ath_hw_cycle_counters_update only needs to be called if the power state
changes. Most of the time this does not happen, even when ps_usecount
goes down to 0.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
During a reset, rx buffers are flushed after rx has been disabled. To avoid
race conditions, rx needs to stay disabled during the reset, so avoid any
calls to ath9k_hw_rxena in that case.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The interrupt handler increases the interrupt disable refcount, so the
tasklet needs to always call ath9k_hw_enable_interrupts.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
On MIMO chips this can be used to enable/disable hardware chains, ensuring
that the MCS information is updated accordingly.
On non-MIMO chips with rx diversity (e.g. 9285), this configures the rx
input antenna.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
reduces unnecessary code duplication. Also takes the sc_pcu_lock within
ath_reset instead of callsites, which makes it possible to always cancel
all queued work items before the reset, possibly fixing a few race
conditions (work items vs reset) along with it.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This makes it much easier to add further rework to avoid race conditions
between reset and other work items.
Move other functions to make ath_reset static.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Helps with making ath_reset static in the next commit
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
we already have ah->{rx,tx}chainmask for the same purpose
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch keep track of number of samples that includes
DMA debugs registers, PCU observe, CR, channel noise,
cycle conters, noisefloor history buffer and last N number
of tx and rx descriptor status. These samples are grouped
in table manner which dumping in debgufs.
Debugfs file location:
<debugfs_mnt>/ieee80211/phy#/ath9k/samples
Signed-off-by: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
this useful for debugging and to keep track of success/failure of
frames such as ACK, RTS and FCS error count in a noisy environment
Signed-off-by: Mohammed Shafi Shajakhan <mohammed@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
While switching b/w HT20/40, the current channel's nf values
are updated into history buffer. Since the current channel's
channel type, channel flag got updated before reading
nf value from hw. This channel type mismatch is causing invalid
readings when hw is on ht20 but getnf tries to read on extn chains.
Signed-off-by: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Whenever RXEOL is received, both RXORN and RXEOL got cleared
to avoid rx overrun interrupt storm. This was handled only for
edma chips. The same scenario was also observered with AR9280,
doing frequent channel type switch b/w HT20/40 with bidi traffic
that is causing failure to stop rx dma. This patch clears
the RXEOL & RXORN interrupts for all chips.
ath: DMA failed to stop in 10 ms AR_CR=0x00000024 AR_DIAG_SW=0x42000020
DMADBG_7=0x000062c0
ath: Could not stop RX, we could be confusing the DMA engine when we
start RX up
------------[ cut here ]------------
WARNING: at drivers/net/wireless/ath/ath9k/recv.c:532
ath_stoprecv+0x110/0x120 [ath9k]()
Call Trace:
[<ffffffff8104a55a>] warn_slowpath_common+0x7a/0xb0
[<ffffffff8104a5a5>] warn_slowpath_null+0x15/0x20
[<ffffffffa0560380>] ath_stoprecv+0x110/0x120 [ath9k]
[<ffffffffa055e6fa>] ath_reset+0x6a/0x200 [ath9k]
Signed-off-by: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Disable ASPM in pci ->probe on upstream (device) and downstream
(PCIe port) component. According to e1000e driver authors this is
required. I did not find that requirement in PCIe spec, but it seems
to be logical for me.
This need to be fixed for CONFIG_PCIEASPM, that will be done later ...
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
We always call ->config_pci_powersave() with both restore and power_off
arguments equal to 0 or both equal to 1, so merge them into one
argument.
Signed-off-by: Stanislaw Gruszka <sgruszka@redhat.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Let us enable/disable interrupts based on reference count.
By doing this we can ensure that interrupts are never be
enabled in the middle of tasklet processing. Instead of
addressing corner cases like "ath9k: avoid enabling interrupts
while processing rx", this approach handles it in generic manner.
Signed-off-by: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
At preset set_interrupt also enables interrupt after changing
mask. This is not necessary in all cases and also sometime it
breaks the assumption that interrupt was disabled. So let us
enable the interrupt explicity if it was disabled earlier.
This could also avoid unnecessary register ops and also helps
the follow up patch to have global ref count for interrupts ops.
