This patch moves SPI controller reset out of spin lock.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

With SW CS, during the transfer completion CS is de-asserted by writing
default command1 register value to SPI_COMMAND1 register. With this both
mode and CS state are set at the same time and if current transfer mode
is different to default SPI mode and if mode change happens prior to CS
de-assert, clock polarity can change while CS is active before transfer
finishes.

This causes Slave to see spurious clock edges resulting in data mismatch.

This patch fixes this by de-asserting CS before writing SPI_COMMAND1 to
its default value so through out the transfer it will be in same SPI mode.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Convert GPIO chip selects in the Renesas MSIOF SPI driver from legacy
GPIO numbers to GPIO descriptors.

Notes:
  - The board file for the SH7724-based Ecovec24 development board now
    registers a GPIO descriptor lookup, instead of passing a GPIO number
    through controller_data,
  - sh_msiof_get_cs_gpios() must release all GPIOs, else
    spi_get_gpio_descs() cannot claim them during SPI controller
    registration.
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Mark Brown <broonie@kernel.org>

Fixes: 5fd917af ("spi: bcm2835aux: make the polling duration limits configurable")
Signed-off-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

While devm_gpiod_get_index_optional() returns NULL if the GPIO is not
present (i.e. -ENOENT), it may still return other error codes, like
-EPROBE_DEFER.  Currently these are not handled, leading to
unrecoverable failures later in case of probe deferral:

    gpiod_set_consumer_name: invalid GPIO (errorpointer)
    gpiod_direction_output: invalid GPIO (errorpointer)
    gpiod_set_value_cansleep: invalid GPIO (errorpointer)
    gpiod_set_value_cansleep: invalid GPIO (errorpointer)
    gpiod_set_value_cansleep: invalid GPIO (errorpointer)

Detect and propagate errors to fix this.

Fixes: f3186dd8 ("spi: Optionally use GPIO descriptors for CS GPIOs")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Mark Brown <broonie@kernel.org>

The conversion from GPIO numbers to GPIO descriptors removed the use of
spi->controller_data, but forgot to update a comment referring to it.

Fixes: 9b00bc7b ("spi: spi-gpio: Rewrite to use GPIO descriptors")
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Mark Brown <broonie@kernel.org>

To estimate efficiency add statistics on transfer types
(polling and interrupt) used to debugfs.
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Signed-off-by: Mark Brown <broonie@kernel.org>

In accordance with hardware specification Ver 1.0, reset register
transmission / reception setting before transfer.
Signed-off-by: Hiromitsu Yamasaki <hiromitsu.yamasaki.ym@renesas.com>
[geert: Use readl_poll_timeout_atomic()]
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Replace the open-coded loop in sh_msiof_modify_ctr_wait() by a call to
the readl_poll_timeout_atomic() helper macro.
Suggested-by: Wolfram Sang <wsa@the-dreams.de>
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Add the missing complete operations for dma_completion to fix the
problem of blocking at the wait_for_completion_interruptible()
function when use spi_slave_abort().
Signed-off-by: Clark Wang <xiaoning.wang@nxp.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Drop unused pdata copy in struct spi_gpio. No functional change
intended.
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Chris Healy <cphealy@gmail.com>
Cc: linux-spi@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>

Swap branches of the if statement in order to simplify it's logical
condition being checked. No functional change intended.
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Chris Healy <cphealy@gmail.com>
Cc: linux-spi@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>

Use a local "struct spi_bitbang *bb" in spi_gpio_probe() for
brevity. No functional change intended.
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Chris Healy <cphealy@gmail.com>
Cc: linux-spi@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>

Use a local "struct device *dev" in spi_gpio_probe() for brevity. No
functional change intended.
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Chris Healy <cphealy@gmail.com>
Cc: linux-spi@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>

Spi_to_pdata() is not used anywhere in the code. Drop it.
Signed-off-by: Andrey Smirnov <andrew.smirnov@gmail.com>
Cc: Mark Brown <broonie@kernel.org>
Cc: Chris Healy <cphealy@gmail.com>
Cc: linux-spi@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Signed-off-by: Mark Brown <broonie@kernel.org>

Under some circumstances the default 30 us polling limit is not optimal
and may lead to long delays because we are waiting on an interrupt.
with this patch we have the possibility to influence this policy.

