When refilling the FIFO with more TX data (using a new
SPI transaction), the driver must prefix the TX data with
a write to the FIFO I/O Register.  This tells the trf7970a
that the following data is destined for the FIFO so it can
be transmitted.

To accomplish this, the driver cannot simply push the
prefix data just before the next set of TX data that
is to be transmitted because that will overwrite part
of the TX data provided by the digital layer.  Instead,
separate the prefix data and the TX data when calling
trf7970a_transmit().  trf7970a_transmit() can then send
the prefix and TX data from different memory locations
with one spi_sync() operation.  This also means that
the driver doesn't require any skb "tx_headroom" as
provided by the digital layer (see
nfc_digital_allocate_device() and digital_skb_alloc()).

Also ensure that the prefix is of type 'u8' and not
'char'.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

It is no longer necessary to reset the FIFO and
read the 'RSSI Levels and Oscillator Status Register'
so remove that code.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

The trf7970a occasionally generates spurious interrupts
which can confuse the driver.  To help alleviate this,
clear any interrupts by reading the 'IRQ Status Register'
before starting a new transaction.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Some of the timeout values used in the driver
are not long enough to handle worst-case scenarios
so they need to be recalculated.

The time to wait for the FIFO to drain past the
low-watermark is now 20 ms because it can take
around 14.35 ms to send 95 bytes (127 bytes in
full FIFO minus 32 bytes where the low-watermark
interrupt will fire).  95 bytes will take around
14.35 ms at 6.62 kbps (the lowest supported bit
rate used by ISO/IEC 15693) so 20 ms should be a
safe value.

The time to wait before issuing an EOF to complete
an ISO/IEC 15693 write or lock command is 40 ms--
20 ms to drain the FIFO and another 20 ms to ensure
the wait is long enough before sending an EOF.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

When turning on the RF field, the driver must wait
an RF-technology-specific amount of time (known as
the guard time) before modulating the field.
Currently, the driver waits 5 ms but that is too
short for NFCF and too long for ISO/IEC 15693.
Fix this by determining the guard time when the
RF technology is set and delaying that amount
of time when turning on the RF field.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Currently, the trf7970a is reset & initialized only
when the pm_runtime resume hook is called.  Instead,
initialize it every time the RF is enabled to ensure
that the trf7970a is quiesced and in a known state
before being set up for another RF technology.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Currently, support for providing the external
SYS_CLK signal on pin 27 is not supported so
turn it off by writing to the 'Modulator and
SYS_CLK Control' register immediately after
reset.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

There is no need to sleep for 1-2 ms before
transmitting a new command.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Despite what the manual says, the FIFO size
on the trf7970a is really 127 bytes so make
the code respect that.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

The Overflow bit in the 'FIFO Status Register' has
proven to be untrustworthy so ignore it.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

To more efficiently handle long continuous reads,
use spi_sync() instead of spi_write_then_read().
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Call spi_setup() to configure SPI communication
with the trf7970a.  This will ensure that the
correct SPI parameters are used.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Currently, the digital layer 'tg_listen_mdaa'
hook is not used and it isn't necessary to have
a stub routine so remove it.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

of_property_read_u32() does not take a reference
to the specified OF node so don't call of_node_put()
in trf7970a_get_autosuspend_delay().
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

NCI RF FRAME interface is used for all kind of tags
except ISODEP ones. So for all other kind of tags the
status byte has to be removed.
Signed-off-by: NVincent Cuissard <cuissard@marvell.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

All contexts have to be initiliazed before calling
nfc_register_device otherwise it is possible to call
nci_dev_up before ending the nci_register_device
function. In such case kernel will crash on non
initialized variables.
Signed-off-by: NVincent Cuissard <cuissard@marvell.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Update nci.h to respect latest NCI specification proposal
(stop using proprietary opcodes). Handle ISO15693 parameters
in NCI_RF_ACTIVATED_NTF handler.
Signed-off-by: NVincent Cuissard <cuissard@marvell.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Make the labels for the two gpio enable lines
more user friendly by prefixing them with the
driver name.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

Current versions of the trf7970a have an erratum where
the EN2 pin must be kept low.  If it isn't kept low,
it will generate an RF field even when in passive
target mode.  To work around this issue, create the
'en2-rf-quirk' device tree property to indicate that
the trf7970a that the driver is using has this erratum.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

A bug has been discovered in the trf7970a where it
will generate an RF field even in passive target
mode when EN2 is asserted.  To work around this,
add support for the 'en2-rf-quirk' device tree
property which indicates that EN2 must remain low.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

The mechanism for specifying that the trf7970a
being used by the driver has the "IRQ Status Read"
erratum has been changed to a device tree property
('irq-status-read-quirk').

