When a function requests to recover, it would normally abort and will
not send any additional commands to the HW. However, other threads may
not be aware of the failure and could try to communicate with the HW
after a recovery was queued, but before the recovery work began.

Fix this by introducing an intermediate state which is set when recovery
is queued, and modify all state checks accordingly.
Signed-off-by: NIdo Yariv <ido@wizery.com>
Signed-off-by: NLuciano Coelho <luca@coelho.fi>

When a single fw is being used for both single-role
and multi-role cases (e.g. 18xx), wl->mr_fw_name is
NULL, which results in NULL dereference while trying
to load the multi-role fw.

In this case, always use the single-role fw, and avoid
redundant fw switch by checking for this case in
wl12xx_need_fw_change() as well.
Signed-off-by: NEliad Peller <eliad@wizery.com>
Signed-off-by: NLuciano Coelho <luca@coelho.fi>

Declare any firmware that might be used by this driver.
If all drivers declare their firmware usage, then a sufficiently
complete list of firmware files can then be used to pare down
the external linux-firmware package to just the files in actual use.

Cc: Luciano Coelho <coelho@ti.com>
Cc: linux-wireless@vger.kernel.org
Cc: netdev@vger.kernel.org
Signed-off-by: NTim Gardner <tim.gardner@canonical.com>
Signed-off-by: NLuciano Coelho <luca@coelho.fi>

Remove the control.sta pointer from ieee80211_tx_info to free up
sufficient space in the TX skb control buffer for the upcoming
Transmit Power Control (TPC).
Instead, the pointer is now on the stack in a new control struct
that is passed as a function parameter to the drivers' tx method.
Signed-off-by: NThomas Huehn <thomas@net.t-labs.tu-berlin.de>
Signed-off-by: NAlina Friedrichsen <x-alina@gmx.net>
Signed-off-by: NFelix Fietkau <nbd@openwrt.org>
[reworded commit message]
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

avoid using the skb CB for getting the appropriate sta.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Acked-by: NLuciano Coelho <coelho@ti.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

There was a bug hiding here since the hlid was sometimes inferred from
the sta, which might be invalid at this point.

Instead, propagate the hlid from the skb-queue where we got the skb
in the first place.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Acked-by: NLuciano Coelho <coelho@ti.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

In some rare cases, the CMD_ROC completion may take over 1 second.
The timeout had earlier been increased to 1000ms (from 750ms), but it
is still not enoug.  Increase it to 1500ms.
Signed-off-by: NLuciano Coelho <coelho@ti.com>

We need to wait for the command completion event when we send the
CMD_ROLE_STOP event otherwise we may try to send CMD_ROLE_START too
soon and get out-of-sync with the firmware.

In some cases, the firmware may not send the event, so we wait for the
event or for the timeout, whichever comes first.

This patch is based on an earlier version by Eliad.

Cc: Eliad Peller <eliad@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

handle errors of nla_put() inside the if(nla_put...) {}

This makes the code simpler and clearer because:
we take advantage from the fact that we have only one nla_put
in our routines (so no real need for goto label).
this avoids ugly goto forward followed by goto backward.
Signed-off-by: NYair Shapira <yair.shapira@ti.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

this completes the calibrator based fem detect logic in driver:
driver starts (by calibrator) in plt_mode PLT_FEM_DETECT
wlcore inits and starts plt on wl12xx
wl12xx fetches fem number from firmware and stores it in wl->fem_manuf
wl12xx immediatly returns (doesn't start radio, etc...)
wlcore returns the fem_manuf to calibrator using WL1271_TM_ATTR_DATA
plt_mode is stopped
Signed-off-by: NYair Shapira <yair.shapira@ti.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

add wl->plt_mode that is used to indicate different plt
working modes: this will be used to implement calibrator side
auto fem detection where driver asks firmware to detect
the wlan fem radio type and returns it to calibrator.

this is not implemented yet and plt_modes: PLT_ON and
PLT_FEM_DETECT currently behave the same.
Signed-off-by: NYair Shapira <yair.shapira@ti.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

After the latest mac80211 changes, the sta has
the ap's sta pointer even before association.

This cause the auth and assoc frames to be sent
with the standard ap's rates, rather than the
basic rates.

