Hardware doesn't place any restrictions on the buffer alignment,
consider this TODO resolved.
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

If card reset with firmware download executed, followed by reset
with use of firmware from build in flash, firmware download indication
remains in the hardware register.
When running firmware download flow,
the SW download indication is written by the driver to bit 0 in usage_6:
wil_fw_load(), "S(RGF_USER_USAGE_6, 1);"
This register, like all USER RGF, wasn't reset in SW reset flow.
Therefore the driver must clear it on SW reset flow.
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

In the wil_tx_ring, error message printed when tx attempted
while vring has no space to accommodate all fragments of frame.
Normally, such situation handled by stopping tx queue.
But, if tx queue is by-passed (like pktgen does), this error
will be triggered at high rate and dmesg will be flooded with
this message. Whole system may become unstable and hang with
no possible recover except power cycle.

Rate-limit it to prevent dmesg flooding.
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

In some compilation environments, result of pointer arithmetic interpreted as int
while in others it is long int. Force conversion to long.
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Allow to configure VRING size for both Rx and Tx via module parameters:
rx_ring_order and tx_ring_order. Parameters are ring size orders, i.e.
ring size calculated as 1 << order.
Defaults for both Tx and Rx are order 9, i.e. size 512
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

There is no need to obtain physical device through
wil->pdev->dev path, as it is done by this macro.
The same device already stored as wiphy's device, thus
wil_to_dev() returns the same device as wil_to_pcie_dev()

Remove unnecessary macros, this allows to drop dependency
by pci.h in the firmware download code.
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Scan timeout treated as indication for firmware error;
and should be handled in the same way.

Recovery state machine does not perform as designed because
its state is not updated in case of scan timeout.

Fix is to set recovery state machine into the proper state.
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

RX_HTRSH interrupt is handled in exactly the same manner
as RX_DONE interrupt - fetching accumulated packets from RX
ring. In addition there's a rate limitted warning message.
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Propagate reason for the disconnect through the relevant call chains:
- report to cfg80211 reason as reported by the firmware
- provide to the firmware reason as requested by cfg80211
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Instead of killing interrupts during reset when the first one happens,
kill them before issuing the reset.
This fixes an easy to reproduce crash with multiple cards sharing the
same IRQ.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Closes another small IRQ handler race
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

IRQs are suppressed if ah == NULL and ATH_OP_INVALID being set in
common->op_flags. Close a short time window between those two.
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Like with ath9k, ath5k queues also need to be ordered by priority.
queue_info->tqi_subtype already contains the correct index, so use it
instead of relying on the order of ath5k_hw_setup_tx_queue calls.

Cc: stable@vger.kernel.org
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Hardware queues are ordered by priority. Use queue index 0 for BK, which
has lower priority than BE.

Cc: stable@vger.kernel.org
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

The driver passes the desired hardware queue index for a WMM data queue
in qinfo->tqi_subtype. This was ignored in ath9k_hw_setuptxqueue, which
instead relied on the order in which the function is called.

Cc: stable@vger.kernel.org
Reported-by: NHubert Feurstein <h.feurstein@gmail.com>
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Fix missing memory deallocation on error paths in wil_write_file_wmi()
and wil_write_file_txmgmt().
Reported-by: NAhmed Tamrawi <ahmedtamrawi@gmail.com>
Signed-off-by: NLino Sanfilippo <LinoSanfilippo@gmx.de>
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Don't bother with fancy arithmetic and just
hardcode the final values.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In theory it was possible to starve the system if
a tx/rx handler could implicitly trigger more
tx/rx pci events.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Pass the eid argument via skbuff control buffer.
This will make it possible to work with queues of
HTC event buffers.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This wasn't used since forever and there are no
plans on using it.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The service bitmap can be used to determine
firmware capabilities.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This shouldn't really happen but take into account
the original service bitmap length when mapping
service ids.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The dump print uses sizeof() but since service_map
was a mere pointer the dump was too short.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Traces shouldn't modified passed data. This will
make it possible to pass const arguments to
traces.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Increase the rts threshold from the legacy value of 2347 to support higher
threshold limit.
Signed-off-by: NVivek Natarajan <nataraja@qti.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The number of peers isn't directly translatable to
the number of stations because ath10k needs to
reserve a few extra peers for special cases like
multi-vif concurrency.

The previous limit was 126 and 15 stations in AP
mode for 10.x and main firmware branches
respectively. The limit is now 128 and 16 which
was the original intention.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The var was supposed to be protected by data_lock
but it wasn't so in all instances. It's actually
not necessary to have a spinlock protected
num_peers so drop it.