Signed-off-by: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The received tx status of aggregated frame without BlockAck may
cause deaf state in AR5416 cards. So the driver does a reset to
recover. When this happens, we release the pcu_lock before doing
a reset as ath_rest acquires pcu_lock. This is ugly and also not
atomic. Fixing this addresses the TX DMA failure also.
ath_tx_complete_aggr can be called from different paths which
takes different variants of spin_lock. This patch also addresses
the following warning.
WARNING: at kernel/timer.c:1011 del_timer_sync+0x4e/0x50()
Call Trace:
<IRQ> [<ffffffff8104be3a>] warn_slowpath_common+0x7a/0xb0
[<ffffffff8104be85>] warn_slowpath_null+0x15/0x20
[<ffffffff8105915e>] del_timer_sync+0x4e/0x50
[<ffffffffa03726be>] ath_reset+0x3e/0x210 [ath9k]
[<ffffffff8135cdaf>] ? _raw_spin_unlock_bh+0x1f/0x30
[<ffffffffa037760a>] ath_tx_complete_aggr.isra.26+0x54a/0xa40 [ath9k]
Signed-off-by: Rajkumar Manoharan <rmanohar@qca.qualcomm.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
during suspend/resume ath9k_stop will always call ath_radio_disable
which will inturn call ath9k_hw_configpcipowersave, so no need to call
it explicitly in ath9k_stop
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Add few debug messages for some of the possible scenarios
where we can detect PAPRD failures. this will help us to be
sure that we had really enabled PAPRD
Cc: susinder@qca.qualcomm.com
Cc: senthilb@qca.qualcomm.com
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
this helps the user to start/stop ANI dynamically.
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
In PAPRD table initialization path we do some register read, so
make sure the chip is awake during that. Currently PAPRD is disabled.
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Configure ibss node's bss_changes under BSS_CHANGED_IBSS.
And also start/stop ani timer only if the station join/leave
the group.
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
EDMA based chips (AR9380+) have 8 Tx FIFO slots, which are used to fix the
tx queue start/stop race conditions which have to be worked around for
earlier chips by keeping the last descriptor in the queue. The current code
stores all frames that do not fit onto the 8 FIFO slots in a separate
list. Whenever a FIFO slot is freed up, the next frame (or A-MPDU) from the
pending queue gets moved to that slot.
This process is not only inefficient, but also unnecessary. The code can
be improved visibly by keeping the pending queue fully linked, and moving
the contents of the entire queue to a FIFO slot as it becomes available.
This patch makes the necessary changes for that and also merges some code
that was duplicated for EDMA vs non-EDMA. It changes txq->axq_link to point
to the last descriptor instead of the link pointer, so that
ath9k_hw_set_desc_link can be used, which works on all chips.
With this patch, a small performance increase for non-aggregated traffic
was observed on AR9380 based embedded hardware.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Resetting hardware helps to recover from baseband
hang/panic for AR9003 based chips.
Cc: stable@kernel.org
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
we are not doing anything by tracking the number of pending frames.
bail out when we first find a pending frame in any one of the 10 queues.
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Only leave filtering enabled for AP or VLAN interfaces, clear the
destination mask for all other interfaces.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Reported-by: Ben Greear <greearb@candelatech.com>
Tested-by: Ben Greear <greearb@candelatech.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Mesh beaconing on ath9k was broken by this commit:
commit 4801416c76
Author: Ben Greear <greearb@candelatech.com>
Date: Sat Jan 15 19:13:48 2011 +0000
This patch assigns the right opmode when the device is used in mesh
mode.
Reported-by: Fabrice Deyber fabricedeyber@agilemesh.com
Signed-off-by: Javier Cardona <javier@cozybit.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This beacon rssi will be used to set noisefloor during ani reset.
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Processing TSF out of range before RX helps to update beacon
timers so early in the succeeding rx process.
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
The assumsion is that while processing ath9k tasklet,
interrupts were already disabled and it will be enabled
at the completion of ath9k tasklet. But whenever TSFOOR is raised,
the driver configures the beacon timers after having received a
beacon frame from the AP which inturn enables the interrupts.