Change the tx rate policy logic to use the regular
ap rates only for data packets (so control and mgmt
packets will be sent with basic rates)
Signed-off-by: NEliad Peller <eliad@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

During interface removal, don't adjust sleep_auth if we are during
recovery. Since the FW is potentially dead we shouldn't talk to it.
Reported-by: NYossi Wortzel <yossiw@ti.com>
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

Send EAPOLs using minimum basic rate for AP, STA, p2p GO and Client.
The patch fixes p2p connection issue with Realtek device in p2p
certification test 5.1.13 (DEVUT reinvokes Persistent Group).
Signed-off-by: NIgal Chernobelsky <igalc@ti.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

suspend and resume callbacks configure wakeup conditions to the FW
which may be different between suspend and resume.
This feature is currently not utilized as both in suspend and resume
FW wakeup every 1 DTIM. Avoid waking up the chip and doing the FW command
unless there's an actual difference in the wakeup conditions.
Signed-off-by: NEyal Shapira <eyal@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

We should free "chunk" here before returning the error code.
Signed-off-by: NDan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

commit 587cc286 ("wlcore: compare ssid_len before comparing
ssids") introduced a new bug - the ssid length from the
request struct was compared against the ssid length of
another request, instead the one of the cmd.

This might cause the sched scan request to fail
(with -EINVAL) in many cases.
Signed-off-by: NEliad Peller <eliad@wizery.com>

Avoid using the IEEE80211_NUM_BANDS constant for arrays sizes etc, as
this can contain bands unsupported by the driver (e.g. 60Ghz). Use an
internal constant to determine the number of bands.
Signed-off-by: NArik Nemtsov <arik@wizery.com>

If some IO read/write fails while the FW is not loaded, a recovery
will not take place. This means the SDIO_FAILED flag will stay in place
forever and prevent further read/writes.

This can happen if a check for STATE_OFF was forgotten in some routine.

Take this opportunity to rename the flag to IO_FAILED, since we support
other buses as well.
Reported-by: NIdo Yariv <ido@wizery.com>
Signed-off-by: NArik Nemtsov <arik@wizery.com>

wlcore needs to wait for certain events for example
for roc complete event. Usually the events are received
from the FW very fast, therefore wlcore can poll with
a short delay and if after a second the event was
not received yet poll with a long (1-5 msec) delay.

This implementation is similar to the sending of
commands to the FW.

Empirically the change reduced the wait for roc event
from ~10-40msec to 100s of usecs.

[replace udelay/msleep with usleep_range - Arik]
Signed-off-by: NYoni Divinsky <yoni.divinsky@ti.com>
Signed-off-by: NArik Nemtsov <arik@wizery.com>

The interrupt line is disabled in op_stop using disable_irq. Since
pending interrupts are synchronized, the mutex has to be released before
disabling the interrupt to avoid a deadlock with the interrupt handler.

In addition, the internal state of the driver is only set to 'off'
after the interrupt is disabled. Otherwise, if an interrupt fires after
the state is set but before the interrupt line is disabled, the
interrupt handler will not be able to acknowledge the interrupt
resulting in an interrupt storm.

The driver's operations might be called during recovery. If these
acquire the mutex after it was released by op_stop, but before the
driver's state is changed, they may queue new work items instead of just
failing. This is especially problematic in the case of scans, in which a
new scan may be scheduled after all scan requests were cancelled.
Signed-off-by: NIdo Yariv <ido@wizery.com>
Signed-off-by: NArik Nemtsov <arik@wizery.com>

implement the .flush() callback by simply calling wl1271_tx_flush().
Signed-off-by: NEliad Peller <eliad@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

We have some API changes and new features in the new firmwares that
are not compatible with older drivers.  Increase the version of the FW
filenames for wl12xx to 5.

Additionally, remove the duplicate definitions from wlcore_i.h and
remove the MODULE_FIRMWARE macro calls from the SDIO and SPI modules,
since they're irrelevant there.
Signed-off-by: NLuciano Coelho <coelho@ti.com>

The driver configures the firmware template for probe requests during
the scan process.  If the same template is used for one-shot and sched
scans they will override each other when running scans simultaneously.

This fix works only on firmwares later than X.3.9.2.112 for single
role and X.3.9.2.23 for multi-role.

[Some cleaning-up and renaming of the quirk to something smaller --
Luca.]
Signed-off-by: NYoni Divinsky <yoni.divinsky@ti.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

If recovery is called when the FW is off, we should clear the recovery
flag. Otherwise we risk booting the driver in permanent pending-recovery
state.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

Don't read the FW panic log or print other debug data when recovery is
intended (i.e. FW type switch). This takes valuable time and can be
confusing to the user.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

This command is buggy (doesn't take the mutex) and unused. Instead, the
"start_recovery" file is used for the same purpose. Remove the code but
keep the command constant to avoid breaking the testmode ABI.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

If a Tx queue is currently stopped because of our Tx watermark flow
control, don't stop it again. This causes a warning to appear.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

Refuse to boot if the FW version is too old. The minimum version is set
per chip, with the option of setting it per PG in the future.

