With addition of flash_parameters structure, and supporting API call
to retrieve it, it is no longer needed to store write_block_size as
a part of flash_driver_api and it should be part of flash_parameters.
Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
With addition of flash_get_parameters API call, it is needed to provide
support for the API to flash drivers.
Signed-off-by: Dominik Ermel <dominik.ermel@nordicsemi.no>
The Nordic QSPI peripheral uses DMA transfers so data to write must be
located in SRAM. Add a Kconfig that enables copying data from NVMC to
a stack SRAM buffer so it can be written to flash.
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
mcumgr and possibly mcuboot write single byte values to update the
state of objects. Rather than fail to do the write of values too
short for this peripheral detect the situation and write from a stack
buffer that meets the length criteria.
Signed-off-by: Sigvart Hovland <sigvart.m@gmail.com>
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
Currently user is forced to configure an array of 4 IO pins. This makes
no sense when there are only 2 IO pins connected on board.
Configure 3rd and 4th pin in internal structure as
NRF_QSPI_PIN_NOT_CONNECTED if only 2 were specified in device-tree.
Signed-off-by: Marcin Niestroj <m.niestroj@grinn-global.com>
Erase can only succeed if the address is sector-aligned, and the span
to erase is an integer multiple of the sector size. Validate this
before starting the process of erasing things.
Also reduce the check that the affected region is within the device
span to its minimal equivalent conditions.
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
Having a completion wait function release a lock internally only when
the operation fails is confusing. Remove that feature, and make the
lock and unlock operations explicit and paired.
This makes it much more clear how to properly handle transactions that
require multiple calls to the HAL.
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
The stack buffer used to pad small reads must still be aligned to a
word address as the underlying driver uses DMA transfers with that
requirement. Don't assume the compiler will naturally align it.
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
The check for small transfers inadvertently allowed a transfer of zero
bytes, which should be an error (invalid parameter).
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
mcuboot and possibly other tools read single byte values to determine
the state of objects. Rather than fail to do the read of values too
short for this peripheral detect the situation and read into a stack
buffer that meets the length criteria, and on success copy the data
into the provided buffer.
Signed-off-by: Peter Bigot <peter.bigot@nordicsemi.no>
The code generally invoked qspi_wait_for_completion() within a locked
region without verifying that an operation was successfully initiated.
This caused a deadlock whenever the operation failed, e.g. because the
data buffer was not 4-byte aligned. Update that function to take the
result of the operation and either wait for completion or release the
lock, depending on its value.
Also uniformly use the correct type for Nordic HAL error values, and
refactor the erase module so that the correct values are displayed in
the diagnostic when something goes wrong.
Also check the alignment requirements for the flash address and
transfer size, which are highly constrained on this device. This
driver also requires 4-byte aligned data buffer; this is checked in
the Nordic HAL.
Signed-off-by: Peter A. Bigot <pab@pabigot.com>
Replace all occurences of BUILD_ASSERT_MSG() with BUILD_ASSERT()
as a result of merging BUILD_ASSERT() and BUILD_ASSERT_MSG().
Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
This reverts commit 8739517107.
Pull Request #23437 was merged by mistake with an invalid manifest.
Signed-off-by: Carles Cufi <carles.cufi@nordicsemi.no>
Replace all occurences of BUILD_ASSERT_MSG() with BUILD_ASSERT()
as a result of merging BUILD_ASSERT() and BUILD_ASSERT_MSG().
Signed-off-by: Oleg Zhurakivskyy <oleg.zhurakivskyy@intel.com>
Most JEDEC NOR flash devices uses not only typical SPI mode
(MISO,MOSI,SCK and CS), but also QSPI mode (IO0,IO1,IO2,IO3,SCK and CS).
QSPI mode uses more data lines and as a result provide higher
throughput. If this were not enough, Nordic chips provide
hardware acceleration for read/write/erase functions, what
gives significant performance boost.
It does a lot of things "behind the scene", i.e when user has written
some data to the flash and would like to read them back, it has to wait
until the flash is ready by reading WIP bit in Status Register.
This driver does it automatically.
Signed-off-by: Kamil Lazowski <Kamil.Lazowski@nordicsemi.no>