Per comments from Bjorn Andersson <bjorn.andersson@sonymobile.com>,
split DMA mask register writing as seperate patch in case we need
bi-sect in the furture.
Signed-off-by: NFengwei Yin <fengwei.yin@linaro.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Lawrence reported that git clone could make system crash on a
Qualcomm ARM soc based device (DragonBoard, 1G memory without
swap) running 64bit Debian.

It's turned out the crash is related with rx skb allocation
failure. git could consume more than 600MB anonymous memory.
And system is in extremely memory shortage case.

But driver didn't handle the rx allocation failure case. This patch
doesn't submit skb to upper layer if rx skb allocation fails.
Instead, it reuse the old skb for rx DMA again. It's more like
drop the packets if system is in memory shortage case.

With this change, git clone is OOMed instead of system crash.
Reported-by: NKing, Lawrence <lking@qti.qualcomm.com>
Signed-off-by: NFengwei Yin <fengwei.yin@linaro.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

There is a type bug so it always returns success.

Fixes: 6fa658fd ('ath9k: Simplify and fix eeprom endianness swapping')
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Ath9k driver does not modify tx skbs, so SUPPORTS_CLONED_SKBS
flag can be set. Enabling this flag significant reduce number
of copy operation during TCP Tx. This is especially noticeable
on platforms with slower CPU (lower CPU usage brings
profits in better TCP Tx troughput results).

Tested on MIPS with 560 MHz clock
Without CLONED_SKBS flag:
TCP Tx 145 Mb/s (iperf result)
__copy_user_common consumes 12.9% of CPU (result from perf tool)
0% CPU Idle

With CLONED_SKBS flag:
TCP Tx 170 Mb/s (iperf result)
__copy_user_common consumes 1.8% of CPU (result from perf tool)
12% CPU Idle
Signed-off-by: NPawel Kulakowski <pawel.kulakowski@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Added a simple interface for platform to perform crash
recovery.
When firmware crashes, wil driver can notify the platform
which can trigger a crash recovery process. During
the process the platform can request a ram dump
from the wil driver as well as control when firmware
recovery will start. This interface allows the platform
to implement a more advanced crash recovery, for
example to reset dependent subsystems in proper order, or
to provide its own notifications during the recovery process.
Signed-off-by: NLior David <qca_liord@qca.qualcomm.com>
Signed-off-by: NMaya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When network interface is stopping, some resources may
be already released by the network stack, and Rx frames
cause kernel OOPS (observed one is in netfilter code)

Proper solution is to drop packets pending in reorder buffer.
Signed-off-by: NHamad Kadmany <qca_hkadmany@qca.qualcomm.com>
Signed-off-by: NVladimir Kondratiev <qca_vkondrat@qca.qualcomm.com>
Signed-off-by: NMaya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Use SMPS disabled as default because FW does not indicate
any support of SMPS.

This change will help STAs out that don’t support SMPS from
sticking on 1SS, since they don’t have method to change it
back to multiple chains.

This change also should not affect power consumption of STAs
supporting SMPS, because they are capable to switch the mode
to dynamic or static either at the end of frame sequence or
by using SMPS action frame.
Signed-off-by: NPeter Oh <poh@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

"iter" is -1 at the end of the loop and not zero.  It means we don't
print a warning message.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The function can return negative values in case of error.
Its result should be then tested for such case.

The problem has been detected using proposed semantic patch
scripts/coccinelle/tests/assign_signed_to_unsigned.cocci [1].

[1]: http://permalink.gmane.org/gmane.linux.kernel/2046107Signed-off-by: NAndrzej Hajda <a.hajda@samsung.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

For devices that does not support PCI power save, force wake up
API is used. So move powersave check inside force wakeup to reduce
one level indentation.
Signed-off-by: NRajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

commit 1aaf8efb ("ath10k: disable PCI PS for QCA988X
and QCA99X0") partially reverts pci soc powersave support added by
commit 77258d40 ("ath10k: enable pci soc powersaving"). While
reverting the change, pci wake up function is called after accessing
pci registers instead of prior to access. The assumption is that chip
is woken up before accessing its registers.Though this change does not
fix any known issues, this might help to avoid unknown or low power
platform specific issues.
Signed-off-by: NRajkumar Manoharan <rmanohar@qti.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Mesh Point requires peer security key install when running
in secured mode since it's a type of peer links, otherwise peer
link will be removed due to key install failure.

MFP feature set is required to run Mesh in secured mode and
QCA988X firmware, 10.2.4.70.14-2 and above, is the only one
supporting secured Mesh at this moment.
Signed-off-by: NPeter Oh <poh@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

minor change, indenting is one tab out.
Signed-off-by: NColin Ian King <colin.king@canonical.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This patch is derived from
commit 6301566e ("ath9k: export HW random number generator"),

We evaluated the entropy of the ADC data on QCA9531, QCA9561, QCA955x,
and AR9340, and it has sufficient quality random data (at least 10 bits
and up to 22 bits of min-entropy for a 32-bit value). We conservatively
assume the min-entropy is 10 bits out of 32 bits. Thus, ATH9K_RNG_BUF_SIZE
is set to 320 (u32) i.e., 1.25 kilobytes of data is inserted to fill up
the pool as soon as the entropy counter becomes 896/4096 (set by random.c).
Since ADC was not designed to be a dedicated HW RNG, we do not want to bind
it to /dev/hwrng framework directly. This patch feeds the entropy directly
from the WiFi driver to the input pool. The ADC register output is only
used as a seed for the Linux entropy pool. No conditioning is needed,
since all the conditioning is performed by the pool itself.
Signed-off-by: NMiaoqing Pan <miaoqing@codeaurora.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When rx stopped, AR_RX_FILTER should be cleared, but in
ath9k_hw_setrxfilter(), ATH9K_RX_FILTER_CONTROL_WRAPPER will always
be set for ar9462/ar9565.

