_radio_postamble 0x00000000ada2b114 ar9300Modes_lowest_ob_db_tx_gain_table_2p0 0x00000000e0bc2c84 ar9300Modes_fast_clock_2p0 0x00000000056eaf74 ar9300_2p0_radio_core 0x0000000000000000 ar9300Common_rx_gain_table_merlin_2p0 0x0000000078658fb5 ar9300_2p0_mac_postamble 0x0000000023235333 ar9300_2p0_soc_postamble 0x0000000054d41904 ar9200_merlin_2p0_radio_core 0x00000000748572cf ar9300_2p0_baseband_postamble 0x000000009aa5a0a4 ar9300_2p0_baseband_core 0x000000003dffa526 ar9300Modes_high_power_tx_gain_table_2p0 0x000000001cfda724 ar9300Modes_high_ob_db_tx_gain_table_2p0 0x0000000011302700 ar9300Common_rx_gain_table_2p0 0x00000000e3eab114 ar9300Modes_low_ob_db_tx_gain_table_2p0 0x00000000c9d66d40 ar9300_2p0_mac_core 0x000000001e1d0800 ar9300Common_wo_xlna_rx_gain_table_2p0 0x00000000a0c54980 ar9300_2p0_soc_preamble 0x00000000292e2544 ar9300PciePhy_pll_on_clkreq_disable_L1_2p0 0x000000002d3e2544 ar9300PciePhy_clkreq_enable_L1_2p0 0x00000000293e2544 ar9300PciePhy_clkreq_disable_L1_2p0 $ ./initvals -f ar9003-2p2 0x00000000c2bfa7d5 ar9300_2p2_radio_postamble 0x00000000ada2b114 ar9300Modes_lowest_ob_db_tx_gain_table_2p2 0x00000000e0bc2c84 ar9300Modes_fast_clock_2p2 0x00000000056eaf74 ar9300_2p2_radio_core 0x0000000000000000 ar9300Common_rx_gain_table_merlin_2p2 0x0000000078658fb5 ar9300_2p2_mac_postamble 0x0000000023235333 ar9300_2p2_soc_postamble 0x0000000054d41904 ar9200_merlin_2p2_radio_core 0x000000008475a084 ar9300_2p2_baseband_postamble 0x000000009aaafd90 ar9300_2p2_baseband_core 0x000000003dffa526 ar9300Modes_high_power_tx_gain_table_2p2 0x000000001cfda724 ar9300Modes_high_ob_db_tx_gain_table_2p2 0x0000000011302700 ar9300Common_rx_gain_table_2p2 0x00000000a9a2b114 ar9300Modes_low_ob_db_tx_gain_table_2p2 0x00000000a9d66d40 ar9300_2p2_mac_core 0x000000001e1d0800 ar9300Common_wo_xlna_rx_gain_table_2p2 0x00000000a0c531c8 ar9300_2p2_soc_preamble 0x00000000292e2544 ar9300PciePhy_pll_on_clkreq_disable_L1_2p2 0x000000002d3e2544 ar9300PciePhy_clkreq_enable_L1_2p2 0x00000000293e2544 ar9300PciePhy_clkreq_disable_L1_2p2 Signed-off-by:
Luis R. Rodriguez <lrodriguez@atheros.com> Signed-off-by:
-
Luis R. Rodriguez authored
All AR9003 devices are PCI-E only, the extra delay here is not required and only reduces the delay for loading the initial register values by at least 14ms. Signed-off-by:
-
Luis R. Rodriguez authored
AR9003 has been tested with the new ANI implementation and so ANI can now be enabled for that family. Signed-off-by:
-
Luis R. Rodriguez authored
This adds support for ANI for AR9003. The implementation for ANI for AR9003 is slightly different than the one used for the older chipset families. It can technically be used for the older families as well but this is not yet fully tested so we only enable the new ANI for the AR5008, AR9001 and AR9002 families with a module parameter, force_new_ani. The old ANI implementation is left intact. Details of the new ANI implemention: * ANI adjustment logic is now table driven so that each ANI level setting is parameterized. This makes adjustments much more deterministic than the old procedure based logic and allows adjustments to be made incrementally to several parameters per level. * ANI register settings are now relative to INI values; so ANI param zero level == INI value. Appropriate floor and ceiling values are obeyed when adjustments are combined with INI values. * ANI processing is done once per second rather that every 100ms. The poll interval is now a set upon hardware initialization and can be picked up by the core driver. * OFDM error and CCK error processing are made in a round robin fashion rather than allowing all OFDM adjustments to be made before CCK adjustments. * ANI adjusts MRC CCK off in the presence of high CCK errors * When adjusting spur immunity (SI) and OFDM weak signal detection, ANI now sets register values for the extension channel too * When adjusting FIR step (ST), ANI now sets register for FIR step low too * FIR step adjustments now allow for an extra level of immunity for extremely noisy environments * The old Noise immunity setting (NI), which changes coarse low, size desired, etc have been removed. Changing these settings could affect up RIFS RX as well. * CCK weak signal adjustment is no longer used * ANI no longer enables phy error interrupts; in all cases phy hw counting registers are used instead * The phy error count (overflow) interrupts are also no longer used for ANI adjustments. All ANI adjustments are made via the polling routine and no adjustments are possible in the ISR context anymore * A history settings buffer is now correctly used for each channel; channel settings are initialized with the defaults but later changes are restored when returning back to that channel * When scanning, ANI is disabled settings are returned to (INI) defaults. * OFDM phy error thresholds are now 400 & 1000 (errors/second units) for low/high water marks, providing increased stability/hysteresis when changing levels. * Similarly CCK phy error thresholds are now 300 & 600 (errors/second) Signed-off-by: -
