This fix the regerssion introduced after "ath10k_wmi_requests_stats()
call to use more stat bits", both the 10.2.X FW and community FW 10.1.X
will return time out when access the fw_stats with the STAT_VDEV and
STAT_PDEV. Add the condition to use different fw_stats mask for different
FW versions.

Fixes: 7777d8c7 ("ath10k: implement fw stats for wmi-tlv")
Signed-off-by: NYanbo Li <yanbol@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The default keyidx callback may be called after
more than 1 key is installed. This led to only 1
WEP key being reinstalled only. This caused Rxed
traffic encrypted with other WEP keys to be
dropped in client mode.
Tested-by: NBartosz Markowski <bartosz.markowski@tieto.com>
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The variable should be cleared regardless of
whether there's a peer associated with the key or
not.

This fixes case when user first associates with 2
WEP keys and then disconnects and connects with 1
WEP key. This resulted in WEP key count being 2 in
the driver leading to default keyidx fixup
failure.
Tested-by: NBartosz Markowski <bartosz.markowski@tieto.com>
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There's always at most 2 credits and it makes
little sense to set the
ATH10K_HTC_FLAG_NEED_CREDIT_UPDATE flag
conditionally.

This seems to fix some random issues with tx
credit starvation on WLAN.RM.2.0-00073 I've been
seeing. Note: this isn't related to wmi mgmt tx.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Some trace messages were truncated and a kernel
splat could be seen in the log:

  WARNING: CPU: 3 PID: 0 at /devel/src/linux/drivers/net/wireless/ath/ath10k/./trace.h:114 ftrace_raw_event_ath10k_log_dbg+0x20e/0x220 [ath10k_core]()
  Modules linked in: ath10k_pci(O) ath10k_core(O) ath iwldvm iwlwifi [last unloaded: iwlwifi]
  CPU: 3 PID: 0 Comm: swapper/3 Tainted: G        W  O    4.0.0-rc3-wl-ath+ #703
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.7.5-20140531_083030-gandalf 04/01/2014
   ffffffffa01d4cb0 ffff88001fd83998 ffffffff8191b86c ffffffff81e3b718
   0000000000000000 ffff88001fd839d8 ffffffff8105573a ffff88001c0a5528
   ffff88001bea9ae0 ffff88001c3dd940 00000000000d0018 ffff88001fd83a80
  Call Trace:
   <IRQ>  [<ffffffff8191b86c>] dump_stack+0x45/0x57
   [<ffffffff8105573a>] warn_slowpath_common+0x8a/0xc0
   [<ffffffff8105582a>] warn_slowpath_null+0x1a/0x20
   [<ffffffffa01c61fe>] ftrace_raw_event_ath10k_log_dbg+0x20e/0x220 [ath10k_core]
   [<ffffffffa01aaa6b>] ath10k_dbg+0xbb/0xd0 [ath10k_core]
   [<ffffffff810eed89>] ? trace_clock_local+0x9/0x10
   [<ffffffffa01bd069>] ath10k_wmi_event_service_ready+0x479/0x520 [ath10k_core]
   [<ffffffff810fdf90>] ? trace_buffer_unlock_commit+0x50/0x60
   [<ffffffffa01c2113>] ath10k_wmi_tlv_op_rx+0x6b3/0x8b0 [ath10k_core]

This could be reproduced with:

  trace-cmd record -e ath10k
  ifconfig wlan0 down
  ifconfig wlan0 up

Fixes: 5c01aa3d ("ath10k: deduplicate wmi service ready logic")
Fixes: ca996ec5 ("ath10k: implement wmi-tlv backend")
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Firmwares that indicate BURST_SERVICE as enabled
could overwrite some wmm parameters (eg. txop).
This could lead IOT and balancing issues.
Hence disable this by default.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

WLAN.RM.2.0-00073 firmware requires self-peer to
be created prior to issuing scan command. Without
this wmi-tlv with qca6174 firmware crashes after
submitting a scan request.

Creating the peer as soon as add_interface()
shouldn't be a problem.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The code was a bit convoluted. Clean it up and
prepare for future changes.

While at it this fixes incorrect verification of
'single nss' case when ss2 rates were missing
while ss1 and ss3 were requested resulting in
nss=3 being set:

  iw wlan1 set bitrates legacy-5 ht-mcs-5 0 1 2 3 4 5 6 7 16 17 18 19 20 21 22 23 vht-mcs-5 1:0-9 3:0-9
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The number of 3 vdevs wasn't enough to handle the
worst case for interface combinations in practice.

wpa_supplicant may need up to 4 vifs to have 2
vifs actually connected, i.e. p2pdev + client + 2x
p2p (either p2p client or p2p go).

This fixes worst case warning:

  Free vdev map is empty, no more interfaces allowed.

This keeps the ability to associate 32 stations in
AP mode at the cost of not being able to guarantee
that under all circumstances, i.e. some
combinations may consume additional fw peer
entries for internal purposes leaving less
resource for stations in AP mode.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This effectively enables multi-channel operation
on qca6174 WLAN.RM.2.0-00073 (and possibly any
newer firmware release for qca6174).

This adds appopriate interface combinations and
initializes firmware channel scheduler.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

qca6174 wmi-tlv firmware defines a new wmi event
for host tx pausing (i.e. stop/wake tx queues).

Map these events to ath10k/mac80211 tx queue
control.

This is important for multi-channel throughput
performance.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Tx queue locking was very simple until now.
Multi-channel support will require a more flexible
and fine grained control.