Fix this by moving the code in ath9k_hw_setrxfilter() to
ath_calcrxfilter().
Signed-off-by: NMiaoqing Pan <miaoqing@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This gives a user a chance to know why a firmware load
is failing.
Signed-off-by: NBen Greear <greearb@candelatech.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

My previous patch incorrectly reported the antenna
for 2x2 devices. It should be a mask instead of
a numeric count. This patch fixes that.
Signed-off-by: NBen Greear <greearb@candelatech.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Remove ath9k_mod_tsf64_tu() function while we could
use div_u64_rem() function.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This is useful for MCC debugging and bug fixing.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In case of MCC and AP interface, add also NOA attr
that will inform stations about absence of an AP.
There is a chance that some stations will handle
this NOA attr correctly and will know exactly when
AP is present/absent.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In case we get BSS_CHANGED_P2P_PS early, from
mac80211, we didn't set NOA timer correctly,
while p2p_ps_vif was NULL.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When adding NOA attr enable Opportunistic Power Save.
Before we calculate ctwindow but didn't enable oppps.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

After queue nullfunc for MCC case, we hit WARN_ON
in xmit.c:2398 while skb priority wasn't set.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Use u32 while ath9k_hw_gettsf32() and
ath9k_hw_gen_timer_start() require u32.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

AP/GO will aprear after NOA, wait 1ms to be sure AP
could receive/answer this frames.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

While mac80211 using null frames when connection polling,
we should queue this frames while NOA could be there, and
AP, P2P_GO could be not present.

Without this patch, with no traffic we often saw disconnections
while we try to send nullfunc when AP/GO wasn't present.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In case of low HZ before this patch we saw wrong
values in debug message. Print real timeout value.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Add/extend debug messages when chanctx used.
Signed-off-by: NJanusz Dziedzic <janusz.dziedzic@tieto.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

In __wmi_send we check if fw is ready at the beginning of the function.
While we wait for the completion of the previous command, system suspend
can be invoked and reset the HW, causing __wmi_send to read from HW
registers while it is not ready.
Taking the wmi_mutex in the reset flow when setting the FW ready bit
to zero will prevent the above race condition.
Signed-off-by: NMaya Erez <qca_merez@qca.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The rates in the tx control rateset do not have the protection
flags applied, so RTS/CTS would never get enabled if requested.

Fix by using the rate flags in the rates returned by
ieee80211_get_tx_rates().
Signed-off-by: NBob Copeland <me@bobcopeland.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

An error message ath6kl_htc_rxmsg_pending_handler isn't appropate for when
the error is ECANCELED. This could be the result of a perfectly appropriate
RX cancel due to shutdown or suspend. This allows the right cleanup to
continue, but without an alarming error message in this particular case.
Signed-off-by: NSteve deRosier <steve.derosier@lairdtech.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The three eeprom implementations had quite some duplicate code when it
came to endianness swapping.
Additionally there was a bug in eeprom_4k and eeprom_9287 which
prevented the endianness swapping from working correctly, because the
swapping code was guarded within an "if (!ath9k_hw_use_flash(ah))". In
eeprom_def this check did not exist, so it seems that eeprom_def was the
only implementation where endianness swapping worked.

This patch takes the duplicate code and moves it from eeprom_* to
eeprom.c. The new code is derived from eeprom_def, while taking into
account the specifics from the other implementations.
Signed-off-by: NMartin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Coherent memory is more expensive to allocate (and constrained on some
architectures where it has to be pre-allocated). It is also completely
unnecessary, since the host has no reason to even access these allocated
memory spaces
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

The default values of max_num_amsdu / max_num_amdpu is assigned a
default value as part of 'ath10k_core_init_firmware_features'
Signed-off-by: NMohammed Shafi Shajakhan <mohammed@qti.qualcomm.com>
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

When debugging firmware problems it's useful to check checksums of each
component. Add a debugfs interface to retrieve crc32 checksums:

# cat /sys/kernel/debug/ieee80211/phy0/ath10k/fw_checksums
firmware-N.bin          cc3fb466
athwlan                 c0089f21
otp                     f3efeb4f
codeswap                00000000
board-N.bin             bebc7c08
board                   bebc7c08
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

To detect cases if the firmare or board file is corrupted or otherwise modified print crc32 value
of both. Now the output looks like:

ath10k_pci 0000:02:00.0: pci irq msi interrupts 1 irq_mode 0 reset_mode 0
ath10k_pci 0000:02:00.0: qca99x0 hw2.0 target 0x01000000 chip_id 0x003801ff sub 168c:0002
ath10k_pci 0000:02:00.0: kconfig debug 1 debugfs 1 tracing 1 dfs 1 testmode 1
ath10k_pci 0000:02:00.0: firmware ver 10.4.1.00030-1 api 5 features no-p2p crc32 d2901e01
ath10k_pci 0000:02:00.0: board_file api 1 bmi_id 1:1 crc32 7e56fd07
ath10k_pci 0000:02:00.0: htt-ver 2.2 wmi-op 6 htt-op 4 cal otp max-sta 512 raw 0 hwcrypto 1
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

Earlier we printed all the info messages after a successful device
initialisation and firmware boot, but that's problematic if something goes
wrong and there's no easy way to know what firmware version was used and so on.

Split the info messages into smaller pieces and print them as soon as we have
the info available.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

That way we can split the messages to be printed in different stages of
device initialisation.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

This is to make it easier to split them later and also make room for crc32
checksums.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>

We check board id from all board types, not just qca99x0, so the error message
was misleading.
Signed-off-by: NKalle Valo <kvalo@qca.qualcomm.com>