Convert the pump_transfers() transfer tasklet to transfer_one() hook the
SPI core calls to process single transfer instead of handling message
processing and chip select handling in the driver. This not only
simplifies the driver but also brings transfer statistics from the core.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Current DMA engine implementation of pxa2xx_spi_dma_prepare() don't use
the dma_burst argument. Remove it since it became unused after
commit 6356437e65 ("spi: spi-pxa2xx: remove legacy PXA DMA bits").
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Convert to generalized SPI controller API introduced by the
commit 8caab75fd2 ("spi: Generalize SPI "master" to "controller"").
Inside driver variable name "master" is still used to indicate the driver
is master only.
While at it, change "unsigned cs" to "unsigned int cs" in
pxa2xx_spi_fw_translate_cs() to suppress checkpatch warning.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Transfer state machine in this driver does not need to set/unset pointer
to chip data between queueing and finalizing message as it is not
actually used as a state info itself but just pointer passing.
Since this per SPI device specific chip data is already carried in
ctldata use that and remove pointer to chip data from driver data.
While at it, group initialized variables before uninitialized variables
in pump_transfers().
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
There is no need to carry pointer to current SPI message in driver data
because cur_msg in struct spi_master holds it already when driver is using
the message queueing infrastructure from the SPI core.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
All of these variables are unconditionally set before their use.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
SPI core provides DMA mapping with scatterlists. Start using it instead
of own implementation in spi-pxa2xx. Major difference in addition to
bunch of removed boilerplate code is that SPI core does
mapping/unmapping for all transfers in a message before and after the
message sending where spi-pxa2xx did mapping/unmapping for each
transfers separately.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Dummy buffer is used for half duplex transfers that don't have TX or RX
buffer set. Instead of own dummy buffer management here let the SPI core to
handle it by setting the SPI_MASTER_MUST_RX and SPI_MASTER_MUST_TX flags.
Then core makes sure both transfer buffers are set.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
spi-pxa2xx-dma.c DMA engine implementation stopped using PIO for
unaligned trailing bytes in the commit 111e0a9dc7 ("spi/pxa2xx: Prevent
DMA from transferring too many bytes"). This means there is no need to
update tx/rx transfer buffer pointers after DMA completion. These buffer
pointers will be set to new buffers when handling the next transfer.
Because this buffer pointer update was only remaining use for rx_map_len
and tx_map_len members in struct driver_data after removing the legacy PXA
DMA implementation they can be removed now.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Mark Brown <broonie@kernel.org>
If by some reason pxa2xx_spi_dma_prepare() fails we have not to ignore its
error. In such case we abort the transfer and return the error to upper
level.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
[Jarkko: Avoid leaking TX descriptors in case RX descriptor allocation
fails. Noted by Robert Jarzmik <robert.jarzmik@free.fr>.
Unmap also buffers in case of failure.]
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Mark Brown <broonie@kernel.org>
Commit b36f09c3c4 ("dmaengine: Add transfer termination
synchronization support") marked dmaengine_terminate_all() as
deprecated and is being replaced by explicit synchronous and asynchronous
terminate functions.
Here DMA termination are done in two cases: FIFO overrun and module
removal.
FIFO overrun is handled in interrupt context and converting
dmaengine_terminate_all() to dmaengine_terminate_async() does the same than
before.
Using synchronous termination in module removal however adds a bit more
robustness as it waits all completion callbacks have finished. Although it
looks all known DMA engines used with spi-pxa2xx don't implement
device_synchronize() callback so this too appears to be a no-op in
practice.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Mark Brown <broonie@kernel.org>
This was leftover from the legacy pxa2xx DMA implementation and not needed
anymore so remove it.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Currently SSP registers are accessed by having an own read and write macros
for each register. For instance read_SSSR(iobase) and write_SSSR(iobase).
In my opinion this hurts readability and requires new macros to be defined
for each new added register. Let's define and use instead common
pxa2xx_spi_read() and pxa2xx_spi_write() accessors.
Signed-off-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
That field has been deprecated in favour of getting the necessary
information from ACPI/DT.
However, we still need to deal systems that are PCI only (no ACPI to back
up). In order to support such systems, we allow the DMA filter function and
its corresponding parameter via pxa2xx_spi_master platform data. Then when
the pxa2xx_spi_dma_setup() doesn't find the channel via ACPI, it falls back
to use the given filter function.
Suggested-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Mark Brown <broonie@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
This is to fix the SPI DMA transfer failure for speed less than 1M.
If using current DMA burst size setting (16), the Rx data bytes are
invalid due to each data byte is multiplied according to the burst
size setting.
Let's said supposedly we shall receive the following 18 bytes of data:
01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 17 18
Instead, the data bytes received consist of "16 bytes of '01' +
2 bytes of '02'" :
01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 01 02 02
Signed-off-by: Chew, Chiau Ee <chiau.ee.chew@intel.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Mark Brown <broonie@linaro.org>
In case we are doing DMA transfer and the size of the buffer is not multiple
of 4 bytes the driver truncates that to 4-byte boundary and tries to handle
remaining bytes using PIO.
Or that is what it tried to do. What actually happens is that it calls
ALIGN() to the buffer size which aligns it to the next 4-byte boundary
(doesn't truncate). Doing this results 1-3 bytes extra to be transferred.
Furthermore we handle remaining bytes using PIO which results one extra
byte to be transferred. In worst case the driver transfers 4 extra bytes.
While investigating this it turned out that the DMA hardware doesn't even
have such limitation so we can solve this by dropping the code that tries
to handle unaligned bytes.
Reported-by: Chiau Ee Chew <chiau.ee.chew@intel.com>
Reported-by: Hock Leong Kweh <hock.leong.kweh@intel.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Mark Brown <broonie@linaro.org>
None of these files are actually using any __init type directives
and hence don't need to include <linux/init.h>. Most are just a
left over from __devinit and __cpuinit removal, or simply due to
code getting copied from one driver to the next.
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Mark Brown <broonie@linaro.org>
pxa2xx_spi_map_dma_buffer() gets called in tasklet context so we can't
sleep when we allocate a new sg table. Use GFP_ATOMIC here instead.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Mark Brown <broonie@linaro.org>
Cc: stable@vger.kernel.org
Now that we have these nice DMA API helper functions we can take advantage
of those instead of open-coding the channel/request line extraction from
ACPI. Use the _compat version which still allows passing the
channel/request line from platform data.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
To be able to use DMA with this driver on non-PXA platforms we implement
support for the generic DMA engine API. This lets user to use different DMA
engines with little or no modification to the driver.
Request lines and channel numbers can be passed to the driver from the
platform specific data.
The DMA engine implementation will be selected by default even on PXA
platform. User can select the legacy DMA API by enabling Kconfig option
CONFIG_SPI_PXA2XX_PXADMA.
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Lu Cao <lucao@marvell.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Jarkko Nikula