So make this limit (in us) configurable via a module parameters
(but also modifyable via /sys/modules/...)
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Setup gpio-cs to the correct levels during setup and also make the
gpio definitely an output GPIO.

This is transparently fixing some badly configured DTs in the process
where cs-gpio is set but the gpios are still configured with native cs.

It also makes 100% sure that the initial CS levels are as expected -
especially on systems with devices on a bus with mixed CS_HIGH/CS_LOW
settings.

Fixes: 1ea29b39 ("spi: bcm2835aux: add bcm2835 auxiliary spi device...")
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

From personal bad experience (even as the author of the original driver)
it shows that native-cs is "somewhat" supported by the spi bus driver
when using a buggy device tree.

So make sure that the driver is warning in dmesg about this fact
that we are running in a not supported mode that may have surprizing
limitations.

Fixes: 1ea29b39 ("spi: bcm2835aux: add bcm2835 auxiliary spi device...")
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

The original driver by default defines num_chipselects as -1.
This actually allicates an array of 65535 entries in
of_spi_register_master.

There is a side-effect for buggy device trees that (contrary to
dt-binding documentation) have no cs-gpio defined.

This mode was never supported by the driver due to limitations
of native cs and additional code complexity and is explicitly
not stated to be implemented.

To keep backwards compatibility with such buggy DTs we limit
the number of chip_selects to 1, as for all practical purposes
it is only ever realistic to use a single chip select in
native cs mode without negative side-effects.

Fixes: 1ea29b39 ("spi: bcm2835aux: add bcm2835 auxiliary spi device...")
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Remove dead code that never can get reached, as we limit count to
a max of 3.
Suggested-by: Hubert Denkmair <h.denkmair@intence.de>
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

On long running tests with a mcp2517fd can controller it showed that
on rare occations the data read shows corruptions for longer spi transfers.

Example of a 22 byte transfer:

expected (as captured on logic analyzer):
FF FF 78 00 00 00 08 06 00 00 91 20 77 56 84 85 86 87 88 89 8a 8b

read by the driver:
FF FF 78 00 00 00 08 06 00 00 91 20 77 56 84 88 89 8a 00 00 8b 9b

To fix this use BCM2835_AUX_SPI_STAT_RX_LVL to determine when we may
read data from the fifo reliably without any corruption.

Surprisingly the only values ever empirically read in
BCM2835_AUX_SPI_STAT_RX_LVL are 0x00, 0x10, 0x20 and 0x30.
So whenever the mask is not 0 we can read from the fifo in a safe manner.

The patch has now been tested intensively and we are no longer
able to reproduce the "RX" issue any longer.

Fixes: 1ea29b39 ("spi: bcm2835aux: add bcm2835 auxiliary spi device...")
Reported-by: Hubert Denkmair <h.denkmair@intence.de>
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

This read of the fifo is a potential candidate for a race condition
as the spi transfer is not necessarily finished and so can lead to
an early read of the fifo that still misses data.

So it has been removed.

Fixes: 1ea29b39 ("spi: bcm2835aux: add bcm2835 auxiliary spi device...")
Suggested-by: Hubert Denkmair <h.denkmair@intence.de>
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Sharing more code between polling and interrupt-driven mode.
Signed-off-by: Martin Sperl <kernel@martin.sperl.org>
Acked-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Export spi_mem_default_supports_op(), so that controller drivers
can use this.
spi-mem driver already exports this using EXPORT_SYMBOL,
but not declared it in spi-mem.h.
This patch declares spi_mem_default_supports_op() in spi-mem.h and
also removes the static from the function prototype.
Signed-off-by: Naga Sureshkumar Relli <naga.sureshkumar.relli@xilinx.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Improve maintainability by converting the register bit, bitmask, and
bitfield definitions from hexadecimal constants to constructs using
BIT(), GENMASK(), or "val << shift".
Suggested-by: Simon Horman <horms+renesas@verge.net.au>
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Taking one interrupt for every byte is rather slow. Since the
controller is perfectly capable of transmitting 32 bits at a time,
change t->bits_per-word to 32 when the length is divisible by 4 and
large enough that the reduced number of interrupts easily compensates
for the one or two extra fsl_spi_setup_transfer() calls this causes.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Mark Brown <broonie@kernel.org>

Commit c9bfcb31 (spi_mpc83xx: much improved driver) introduced
logic to ensure bits_per_word and speed_hz stay the same for a series
of spi_transfers with CS active, arguing that

    The current driver may cause glitches on SPI CLK line since one
    must disable the SPI controller before changing any HW settings.

