The dma driver was determining src_inc and dst_inc from the
config of the first block buffer and ignoring the config
flags for any additional buffers in the chain, which could
lead to incorrect transfers (e.g. in a multiple rx buffer
case, if the first buffer was to receive to NULL,
but the subsequent buffers were not NULL, the bug
would manifest as all transfers being made with
dst_inc of 0). Change the driver to setup
each dma descriptor according to the addr_adj flag
of each block_buffer.
Add check that peripheral transfers have addr_adj set to
NO_CHANGE instead of assuming it, to help catch errors.
Also now check for invalid addr_adj request of
decrement, which this controller doesn't support.
Signed-off-by: Mike J. Chen <mjchen@google.com>
The spi_mcux_flexcomm driver uses a special last DMA blk_cfg
to trigger a release of the SPI chip select. This transfer
is always a 4-byte transfer, regardless of the width specified
during dma_configure().
The way the spi_mcux_flexcomm driver communicated this special
transfer was kind of a hack, where the dma_mcux_lpc driver would
assume that when a blk_cfg with source_addr_adj and dest_addr_adj
both set to NO_CHANGE was for this SPI_TX special case.
However, this is an unsafe hack since it is perfectly valid
to have dma use cases for both src/dest_addr_adj to be NO_CHANGE
that is not for SPI_TX. One example is when transmitting a
fixed/repeating value to a periperhal address (e.g. send 100
bytes of the same value from a single memory address over SPI).
This CL introduces a dma_mcux_lpc specific dma channel_direction
which the two drivers now use to cleary request this special
transfer case.
Signed-off-by: Mike J. Chen <mjchen@google.com>
In spi_mcux_transfer_next_packet(), if the next
transfer start fails, add calls to
spi_context_cs_control() to release cs and
spi_context_complete() with error code -EIO.
Signed-off-by: Mike J. Chen <mjchen@google.com>
The current driver implementation would block even when the async
API was invoked, so it wasn't really async.
This CL also fully chains the DMA transfer using multiple dma blocks
and makes the number of dma blocks available a config value. The
increase in number of dma blocks is needed so that a spi_buf_set
that has many entries can be converted into chained dma transfers
with the last transfer in a separate block that will set the EOT flag.
Also make some improvements:
1) When doing single cnt transfer, don't use the SDK driver but
use a new fucntion spi_mcux_transfer_single_word().
It's much more efficient and does not use an ISR
like the SDK function does just to send one word.
2) Fix calls to spi_context_update_tx/rx() so that the
correct word size is passed in, instead of previously
being hardcoded to 1. This only matters when word size
is two. Evaluate the word_size_bytes and word_size_bits
once and store the values in data instead of computing
it multiple times in various parts of the driver
3) When CONFIG_SPI_MCUX_FLEXCOMM_DMA is defined, we
do not use the IRQ handler, so add #ifdefs to compile
that code out. This reduces code size.
Signed-off-by: Mike J. Chen <mjchen@google.com>
This reverts commit a3530d6a43.
This change incorrect if chip-select is via a GPIO pin,
released by spi_context_cs_control(), and not the controller
default chip-select pin that can be released using a final
FIFOWR.
When GPIO is used for chip-select control, the chip-select could
be released too soon because the transfer callback is invoked
when the last byte is put into the TX FIFO but that byte might
not yet have clocked out yet on the SPI pins. The rx part of
the transfer is really what completes the transfer so ignoring
it is incorrect.
We could try to special case and use ignore only when
spi_context_cs_control() does nothing, but since it is a very
small optimization, it doesn't seem worth the extra
complexity.
Signed-off-by: Mike J. Chen <mjchen@google.com>
If CONFIG_SPI_STATS is enabled, the device state for all SPI controller
drivers must contain the SPI stats. This space is allocated by calling
Z_SPI_INIT_FN as part of the device definition; this is done automatically
when using SPI_DEVICE_DT_DEFINE instead of DEVICE_DT_DEFINE. If space for
statistics is not properly allocated but CONFIG_SPI_STATS is enabled, an
unexpected write to memory outside of the stats region may occur on a SPI
transfer. This commit uses SPI_DEVICE_DT_DEFINE or
SPI_DEVICE_DT_INST_DEFINE for all in-tree SPI controller drivers.
Signed-off-by: Dane Wagner <dane.wagner@gmail.com>
when no rx data need to be read set the rxignore bits.
with this change the dma setup is faster and no unnecessary
dummy writes are done to memory.
Signed-off-by: Johan Carlsson <johan.carlsson@teenage.engineering>
when sending multiple bytes only the DMA_STATUS_COMPLETE
status is interesting. otherwise the semaphore will be signaled twice.
Signed-off-by: Johan Carlsson <johan.carlsson@teenage.engineering>
Remove address-of operator ('&') when assigning `init_fn`
function pointer in `DEVICE_DT_INST_DEFINE` macro.
This change aims to maintain consistency among the drivers in
`drivers/spi`, ensuring that all function pointer assignments
follow the same pattern.
Signed-off-by: Pisit Sawangvonganan <pisit@ndrsolution.com>
fixes an incorrect dma configuration. When lpc dma driver was extended
with gather/scatter support the spi dma driver stopped working.
Signed-off-by: Johan Carlsson <johan.carlsson@teenage.engineering>
The MCUX LPC driver now recognizes the complete_callback
flag. We need to set this flag so we receive a callback
after every block.