Document the new device tree property.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

The quirk indicating whether the trf7970a has
the "IRQ Status Read" erratum or not is currently
implemented using the 'driver_data' member of the
'spi_device_id' structure.  That requires the
driver to be modified to turn the quirk off when
a version of the trf7970a that doesn't have the
erratum is being used.  To fix that, create a
new device tree property called
'irq-status-read-quirk' that indicates that the
trf7970a being used has the erratum.

While at it, rename 'TRF7970A_QUIRK_IRQ_STATUS_READ_ERRATA'
to 'TRF7970A_QUIRK_IRQ_STATUS_READ' to make it
less of an eyesore.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

The trf7970a driver recently had support for the
'vin-voltage-override' property added to it.  This
property is used to override the value given by
the regulator subsystem for the VIN pin's voltage.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

The trf7970a driver uses the voltage from the
power/regulator subsystem to determine what the
voltage on the VIN pin is.  Normally, this is
the right thing to do but sometimes the board
that the trf7970a is on may change the voltage.
This is the case for the trf7970atb board from
Texas Instruments where it boosts the VIN voltage
from 3.3V to 5V (see http://www.ti.com/tool/trf7970atb).

To handle this, add support for the 'vin-voltage-override'
device tree property which overrides the voltage value
given by the regulator subsystem.  When the DT property
is not present, the value from the regulator subsystem
is used.

The value of 'vin-voltage-override' is in uVolts.
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

In order to operate at the fasted bit rate
possible, add initiator-side support for
PSL REQ while in P2P mode.  The PSL REQ
will switch the RF technology to 424F
whenever possible.
Reviewed-by: NThierry Escande <thierry.escande@linux.intel.com>
Tested-by: NThierry Escande <thierry.escande@linux.intel.com>
Signed-off-by: NMark A. Greer <mgreer@animalcreek.com>
Signed-off-by: NSamuel Ortiz <sameo@linux.intel.com>

According to RCU_INIT_POINTER()'s block comment 3.a, it can be used if
"3.   The referenced data structure has already been exposed to readers either
at compile time or via rcu_assign_pointer() -and-
 a.   You have not made -any- reader-visible changes to this structure since
then".

This case fulfills the conditions above because between the rcu_dereference()
call (cvif = rcu_dereference(ar->beacon_iter);) and the rcu_assign_pointer()
call there is no update of the "cvif" variable.
Therefore, this patch makes the replacement.

The following Coccinelle semantic patch was used:
@@
identifier v;
@@

v = rcu_dereference(...);
... when != rcu_dereference(...);
    when != v = ...;
    when != (<+...v...+>)++;
    when != \(memcpy\|memset\)(...);
(
- rcu_assign_pointer
+ RCU_INIT_POINTER
  (..., v);
|
 if(...) {
... when != v = ...;
- rcu_assign_pointer
+ RCU_INIT_POINTER
  (..., v);
... when any
 }
)

Because there are cases where between a “rcu_dereference()” call and a
“rcu_assign_pointer()” call might be updates of the value that interests us,
the Coccinelle semantic patch ignores them and replaces with
"RCU_INIT_POINTER()" only when the update is not happening.
Signed-off-by: NAndreea-Cristina Bernat <bernat.ada@gmail.com>
Acked-by: NChristian Lamparter <chunkeey@googlemail.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Setup the channel context correctly. Not doing this was
causing beacon loss in a P2P-GO/STA concurrent setup.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Signed-off-by: NCanek Peláez Valdés <canek@ciencias.unam.mx>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

If a new context is being added in addition to the current one,
then send the ASSIGN event to abort a running scan since
the addition of a context is usually followed by VIF
assignment and further operations.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

There is no reason why managed/p2p interfaces have to
be limited to one. IBSS is the only type that needs
a restriction.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This will be useful in handling addition/change of new
channel contexts.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Move it inside a CONFIG_ATH9K_CHANNEL_CONTEXT ifdef
since it is not needed otherwise.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

This was introduced in an earlier patch to handle
a compilation warning, but since the channel context
code has been mostly isolated, this is not required now.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

chanctx_work and next_chan are required only when
CONFIG_ATH9K_CHANNEL_CONTEXT is enabled.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

Finally move the 'offchannel' instance in ath_softc
inside a CONFIG_ATH9K_CHANNEL_CONTEXT cage. The offchannel
usage in ath9k_calculate_iter_data() is closed off with
an ifdef for now, since the state/opmode calculation is
common for both the channel context mode and the normal mode.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>

ath9k_vif_iter() was earlier used as an iterator
routine when calling a mac80211 utility. This is no
longer the case and hence we can mention the argument
type explicitly.
Signed-off-by: NSujith Manoharan <c_manoha@qca.qualcomm.com>
Signed-off-by: NJohn W. Linville <linville@tuxdriver.com>