Cc: stable@kernel.org
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
during suspend/S3 state drv_flush is called from mac80211 irrespective of
interface count. In ath9k we queue a work in ath9k_flush which we expect
to be cancelled in the drv_stop call back. during suspend process mac80211
calls drv_stop only when the interface count(local->count) is non-zero.
unfortunately when the network manager is enabled, drv_flush is called
while drv_stop is not called as local->count reaches '0'.
So fix this by simply checking for the device presence in the
drv_flush call back in the driver before queueing work or anything else.
this patch fixes the following WARNING
Call Trace:
[<c014c6e2>] warn_slowpath_common+0x72/0xa0
[<fc133f99>] ? ieee80211_can_queue_work+0x39/0x50 [mac80211]
[<fc133f99>] ? ieee80211_can_queue_work+0x39/0x50 [mac80211]
[<c014c75b>] warn_slowpath_fmt+0x2b/0x30
[<fc133f99>] ieee80211_can_queue_work+0x39/0x50 [mac80211]
[<fc134ed1>] ieee80211_queue_delayed_work+0x21/0x50 [mac80211]
[<fc1e5b22>] ath_tx_complete_poll_work+0xb2/0x100 [ath9k]
[<c016399e>] run_workqueue+0x8e/0x150
[<fc1e5a70>] ? ath_tx_complete_poll_work+0x0/0x100 [ath9k]
[<c0163ae4>] worker_thread+0x84/0xe0
[<c0167a60>] ? autoremove_wake_function+0x0/0x50
[<c0163a60>] ? worker_thread+0x0/0xe0
[<c01677d4>] kthread+0x74/0x80
[<c0167760>] ? kthread+0x0/0x80
[<c0104087>] kernel_thread_helper+0x7/0x10
---[ end trace 2aff81010df9215b ]---
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
when aggregation protection mode is enabled the hardware needs
to send RTS/CTS for each HT frame. Currently its disabled so
remove the unused call backs.
Signed-off-by: Mohammed Shafi Shajakhan <mshajakhan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
While draining the txq in flush, the buffers can be
added into the tx queue by tx_tasklet which leads to
unneccesary chip reset.
This issue was originially found with AR9382 and
running heavy uplink udp traffic with higher bandwidth
and doing frequent bgscan.
Cc: stable@kernel.org
Signed-off-by: Vasanthakumar Thiagarajan <vasanth@atheros.com>
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Requesting beacon sync up to configure beacon timers properly
in hw, has be done after doing beacon config with default values.
Setting the flags in beacon config is causing the device to not
enter into network sleep on idle state.
Signed-off-by: Rajkumar Manoharan <rmanoharan@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
We should protect hw_pll handler with power save wrappers and
also modularize hw_pll handler properly for better readability.
Also add a debug message to track chip resets on pll hang condition.
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
There is no reason why pll work handler should be part of xmit
file. move it to main.c so that reading hw check routines are
all in the same place.
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
We need not wake up the chip even before mutex lock is acquired and also
that it is required only if we are going to drain the txq. So place the
wrappers accordingly and this change is also useful when there are no
pending frames in the txq as we do not wake up the chip unnecessarily.
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Frame filtering relies on having a valid destination index (keycache slot),
to keep track of the destination. Assigning a keycache slot (configured
to unencrypted, with no key data attached) improves powersave handling in
AP mode with no encryption.
The dummy keycache entry for a station is cleared, when a real key gets
added.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
This patch fixes a long standing issue of pending packets in the queue being
sent (and retransmitted many times) to sleeping stations.
This was made worse by aggregation through driver-internal retransmitting
of A-MDPU subframes.
Previously the hardware tx filter was cleared unconditionally for every
single packet - with this patch it uses the IEEE80211_TX_CTL_CLEAR_PS_FILT
for unaggregated frames.
A sta_notify driver op is added to stop aggregation for stations when they
enter powersave mode. Subframes stay buffered inside the driver, to ensure
that the BlockAck window keeps a sane state.
Since the driver uses software aggregation, the clearing of the tx filter
needs to be handled by the driver instead of mac80211 for aggregated frames.
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Felix Fietkau