When boot fails because of an old FW, display a helpful message.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

New wl12xx firmware supports scheduled scans also while connected.
Stop blocking sched scan requests when connected and add a quirk to
block in hardware that don't support it (currently wl18xx doesn't).

This requires FW version 6/7.3.10.2.112 for single-role and
6/7.5.6.0.25 for multi-role.
Signed-off-by: NVictor Goldenshtein <victorg@ti.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

When CONFIG_PM is disabled, no device can possibly
support WoWLAN since it can't go to sleep to start
with. Due to this, mac80211 had even rejected the
hardware registration. By making all the code and
data for WoWLAN depend on CONFIG_PM we can promote
this runtime error to a compile-time error.

Add #ifdef around all WoWLAN code to remove it in
systems that don't need it as they never suspend.

Cc: Kalle Valo <kvalo@qca.qualcomm.com>
Acked-by: NLuciano Coelho <coelho@ti.com>
Signed-off-by: NJohannes Berg <johannes.berg@intel.com>

As recovery queuing can now occur from multiple code paths
it's convenient to know what triggered it in all cases
other than an intended recovery which is part of the
switch between single role to multi role.
Signed-off-by: NEyal Shapira <eyal@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

Following the addition of propagating errors from the bus ops
there's a need to distinguish between bus errors (including timeout)
and a legitimate timeout occuring in cmd_wait_for_event_or_timeout.
In case of real bus errors we need to queue recovery even in cases
where a timeout on a response from the FW to a command is acceptable.
Reported-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NEyal Shapira <eyal@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

commit 14bba17b "wl12xx: Propagate errors from wl1271_raw_write32"
breaks down TX in certain scenarios. wl1271_irq_locked() propagates
errors from wl1271_tx_work_locked however it may return -EBUSY
when the FW queues are full which is a legitimate case and not a
a real error. In this case a recovery is triggered by wl1271_irq
and this keeps repeating itself so TX is completely broken.
Fix it by avoiding propagating return values as errors even if they
aren't. Only bus (SDIO or SPI) ops failures would be progagated
as only these should trigger recovery.
Signed-off-by: NEyal Shapira <eyal@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

We need to release the IRQ if hw_info() or identify_chip() fails.  And
we need unregister the HW with mac80211 if there are any failures
after it's registered.
Signed-off-by: NLuciano Coelho <coelho@ti.com>

Mainline build reports:

warning: (WL12XX) selects WLCORE which has unmet direct dependencies (NETDEVICES && WLAN && WL_TI && GENERIC_HARDIRQS && MAC80211 && INET)

The INET dependency was added in commit
3c6af5b5:
    wl1271_main.c:(.text+0x271052): undefined reference to `unregister_inetaddr_
notifier'
    wl1271_main.c:(.text+0x2714d7): undefined reference to `register_inetaddr_no
tifier'

    Driver is doing some filtering based on IP addresses...

but this driver no longer has that code and it builds fine even when
CONFIG_INET is not enabled, so drop that dependency and eliminate the
kconfig warning message.
Signed-off-by: NRandy Dunlap <rdunlap@xenotime.net>
Cc: Luciano Coelho <luciano.coelho@nokia.com>
Cc: John W. Linville <linville@tuxdriver.com>
Acked-by: NLuciano Coelho <coelho@ti.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Take the mutex early in the resume handler and use the locked version of
the IRQ routine. This ensures any recoveries queued will only take place
after resume has fully completed.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

Separate the threaded IRQ handling routine into two functions.
The outer function takes the mutex and calls recovery on errors. It also
performs a Tx-path optimization to avoid redundant works.

The inner function is simplified - all calls to recovery are removed and
it assumes the lock is taken. The locked variant will be reused elsewhere.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

Propagate some missing return values for Rx-filter related functions.
This and makes sure we always fail the suspend in case of SDIO errors.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>

If an error is detected after mac80211 is already suspended, the recovery
work will not be queued. This will leave the driver in a bad state on
resume.

Detect this in the resume op and re-queue a recovery.
Signed-off-by: NArik Nemtsov <arik@wizery.com>
Signed-off-by: NLuciano Coelho <coelho@ti.com>