This sounds quite reasonable. So this is a quite naive attempt at
relaxing this sanity checking to only ensure that speed_hz is
constant - in the faint hope that if we do not causes changes to the
clock-related fields of the SPMODE register (DIV16 and PM), those
glitches won't appear.

The purpose of this change is to allow automatically optimizing large
transfers to use 32 bits-per-word; taking one interrupt for every byte
is extremely slow.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Mark Brown <broonie@kernel.org>

Use SPI device pointer in the remaining two error and warning prints in
pxa2xx_spi_transfer_one() instead of platform device of the controller

It make prints in the function uniform and more useful especially the
error print here as it can reveal the driver that has mapped the DMA
itself and attempts to transfer more than the maximum supported DMA
transfer length.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Pointer to a SPI device is passed to pxa2xx_spi_transfer_one() so there
is no need to access it through the current SPI message pointer.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Tegra SPI controller supports lsb first mode. Default is MSB bit first
and on selection of SPI_LSB_FIRST through SPI mode transmission happens
with LSB bit first.

This patch adds SPI_LSB_FIRST flag to mode_bits and also configures it
on request.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Fixes: Use packed mode for 32 bits per word transfers to increase
performance as each packet is a full 32-bit word.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

The comment says that we should not allow changes (to
bits_per_word/speed_hz) while CS is active, and indeed the code below
does fsl_spi_setup_transfer() when the ->cs_change of the previous
spi_transfer was set (and for the very first transfer).

So the sanity checking is a bit too strict - we can change it to
follow the same logic as is used by the actual transfer loop.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Mark Brown <broonie@kernel.org>

__spi_validate() in the generic SPI code sets ->speed_hz and
->bits_per_word to non-zero values, so this condition is always true.
Signed-off-by: Rasmus Villemoes <rasmus.villemoes@prevas.dk>
Signed-off-by: Mark Brown <broonie@kernel.org>

Fixes: SPI driver can be built as module so perform SPI controller reset
on probe to make sure it is in valid state before initiating transfer.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Fixes: Configure DMA burst size to be same as SPI TX/RX trigger levels
to avoid mismatch.

SPI FIFO trigger levels are calculated based on the transfer length.
So this patch moves DMA slave configuration to happen before start
of DMAs.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Fixes: Flush TX and RX FIFOs before start of new transfer and on FIFO
overflow or underrun errors.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Fixes: terminate DMA and perform controller reset on transfer timeout
to clear the FIFO's and errors.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Fixes: computation of actual bytes to fill/receive in/from FIFO in unpacked
mode when transfer length is not a multiple of requested bits per word.

unpacked mode transfers fails when the transfer includes partial bytes in
the last word.

Total words to be written/read to/from FIFO is computed based on transfer
length and bits per word. Unpacked mode includes 0 padding bytes for partial
words to align with bits per word and these extra bytes are also accounted
for calculating bytes left to transfer in the current driver.

This causes extra bytes access of tx/rx buffers along with buffer index
position crossing actual length where remain_len becomes negative and due to
unsigned type, negative value is a 32 bit representation of signed value
and transferred bytes never meets the actual transfer length resulting in
transfer timeout and a hang.

This patch fixes this with proper computation of the actual bytes to fill in
FIFO during transmit and the actual bytes to read from FIFO during receive
ignoring 0 padded bytes.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>

Fixes: Clear packed bit when not using packed mode.

Packed bit is not cleared when not using packed mode. This results
in transfer timeouts for the unpacked mode transfers followed by the
packed mode transfers.
Signed-off-by: Sowjanya Komatineni <skomatineni@nvidia.com>
Signed-off-by: Mark Brown <broonie@kernel.org>