This introduces a per-hw and per-vif (each with a
bitmask of reasons) tx queue locking.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This command will be used to configure
multi-channel scheduler in firmware.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The chanctx API will allow ath10k to support
multi-channel operation.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

It doesn't make much sense to reject a valid
firmware configuration combination.

Since SSID isn't known early on it might make
sense to allow driver to start vdev without SSID
and restart it later.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

qca6174 and wmi-tlv doesn't have quiet mode
implemented. Don't even attempt to call it. This
fixes a warning when bringing first interface up:

  failed to set quiet mode period 100 duarion 0 enabled 0 ret -95

Fixes: 8515b5c7 ("ath10k: configure thermal throttle while powering up")
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Number of channels is stored in a separate macro
in a header file and channel list is constructed
independently. The macro is used to define survey
array.

This fixes a recent regression introduced after
adding support for 144 channel. The regression
would lead to a warning and incomplete survey data
on channel 165:

  chan info: invalid frequency 5825 (idx 38 out of bounds)

Also make sure to enforce the sizes and avoid this
kind of problem in the future.

Fixes: 4a7898fe ("ath10k: enable channel 144 on 5GHz band")
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

All of the bringup/init functions called in ath10k_core_start return 0
on success and != 0 on failure. ath10k_wmi_wait_for_service_ready(),
ath10k_wmi_wait_for_unified_ready() and their call sites were adjusted
to fit this model.
The return type of wait_for_completion_timeout is unsigned long not int so
ath10k_wmi_wait_for_service_ready() and ath10k_wmi_wait_for_unified_ready()
were fixed up accordingly.
Signed-off-by: NNicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Return type of wait_for_completion_timeout is unsigned long not int.
An appropriately named unsigned long is added and the assignments fixed up.
Signed-off-by: NNicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Return type of wait_for_completion_timeout is unsigned long not int.
An appropriately named unsigned long is added and the assignments fixed up.
Signed-off-by: NNicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Return type of wait_for_completion_timeout is unsigned long not int.
An appropriately named unsigned long is added and the assignments fixed up.
Missing spaces 1*HZ -> 1 * HZ were also added along the way.
Signed-off-by: NNicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Return type of wait_for_completion_timeout is unsigned long not int.
An appropriately named unsigned long is added and the assignments fixed up.
Signed-off-by: NNicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Return type of wait_for_completion_timeout is unsigned long not int.
An appropriately named unsigned long is added, respectively 'ret'
renamed, and the assignments fixed up.
Signed-off-by: NNicholas Mc Guire <hofrat@osadl.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

While debugging something else I noticed AID was
set to 0. This could lead to powersave issues in
station mode. Maybe this isn't really necessary
but set it properly just to be sure.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Don't use literal values for offsets. While at it
rename a function so it is more clear what it
checks for.

This finally takes care of the last magic
5GHz/OFDM offset.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

It's possible to derive rate index from bitrate
without any additional mapping structures/logic.

This should have little to none impact on
performance since this is only done for management
frames and the previous approach wasn't
particularly optimized.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Instead of using a hacky table and magic values
use supported band information advertised to
mac80211.

This may impact performance a little when dealing
with legacy rx rates depending on system
architecture. It's probably negligible.

This also fixes a highly theoretical corner case
when HT/VHT rates weren't reported correctly if
channel frequency wasn't known.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Use the existing ieee80211_rate array instead of
definining separate one. This gets rid of the ugly
4-index offset when dealing with 5GHz band.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Using raw values is discouraged. Having
defines/enums is preferred.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Prepare defines for future use.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

It was superfluous and confusing. It's better to
define the structure explicitly.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There's no need to store rate values in hw_value.
This frees up the hw_value for better future use.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Prepare the code for future changes so that new
code can refer to rate-related stuff without
forward declarations.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This patch introduces tdls without tdls peer uapsd and tdls channel switching.
Transmitting tdls data frames works only for ethernet type frames, that's why
data addressed to tdls sta is in ethernet format.

This patch depends on commit c23e31cf ("mac80211: initialize rate control
earlier for tdls station").
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NMarek Kwaczynski <marek.kwaczynski@tieto.com>
Signed-off-by: NMarek Puzyniak <marek.puzyniak@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

As a part of tdls implementation introduce
tdls related wmi data structures, constant
values and functions.
Signed-off-by: NMarek Puzyniak <marek.puzyniak@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Currently station counting functions
(inc_num_stations/dec_num_stations) does not handle
tdls type of stations. Tdls station should be counted
because it consumes peer in firmware. Only not tdls
stations are excluded from this counting.
Signed-off-by: NMarek Puzyniak <marek.puzyniak@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Currently number of tdls vdevs supported by firmware
is hardcoded. For future usage it is stored in ath10k
structure based on defined value.
Signed-off-by: NMarek Puzyniak <marek.puzyniak@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Peer type was hardcoded to default value.
For future implementation it is required
to make is configurable.
Signed-off-by: NMarek Puzyniak <marek.puzyniak@tieto.com>
Signed-off-by: NMarek Kwaczynski <marek.kwaczynski@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There are a few different tx paths depending on
firmware and frame itself.

Creating a uniform decision will make it possible
to switch between different txmode easier, both
for testing and for future features as well.
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NMarek Puzyniak <marek.puzyniak@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

If htt.freq isn't cleared and contains garbage fw
may discard tx packets. Prevent this from
happening by clearing htt.freq properly.

Possible manifestation of the problem could be not
being able to send auth request/response frames on
firmware with HTT >= 3.4 (when freq param was
introduced), e.g. on qca6174.

Fixes: 8d6d3624 ("ath10k: fix offchan reliability")
Signed-off-by: NMichal Kazior <michal.kazior@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>