All instances of num_peers are already within
conf_mutex sections so use that.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This prevents warning spamming if peer creation
fails during sta_state in some cases.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Add mem_val debugfs file for dumping the firmware (target) memory and also for
writing to the memory. The firmware memory is accessed through one file which
uses position of the file as the firmware memory address. For example, with dd
use skip parameter for the address.

Beucase target memory width is 32 bits it's strongly recommended to use
blocksize divisable with 4 when using this interface. For example, when using
dd use bs=4 to set the block size to 4 and remember to divide both count and
skip values with four.

To read 4 kB chunk from address 0x400000:

dd if=mem_value bs=4 count=1024 skip=1048576 | xxd -g1

To write value 0x01020304 to address 0x400400:

echo 0x01020304 | xxd -r | dd of=mem_value bs=4 seek=1048832

To read 4 KB chunk of memory and then write back after edit:

dd if=mem_value of=tmp.bin bs=4 count=1024 skip=1048576
emacs tmp.bin
dd if=tmp.bin of=mem_value bs=4 count=1024 seek=1048576
Signed-off-by: NYanbo Li <yanbol@qti.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Debugfs files reg_addr and reg_val are used for reading and writing to the
firmware (target) registers. reg_addr contains the address to be accessed,
which also needs to be set first, and reg_value is when used for reading and
writing the actual value in ASCII.

To read a value from the firmware register 0x100000:

# echo 0x100000 > reg_addr
# cat reg_value
0x00100000:0x000002d3

To write value 0x2400 to address 0x100000:

# echo 0x100000 > reg_addr
# echo  0x2400 > reg_value
#
Signed-off-by: NYanbo Li <yanbol@qti.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

ath10k_tx_wep_key_work() acquires conf_mutex, so
cancelling it when conf_mutex is already taken
in ath10k_remove_interface() is incorrect, so
move it outside the lock.

Snippet from the lockdep report:

kernel: ======================================================
kernel: [ INFO: possible circular locking dependency detected ]
kernel: 3.18.0-rc5-wl-debug #34 Tainted: G           O
kernel: -------------------------------------------------------
kernel: hostapd/451 is trying to acquire lock:
kernel:  ((&arvif->wep_key_work)){+.+...}, at: [<ffffffff810872d5>] flush_work+0x5/0x290
kernel: but task is already holding lock:
kernel:  (&ar->conf_mutex){+.+.+.}, at: [<ffffffffa0b99f00>] ath10k_remove_interface+0x40/0x290 [ath10k_core]
kernel: which lock already depends on the new lock.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

peer->keys needs to be protected by data_lock
since it is also accessed from the WMI path.

Both install() and clear() routines for peer
keys modify the key contents, so use the data_lock
to avoid races.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When static keys are used in shared WEP, when a
station is associated, message 3 is sent with an
encrypted payload. But, for subsequent
authentications that are triggered without a
deauth, the auth frame is decrypted by the HW.

To handle this, check if the WEP keys have already
been set for the peer and if so, mark the
frame as decrypted. This scenario can happen
when a station changes its default TX key and initiates
a new authentication sequence.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Some firmware revisions don't seem to deilver
management frames with FCS error via WMI so narrow
down the HTT rule to not drop corrupted management
frames.

This basically increases number of frames ath10k
reports while sniffing.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When mac80211 requests driver to cancel a hw roc
the driver must not call the expired() callback or
else roc will fail in some cases depending on how
things get scheduled.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

With P2P concurrency requested hw roc duration
time can be very small. Some firmware revisions
refuse scan requests with too small channel dwell
time.

This prevents messages like, e.g. with connected
STA vif and performing P2P Find:

 ath10k_pci 0000:00:05.0: failed to switch to channel for roc scan
 ieee80211 phy3: failed to start next HW ROC (-110)
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

New firmware revisions don't need peer creation
when doing offchannel tx. Earlier revisions would
queue and never release frames without a peer.

This prevent new firmware revisions from stopping
replenishing wmi-htc tx credits and improves
reliability of offchannel tx which would sometimes
silently fail.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Comment was out-of-date. The headroom is no longer
necessary because HTT Tx fragment list is stored
in dma pool item associated with each sk_buff.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When re-associating a station, the nss was set back to
maximum value even if user had configured small number
of tx chains.  So, pay attention to user's config in
this case as well.
Signed-off-by: NBen Greear <greearb@candelatech.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

It appears it takes more than just setting the
hardware's chainmask to make things work well.  Without
this patch, a vdev would only use 1x1 rates when chainmask
was set to 0x3.

Setting the 'nss' (number of spatial streams) on the vdev
helps the firmware's rate-control algorithm work properly.
Signed-off-by: NBen Greear <greearb@candelatech.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>