Signed-off-by: Mahesh Mahadevan <mahesh.mahadevan@nxp.com>
The MCUX DMA controller only supports a single data_size
for a DMA transfer, not separate ones for source and
dest. An older version of the DMA driver used
dest_data_size as the DMA transfer size, but the
current one uses MIN(dest/source) as the trasnfer
size, which breaks case when SPI wants to do 2-byte
transfers.
Signed-off-by: Mike J. Chen <mjchen@google.com>
Fix for bug:
https://github.com/zephyrproject-rtos/zephyr/issues/59575
The dma version of the version of the driver can
invoke multiple intermediate dma transfers, like
when the spi_buf_set count is greater than one.
However, there is a bug where chip select is not kept
asserted for all intermediate dma transfers required
to process the entire spi_buf_set.
Signed-off-by: Mike J. Chen <mjchen@google.com>
Make use of positive status values in the DMA callback to pass
info to the DMA client after a successful DMA operation.
A completed DMA transfer uses the status 0 while a reached
water mark uses the status 1.
Signed-off-by: Cyril Fougeray <cyril.fougeray@worldcoin.org>
The MCUX platform always uses pinctrl, there's no need to keep extra
macrology around pinctrl. Also updated driver's Kconfig options to
`select PINCTRL` (note that some already did).
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
When using a dummy buffer on the RX side, do not
increment the destination memory buffer address.
This issue was uncovered when running the SPI
loopback test.
Signed-off-by: Mahesh Mahadevan <mahesh.mahadevan@nxp.com>
Change automated searching for files using "IRQ_CONNECT()" API not
including <zephyr/irq.h>.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Adds a new spi_transcieve_cb API which enables asynchronous
SPI transactions with callback notification.
The exist spi_transcieve_async API remains and uses the new
spi_transcieve_cb API to provide a k_poll_signal notifier.
The driver API changes to provide a callback and userdata
parameter to async transcieve. All drivers in the tree
have been updated to the change.
Signed-off-by: Tom Burdick <thomas.burdick@intel.com>
This is a bug fix. A pointer to the spi configuration is not saved when
the spi driver is configured for slave operation and it can lead to
runtime errors.
Signed-off-by: Bryce Wilkins <bryce.wilkins@gmail.com>
Adds optional device tree property to specify a default character
to clock out when the TX buffer pointer is NULL. If the property is
not set the existing behavior (default char of 0x00) is used.
I verified the expected behavior using an i.MX RT685 board and
logic analyzer that the def-char character is transmitted when
TX buffer pointer is NULL.
Signed-off-by: Bryce Wilkins <bryce.wilkins@gmail.com>
In order to bring consistency in-tree, migrate all drivers to the new
prefix <zephyr/...>. Note that the conversion has been scripted, refer
to #45388 for more details.
Signed-off-by: Gerard Marull-Paretas <gerard.marull@nordicsemi.no>
Adds optional device tree properties to insert delays between spi chip
select assert/deassert and the clock edges, and also between spi
frames and transfers to the mcux flexcomm spi driver. If the properties
are not set, no additional delay is inserted.
Verified expected behavior on mimxrt685_evk and check with a scope
that the pre- and post-delay could be changed from the device tree
properties.
Signed-off-by: Bryce Wilkins <bryce.wilkins@gmail.com>
In case when we have multiple devices connected to the
one SPI interface the initial state of CS gpios after
MCU reset is floating and it might be low that prevents us from
communicating between particular devices. Fix that by
initializing all provided cs gpios and setting them as inactive.
Signed-off-by: Bartosz Bilas <bartosz.bilas@hotmail.com>
A couple of SPI drivers use CONFIG_KERNEL_INIT_PRIORITY_DEVICE
as init priority for driver initialization. Let's change
it to the dedicated CONFIG_SPI_INIT_PRIORITY to make it
compatible with other ones.
Signed-off-by: Bartosz Bilas <bartosz.bilas@hotmail.com>
Change to spi_context_lock missed one spot in the flexcomm driver and
this causes a build issue. Pass spi_cfg to spi_context_lock to fix
the issue.
Signed-off-by: Kumar Gala <kumar.gala@linaro.org>
Keep locking for SPI_LOCK_ON from the first call of transceive until
spi_release release the lock. Use owner parameter to in the spi_context
to store the owner of the lock.
The locking is in line with the SPI_HOLD_ON_CS
Signed-off-by: Stefan Bigler <stefan@bigler.io>
These are all the case that coccinelle cannot find as they are inside
macro declarations.
Fixed via:
git grep -rlz -E "\(struct device \*" |
xargs -0 sed -i 's/(struct device/(const struct device/g'
Fixes#27399
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Some needed to wrap the device pointer into device's data, where others
needed only device's data to be passed to HAL callback function.
Fixes#27399
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
Now that device_api attribute is unmodified at runtime, as well as all
the other attributes, it is possible to switch all device driver
instance to be constant.
A coccinelle rule is used for this:
@r_const_dev_1
disable optional_qualifier
@
@@
-struct device *
+const struct device *
@r_const_dev_2
disable optional_qualifier
@
@@
-struct device * const
+const struct device *
Fixes#27399
Signed-off-by: Tomasz Bursztyka <tomasz.bursztyka@linux.intel.com>
