The Denali IP adopts the syndrome page layout (payload and ECC are
interleaved). The *_page_raw() and *_oob() callbacks are complicated
because they must hide the underlying layout used by the hardware,
and always return contiguous in-band and out-of-band data.

The Denali IP cannot reuse nand_{read,write}_page_raw_syndrome()
in nand_base.c because its hardware ECC engine skips some of first
bytes in OOB. That is why this driver implements specially-crafted
*_page_raw() and *_oob() hooks.

Currently, similar code is duplicated to reorganize the data layout.
For example, denali_read_page_raw() and denali_write_page_raw() look
almost the same. The complexity is partly due to the DMA transfer
used for better performance of *_page_raw() accessors.

On second thought, we do not need to care about their performance
because MTD_OPS_RAW is rarely used.

Let's focus on code cleanups rather than the performance. This commit
removes the internal buffer for DMA, and factors out as much code as
possible.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

With the recent refactoring, the NAND driver hooks now take a pointer
to nand_chip. Add to_denali() in order to convert (struct nand_chip *)
to (struct denali_nand_info *) directly. It is more useful than the
current mtd_to_denali().

I changed some helper functions to take (struct nand_chip *). This will
avoid pointer conversion back and forth, and ease further development.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

Wrong copy/paste from the previous block, the error message should
refer to #size-cells instead of #address-cells.

Fixes: f88fc122

 ("mtd: nand: Cleanup/rework the atmel_nand driver")
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Tudor Ambarus <tudor.ambarus@microchip.com>

Move the code that choses ECC into _attach_chip, which is executed only
after the chip->ecc.* properties were loaded from device-tree. This way
we know which ECC method was chosen by the device-tree and can set
methods appropriately.

The chip->ecc.*page methods should be set to fsl_elbc_*page only in HW
ECC mode.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

This is needed for SW ECC.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

Move the fsl_elbc_attach_chip function after the definitions of
fsl_elbc_read_page and friends in preparation for the next patch.

Signed-off-by: Marek Behún <marek.behun@nic.cz>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

Sphinx doesn't handle expressions in identifier references.

This fixes the following warnings:

./include/linux/mtd/rawnand.h:1184: WARNING: Inline strong start-string without end-string.
./include/linux/mtd/rawnand.h:1186: WARNING: Inline strong start-string without end-string.

Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

When enabling CONFIG_MTD_NAND_ECC_SW_BCH as a module, the
MTD_NAND_ECC_SW_BCH depends on MTD_NAND, but the module controlled by
MTD_NAND links against the module controlled by MTD_NAND_ECC_SW_BCH.
This leads to the following link failure.

aarch64-linux-gnu-ld: drivers/mtd/nand/raw/nand_base.o: in function `nand_cleanup':
../drivers/mtd/nand/raw/nand_base.c:5886: undefined reference to `nand_bch_free'
aarch64-linux-gnu-ld: ../drivers/mtd/nand/raw/nand_base.c:5886:(.text+0x9928): relocation truncated to fit: R_AARCH64_CALL26 against undefined symbol `nand_bch_free'
aarch64-linux-gnu-ld: drivers/mtd/nand/raw/nand_base.o: in function `nand_set_ecc_soft_ops':
../drivers/mtd/nand/raw/nand_base.c:5093: undefined reference to `nand_bch_calculate_ecc'
aarch64-linux-gnu-ld: ../drivers/mtd/nand/raw/nand_base.c:5093:(.text+0xe914): relocation truncated to fit: R_AARCH64_ADR_PREL_PG_HI21 against undefined symbol `nand_bch_calculate_ecc'
aarch64-linux-gnu-ld: ../drivers/mtd/nand/raw/nand_base.c:5093: undefined reference to `nand_bch_calculate_ecc'
aarch64-linux-gnu-ld: ../drivers/mtd/nand/raw/nand_base.c:5094: undefined reference to `nand_bch_correct_data'
aarch64-linux-gnu-ld: ../drivers/mtd/nand/raw/nand_base.c:5094:(.text+0xe934): relocation truncated to fit: R_AARCH64_ADR_PREL_PG_HI21 against undefined symbol `nand_bch_correct_data'
aarch64-linux-gnu-ld: ../drivers/mtd/nand/raw/nand_base.c:5094: undefined reference to `nand_bch_correct_data'
aarch64-linux-gnu-ld: ../drivers/mtd/nand/raw/nand_base.c:5148: undefined reference to `nand_bch_init'
aarch64-linux-gnu-ld: ../drivers/mtd/nand/raw/nand_base.c:5148:(.text+0xebbc): relocation truncated to fit: R_AARCH64_CALL26 against undefined symbol `nand_bch_init'

Rework CONFIG_MTD_NAND_ECC_SW_BCH from tristate to bool,
and then link the nand_bch.o file into nand.ko if its enabled.

Fixes: 51ef1d0b2095 ("mtd: nand: Clarify Kconfig entry for software BCH ECC algorithm")
Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

Prepare changes that will lay in this file to better express what is
NAND related and what is not in menuconfig.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

MTD_NAND is large and encloses much more than what the symbol is
actually used for: raw NAND. Clarify the symbol by naming it
MTD_RAW_NAND instead.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

This list is a mess, while some items should probably not be in the
raw/ sub-directory, others are definitely at the right place but not
with the right description. Write uniform titles and group IPs by
vendor.