Luis R. Rodriguez authored
The new ANI implementation will use this to skip ANI calibration upon a scan. This cannot be ported to the older ANI implementation unless default ANI values from the ANI are also used upon a scan. This is essentially what one of the things thenew ANI does. Signed-off-by:
-
Luis R. Rodriguez authored
Signed-off-by:
-
Luis R. Rodriguez authored
The AR9003 hardware family will use a slightly modified ANI implementation which has not yet been tested on the other hardware families. To allow for this new ANI implementation a few ANI calls need to be abstracted away. This patch just allows for each hardware family to declare their own ANI ops and annotates the current ANI implementation as old. Signed-off-by:
-
Luis R. Rodriguez authored
We get an MIB interrupt when we hit certain PHY error counter thresholds. If ANI is disabled but the MIB interrupt is enabled we'll keep around the old MIB interrupt causes. Since ath9k disables the MIB interrupt when ANI is disabled this is not a fix, but more of a sanity fix in case we ever need the MIB interrupt enabled but disabling ANI. Signed-off-by:
-
Luis R. Rodriguez authored
The clock rate was assumed to be static but it actually changes depending on the mode of operation, correct this to help improve the calcuation of the listenTime for ANI. This change will help adjust ANI more accurately on different PHY thresholds. Signed-off-by:
-
Luis R. Rodriguez authored
These will be used by the ANI code next. Signed-off-by:
-
Johannes Berg authored
Currently, driver tracing is sometimes invoked after and sometimes before the actual driver callback. This is fine as long as the driver has no tracing itself, but as soon as it does it gets confusing. To make traces containing such information easier to read, introduce a return tracer in mac80211 that essentially brackets any driver tracing, and invoke the real trace before the driver's callback, only showing the return value, if any, afterwards. Since tracing records the process, there's no problem with overlapping calls if that should happen. Signed-off-by:
Johannes Berg <johannes.berg@intel.com> Signed-off-by:
-
Johannes Berg authored
The recent change to processing action frames from the management frame queue had already broken action frame accounting, and my rework didn't help either. So add back accounting and simplify the code with a label rather than duplicating it, and also add accounting for management frames. Signed-off-by:
-
Johannes Berg authored
Even before the recent changes, the documentation for TX aggregation was somewhat out of date. Update it and also add documentation for the RX side. Signed-off-by:
-
Johannes Berg authored
Allow drivers to sleep, and indicate this in the documentation. ath9k has some locking I don't understand, so keep it safe and disable BHs in it, all other drivers look fine with the context change. Signed-off-by:
-
Johannes Berg authored
To prepare for allowing drivers to sleep in ampdu_action, change the locking in the TX aggregation code to use the mutex the RX part already uses. The spinlock is still necessary around some code to avoid races with TX, but now we can also synchronize_net() to avoid getting an inconsistent sequence number. Signed-off-by:
-
Johannes Berg authored
Since we want the code to be able to sleep in the future, it must not be called from the timer directly. To achieve that, simply call the function drivers would call, and also use RCU in the timer to get the struct so we don't need to rely on the spinlock in the future. Signed-off-by:
-
Johannes Berg authored
To prepare for allowing drivers to sleep in ampdu_action, change the locking in the RX aggregation code to use a mutex, so that it would already allow drivers to sleep. But explicitly disable BHs around the callback for now since the TX part cannot yet sleep, and drivers' locking might require it. Signed-off-by:
-
Johannes Berg authored
I noticed that when there was _no_ traffic at all on a given aggregation session, it would never time out. This won't happen unless you forced creating a session, but fix it anyway. Signed-off-by:
-
Johannes Berg authored
Since we want the code to be able to sleep in the future, it must not be called from the timer directly. To prepare, move it out into the aggregation work. Signed-off-by:
-
Johannes Berg authored
Move the block-ack session works into common code, since it will be needed for RX agg too in the next patches. Signed-off-by:
-
Johannes Berg authored
