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: Eyal Shapira <eyal@wizery.com>
Signed-off-by: Luciano 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: Luciano Coelho <coelho@ti.com>

The new PG2 version of the chip has a few differences in terms of FW
API if compared to PG1.  PG1 is just a sample that shouldn't be used
in real life, so to avoid having to handle both separately, mark the
PG1 version as deprecated and bail out during probe.

Signed-off-by: Luciano Coelho <coelho@ti.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: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano 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: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano 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: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano 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: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Set a flag and after the first read/write failure is encountered.
This flag will disallow further SDIO read/writes until op_stop() is
executed, which will clear all flags.

This prevents further errors from occurring, since one error usually
indicates that IO operations won't work anymore until the chip is
rebooted.  By blocking more calls, we avoid extra timeouts and having
to wait for them to occur.

[Added second paragraph explaining why the change is needed. -- Luca]

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

We wish to postpone suspend if recovery is pending. This will make sure
the FW is in a good state and perform wowlan wakeup.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Applications running in the user space needs access to the
memory of the chip. Examples of such access
- read/write global variables
- access to firmware log
- dump memory after firmware panic event

Arbitrary 4-bytes aligned location can be accessed by
read/write file wlcore/mem

[Check return value of wlcore_raw_read/write and wlcore_set_partition
calls as required by the recent IO changes. -- Luca]

Signed-off-by: Arkady Miasnikov <a-miasnikov@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

All io functions' return values should be propagated and handled. Add a
__must_check annotation to verify that the return values are checked and
to avoid future mistakes.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Propagate errors from wl1271_read_hwaddr. This function is only used
when reading the FW log (following a recovery), so don't read the FW log
in case of a bus error.

Also rename prefixes of wlcore functions which their prototypes had to
be changed.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Propagate errors from wl1271_raw_write32 and request for recovery when
appropriate.
Also rename prefixes of wlcore functions which their prototypes had to
be changed.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Propagate errors from wl1271_raw_read32. Since the read functions had no
way of returning errors in-band, change their prototypes.
Also rename prefixes of wlcore functions which their prototypes had to
be changed.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Propagate errors from wl1271_write and request for recovery when
appropriate.
Also rename prefixes of wlcore functions which their prototypes had to
be changed.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Propagate errors from wl1271_read and request for recovery when
appropriate.
Also rename prefixes of wlcore functions which their prototypes had to
be changed.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wlcore_raw_read_data is called when the FW status is read which happens
while handling interrupts and when the FW log is read following a
recovery. Request a recovery in the former case, and don't read the FW
log in case the FW status read failed.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Make wl1271_raw_write and wl1271_raw_read return errors so the driver
could handle these appropriately.
Since the prototype has changed, also rename the prefix of these
functions to wlcore.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

While bus operations may fail, either due to HW or FW issues, these are
never propagated to higher layers. As a result, the core driver has no
way of knowing that the operations failed, and will only recover if high
level logic requires it (e.g. no command completion).

Change read/write bus operations to return errors to let higher layer
functionality handle these.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

In case a recovery is initiated, the FW can no longer be trusted, and
the driver should not handle any new FW events.

Disable the interrupt handler when a recovery is scheduled and balance
it back in the op_stop callback.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

wl12xx_sdio_power_off() manually powers down the card regardless of the
runtime pm state. If wl12xx_sdio_power_on() is called before the card
was suspended by runtime PM, it will not power up the card.

As part of the HW detection, the chip's power is toggled. Since this
happens in the context of probing sdio, the power reference counter will
be higher than zero. As a result, when wl12xx_sdio_power_off() is
called, the chip will be powered down while still having a positive
power reference counter. If the interface is quickly activated, the
driver might try to transfer data to a powered off chip.

Fix this by ensuring that wl12xx_sdio_power_on() explicitly powers on
the chip in case runtime pm claims the chip is already powered on. To
avoid cases in which it is not possible to determine if the chip was
really powered on (card's power reference counter is positive), operate
on the mmc_card instead of the function.

Also verify that the chip is indeed powered on before powering off, to
avoid wrong reference counter values in error cases.

Signed-off-by: Ido Yariv <ido@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

After 934b9d1e (wl18xx: avoid some -Wformat warnings) there was still
a warning with (at least) ARM gcc version 4.4.1:

drivers/net/wireless/ti/wl18xx/main.c: In function 'wl18xx_conf_init':
drivers/net/wireless/ti/wl18xx/main.c:1026: warning: format '%ld' expects type 'long int', but argument 2 has type 'unsigned int'

Fix this by using %zu for the both formats, since the fw->size and the
macro (derived from sizeof()) are size_t.

Signed-off-by: Luciano Coelho <coelho@ti.com>

Remove the cap IEEE80211_HT_CAP_DSSSCCK40 from the 5Ghz variant of
the siso40 HT capabilities. It is meaningless in 5Ghz.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Introduce a default set of HT capabilities that are set according to the
number of antennas on the board. Move the HT setting code down to allow
the number of antennas to be set (and optionally overridden) before it.

Remove the "mimo" HT option, since the default mode now enables MIMO is
possible.