NAND controllers will appear under the list named "Raw/parallel NAND
flash controllers" while the other drivers will appear under
"Misc". Software ECC engines will later be moved out of the raw/
directory.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The software Hamming ECC correction implementation is referred as
MTD_NAND_ECC which is too generic. Rename it
MTD_NAND_ECC_SW_HAMMING. Also rename MTD_NAND_ECC_SMC which is an
SMC quirk in the Hamming implementation as
MTD_NAND_ECC_SW_HAMMING_SMC.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

There is no point in having two distinct entries, merge them and
rename the symbol for more clarity: MTD_NAND_ECC_SW_BCH

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The symbol that is being used in the #if/#endif block is not the one
which is mentioned at the bottom.

Fixes: 93af53b8

 ("nand: omap2: Remove horrible ifdefs to fix module probe")
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

nand_device embeds a nand_ecc_req object which contains the minimum
strength and step-size required by the NAND device.

Drop the chip->ecc_{strength,step}_ds fields and use
chip->base.eccreq.{strength,step_size} instead.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

The same information is provided by nanddev_ntargets().

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

The target size can now be returned by nanddev_get_targetsize(). Get
rid of the chip->chipsize field and use this helper instead.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

Now that we inherit from nand_device, we can use
nand_device->memorg.bits_per_cell instead of having our own field at
the nand_chip level.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

nanddev_mtd_max_bad_blocks() is implemented by the generic NAND layer
and is already doing what we need. Reuse this function instead of
having our own implementation.

While at it, get rid of the ->max_bb_per_die and ->blocks_per_die
fields which are now unused.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

Looking at the field names it's hard to tell what ->data_buf, ->pagebuf
and ->pagebuf_bitflips are for. Clarify that by moving those fields
in a sub-struct named pagecache.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

We plan to move cache related fields to a pagecache struct in nand_chip
but some drivers access ->pagebuf directly to invalidate the cache
before they start using ->data_buf.

Let's provide an helper that returns a pointer to ->data_buf after
invalidating the cache.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

In order to use some of the nanddev_xxx() helpers, we need to
initialize the nand_device object embedded in nand_chip using
nanddev_init(). This requires implementing nand_ops.

We also drop useless mtd->xxx initialization when they're already taken
case of by nanddev_init().

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

If we want to use the generic NAND layer, we need to have the memorg
struct appropriately filled. Patch the detection code to fill this
struct.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The generic NAND layer provides abstraction of NAND devices no matter
the bus that is used to communicate with the chip. Basing the raw NAND
core on this generic layer should avoid duplication of common
operations, like iterating over all pages/blocks for MTD IO/erase
operations.

In order to re-use this layer, we must first inherit from nand_device
and then initialize the nand_device struct appropriately. This patch
is taking care of the former.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

Use the nand_to_mtd() helper to access chip->mtd as done everywhere
else.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

We just have to use nanddev_mtd_max_bad_blocks().

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

Will be used by the raw NAND framework.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

Some drivers in the raw NAND framework seems to need this helper, so
let's just add it instead of open-coding the logic.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

NAND datasheets usually give the maximum number of bad blocks per LUN
and this number can be used to help upper layers decide how much blocks
they should reserve for bad block handling.

Add a max_bad_eraseblocks_per_lun to the nand_memory_organization
struct and update the NAND_MEMORG() macro (and its users) accordingly.

We also provide a default mtd->_max_bad_blocks() implementation.

Signed-off-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Frieder Schrempf <frieder.schrempf@kontron.de>

Specify the oob layout operation to avoid no oob scheme defined for
some nand flash.

Fixes: 8fae856c

 ("mtd: rawnand: meson: add support for Amlogic NAND flash controller")
Signed-off-by: Liang Yang <liang.yang@amlogic.com>
Tested-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

of_match_device can return NULL if there is no matching device. Avoid
a potential NULL pointer dereference by checking for the return value
and passing the error upstream.

Signed-off-by: Aditya Pakki <pakki001@umn.edu>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The generic layout for BBT markers will most likely overlap with our
ECC bytes in the OOB, so move the BBT markers outside the OOB area.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The Ben Nanonote from Qi Hardware expects a specific OOB layout on its
NAND.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The boot ROM of the JZ4725B SoC expects a specific OOB layout on the
NAND, so we use it unconditionally in the ingenic-nand driver.

Also add the jz4725b-bch driver to support the JZ4725B-specific BCH
hardware.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

Add support for probing the ingenic-nand driver on the JZ4740 SoC from
Ingenic, and the jz4740-ecc driver to support the JZ4740-specific
ECC hardware.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

Use the 'ecc-engine' standard property instead of the custom
'ingenic,bch-controller' custom property, which is now deprecated.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The ingenic-nand driver uses an API provided by the jz4780-bch driver.
This makes it difficult to support other SoCs in the jz4780-bch driver.
To work around this, we separate the API functions from the SoC-specific
code, so that these API functions are SoC-agnostic.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The jz4780_bch_init name was confusing, as it suggested that its content
should be executed once at init time, whereas what the function really
does is reset the hardware for a new ECC operation.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

The jz4780_nand driver will be modified to handle all the Ingenic
JZ47xx SoCs that the upstream Linux kernel supports (JZ4740, JZ4725B,
JZ4770, JZ4780), so it makes sense to rename it.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>

Use SPDX license notifiers instead of GPLv2 license text in the headers.

Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Reviewed-by: Boris Brezillon <bbrezillon@kernel.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>