When the driver or rate control requests starting or stopping an aggregation session, that currently causes a direct callback into the driver, which could potentially cause locking problems. Also, the functions need to be callable from contexts that cannot sleep, and thus will interfere with making the ampdu_action callback sleeping. To address these issues, add a new work item for each station that will process any start or stop requests out of line. Signed-off-by:
-
Johannes Berg authored
mac80211 currently maintains the ampdu_lock to avoid starting a queue due to one aggregation session while another aggregation session needs the queue stopped. We can do better, however, and instead refcount the queue stops for this particular purpose, thus removing the need for the lock. This will help making ampdu_action able to sleep. Signed-off-by:
-
Johannes Berg authored
The non-irqsafe aggregation start/stop done callbacks are currently only used by ath9k_htc, and can cause callbacks into the driver again. This might lead to locking issues, which will only get worse as we modify locking. To avoid trouble, remove the non-irqsafe versions and change ath9k_htc to use those instead. Signed-off-by:
-
Johannes Berg authored
Currently we allocate some memory for each TX aggregation session and additionally keep a state bitmap indicating the state it is in. By using RCU to protect the pointer, moving the state into the structure and some locking trickery we can avoid locking when the TX agg session is fully operational. Signed-off-by:
-
Johannes Berg authored
Currently we allocate some memory for each RX aggregation session and additionally keep a flag indicating whether or not it is valid. By using RCU to protect the pointer and making sure that the memory is fully set up before it becomes visible to the RX path, we can remove the need for the bool that indicates validity, as well as for locking on the RX path since it is always synchronised against itself, and we can guarantee that all other modifications are done when the structure is not visible to the RX path. The net result is that since we remove locking requirements from the RX path, we can in the future use any kind of lock for the setup and teardown code paths. Signed-off-by:
-
Johannes Berg authored
This moves the aggregation callback processing to the per-sdata skb queue and a work function rather than the tasklet. Unfortunately, this means that it extends the pkt_type hack to that skb queue. However, it will enable making ampdu_action API changes gradually, my current plan is to get rid of this again by forcing drivers to only return from ampdu_action() when everything is done, thus removing the callbacks completely. Signed-off-by:
-
Johannes Berg authored
There's a corner case where we receive a fragmented frame during a blockack session, in which case we will terminate that session. To simplify future work in this area that will culminate in allowing the driver callbacks for aggregation to sleep, move the processing of this case out of the RX path into the interface work. This will simplify future work because the new place for this code doesn't require that the function will always be atomic, which the RX path needs. Signed-off-by:
-
Johannes Berg authored
To prepare for making the ampdu_action callback sleep, make mac80211 always process blockack action frames from the skb queue. This gets rid of the current special case for managed mode interfaces as well. Signed-off-by:
-
Johannes Berg authored
Some code is duplicated between ibss, mesh and managed mode regarding the queueing of management frames. Since all modes now use a common skb queue and a common work function, we can pull the queueing code into the rx handler directly and remove the duplicated length checks etc. Signed-off-by:
-
Johannes Berg authored
All the management processing functions free the skb after they are done, so this can be done in the new common code instead. Signed-off-by:
-
Johannes Berg authored
Even with the previous patch, IBSS, managed and mesh modes all attach their own work function to the shared work struct, which means some duplicated code. Change that to only have a frame processing function and a further work function for each of them and share some common code. Signed-off-by:
-
Johannes Berg authored
IBSS, managed and mesh modes all have their own work struct, and in the future we want to also use it in other modes to process frames from the now common skb queue. This also makes the skb queue and work safe to use from other interface types. Signed-off-by:
-
Johannes Berg authored
IBSS, managed and mesh modes all have an skb queue, and in the future we want to also use it in other modes, so make them all use a common skb queue already. Signed-off-by:
-