Use this opportunity to add a helper function for setting HT
capabilities and reduce the volume of the code a bit.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

The 18xx chip does not support MIMO in 5Ghz. Use the siso20 HT cap as
fallback in 5Ghz when "mimo" is requested.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

FW uses a few memory blocks as a buffer to accumulate FW logs before
transmitting them to the host over SDIO. When FW WatchDog recovery
occurs, the last FW traces are still pending in the buffer. Driver is
to read these FW traces whether log mode is continuous or on demand.

FW memory blocks allocated for the log buffer are handled as a link list:
the first 4 bytes in each memory block contain FW address to the next block.
The end of list condition depends on FW log mode:
- on demand: the list is cyclic, the next address is equal to the first address
- continuous: the address is  equal to 0x2000000

Log data resides inside FW memory block with offset depending on
logger mode:
- on demand:  4 bytes (address of the next memory block)
- continuous: 4 bytes and Rx Descriptor structure size

Described FW logger API is backward compatible with previous FW versions.

Signed-off-by: Igal Chernobelsky <igalc@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

The get survey op expects the low level driver to report
the noise level for a a given channel.

The noise calculated in wlcore is (rssi-snr/2), but since
the snr reported by the FW is a derivative from the rssi
this calculation is useless, and should not be reported
to the user space.

Reporting incorrect noise, results in the wpa_supplicant
miscalculating the roaming candidate priority, thus causing
a situation where an AP with a lower rssi level would be
chosen over a better AP.

Signed-off-by: Yoni Divinsky <yoni.divinsky@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Before patch b5d6d9b2 (wlcore/wl12xx/wl18xx: don't use TX align quirk
for wl127x), this was automatically set for all platforms. As this
should now be set explicitly, set it for PG2 as well.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

New PG2 firmwares have additional Rx descriptors.

Add a module parameter to manually set the number of Rx descriptors for
older versions (PG1). We cannot discriminate based on chip-id, since
this value must be set on probe.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

The sleep_auth value of the last interface to be set up prevailed when
an interface was removed. Take care of this by correctly configuring the
value according to the remaining STA/AP interfaces.

Take this opportunity to refactor the sleep_auth setting code for better
readability.

[Small style fix. -- Luca]

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Hold a value for sta_sleep_auth that is amenable to change by debugfs.
When detecting a legal value in this variable on interface init, use it
as an override value for sleep_auth.

This makes debugging more intuitive using the debugfs value.

Increment the conf version since we added an element to the conf
structure.

Note: An AP going up will always set sleep_auth to PSM_CAM.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

For FW debugging purposes, we may need to change the sleep mode
(aka. sleep_auth) by hand, and set it to the mode we want.  To allow
this, a debugfs entry is added.

Now we store the sleep_auth value that has been set and use that
instead of the quirk to decide whether we should enter ELP or not.

Signed-off-by: Luciano Coelho <coelho@ti.com>
Signed-off-by: Arik Nemtsov <arik@wizery.com>

The ampdu_action() function is called on the reconfig() path to remove
existing Rx BA sessions. Since these don't exist for the low level
driver, we output an error message. Turn the message into a debug
message for now, until the mac80211 reconfig flow is changed.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Advertise to the stack that the wlcore driver
supports multiple interfaces for a single device.
This is required in order to be able to run
multirole with mac80211.

Signed-off-by: Eliad Peller <eliad@wizery.com>
Signed-off-by: Yoni Divinsky <yoni.divinsky@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

FEM BIP calibration may fail with fw/phy radio status. In order to
recognize these failures a log is added to the calibration answer
(TEST_CMD_P2G_CAL)

Signed-off-by: Yair Shapira <yair.shapira@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Add support for HP (High Performance TQS fem type 3) and SKW
(fem type 2). This is done by increasing the number of FEM
manufacturers to 4.

Usually FEM parameters from ini file are read from nvs file and
passed to firmware using TEST_CMD_INI_FILE_RADIO_PARAM. Still,
because the nvs file has only place for 2 FEMs, we need to pass the
new FEM types information in one of the available entries.

This is done by mapping new fem types 2,3 to entry 0. This solution
works for manual FEM selection. AutoDetect-FEM still support only
fem types 0 and 1.

Signed-off-by: Yair Shapira <yair.shapira@ti.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

On channel switch we have to update the basic rates, in
order to reflect possible band changes (otherwise, we
might start beaconing on 11a with the default rates
of 11g).

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

EAPOLs are sent at high rates as they are considered
data packets. Some APs like Motorola Symbol AP7131 and AP650
don't respond well to these rates and don't respond with
EAPOL 3/4 consistently. When sending EAPOL 2/4 at 54Mbps
we've seen approx 30% success rate in getting EAPOL 3/4 response
while using 11Mbps we got 100% success.
To increase the chances of successful 4-Way handshake with
such APs, send EAPOLs with basic rate policy in order to avoid
high rates.

Signed-off-by: Eyal Shapira <eyal@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

NET_IP_ALIGN can be overriden on different architectures
and therefore cannot be used in the RX path to account
for the 2 bytes added for alignment (either by the FW
in the case of 18xx or by the host for 12xx).
Instead use an internal define.

Signed-off-by: Eyal Shapira <eyal@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>

Don't spew errors when we can't find the NVS file in wlcore. Instead
fail the wl12xx boot HW op if the NVS isn't found.

Take this opportunity to remove some dead code from register_hw()
which looks for the NVS again needlessly.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
Signed-off-by: Luciano Coelho <coelho@ti.com>