platform/kernel/linux-rpi.git
6 years agoMerge tag 'nand/for-4.19' of git://git.infradead.org/linux-mtd into mtd/next
Boris Brezillon [Sat, 11 Aug 2018 10:15:19 +0000 (12:15 +0200)]
Merge tag 'nand/for-4.19' of git://git.infradead.org/linux-mtd into mtd/next

Pull NAND updates from Miquel Raynal:

"
 NAND core changes:
 - Add the SPI-NAND framework.
 - Create a helper to find the best ECC configuration.
 - Create NAND controller operations.
 - Allocate dynamically ONFI parameters structure.
 - Add defines for ONFI version bits.
 - Add manufacturer fixup for ONFI parameter page.
 - Add an option to specify NAND chip as a boot device.
 - Add Reed-Solomon error correction algorithm.
 - Better name for the controller structure.
 - Remove unused caller_is_module() definition.
 - Make subop helpers return unsigned values.
 - Expose _notsupp() helpers for raw page accessors.
 - Add default values for dynamic timings.
 - Kill the chip->scan_bbt() hook.
 - Rename nand_default_bbt() into nand_create_bbt().
 - Start to clean the nand_chip structure.
 - Remove stale prototype from rawnand.h.

 Raw NAND controllers drivers changes:
 - Qcom: structuring cleanup.
 - Denali: use core helper to find the best ECC configuration.
 - Possible build of almost all drivers by adding a dependency on
   COMPILE_TEST for almost all of them in Kconfig, implies various
   fixes, Kconfig cleanup, GPIO headers inclusion cleanup, and even
   changes in sparc64 and ia64 architectures.
 - Clean the ->probe() functions error path of a lot of drivers.
 - Migrate all drivers to use nand_scan() instead of
   nand_scan_ident()/nand_scan_tail() pair.
 - Use mtd_device_register() where applicable to simplify the code.
 - Marvell:
   * Handle on-die ECC.
   * Better clocks handling.
   * Remove bogus comment.
   * Add suspend and resume support.
 - Tegra: add NAND controller driver.
 - Atmel:
   * Add module param to avoid using dma.
   * Drop Wenyou Yang from MAINTAINERS.
 - Denali: optimize timings handling.
 - FSMC: Stop using chip->read_buf().
 - FSL:
   * Switch to SPDX license tag identifiers.
   * Fix qualifiers in MXC init functions.

 Raw NAND chip drivers changes:
 - Micron:
   * Add fixup for ONFI revision.
   * Update ecc_stats.corrected.
   * Make ECC activation stateful.
   * Avoid enabling/disabling ECC when it can't be disabled.
   * Get the actual number of bitflips.
   * Allow forced on-die ECC.
   * Support 8/512 on-die ECC.
   * Fix on-die ECC detection logic.
 - Hynix:
   * Fix decoding the OOB size on H27UCG8T2BTR.
   * Use ->exec_op() in hynix_nand_reg_write_op().
"

6 years agoMAINTAINERS: drop Wenyou Yang from Atmel NAND driver support
Miquel Raynal [Wed, 25 Jul 2018 14:56:28 +0000 (16:56 +0200)]
MAINTAINERS: drop Wenyou Yang from Atmel NAND driver support

Mails to wenyou.yang@microchip.com are not deliverable.
Drop him as Microchip/Atmel NAND controller driver maintainer.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Acked-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Alexandre Belloni <alexandre.belloni@bootlin.com>
6 years agomtd: rawnand: allocate dynamically ONFI parameters during detection
Miquel Raynal [Wed, 25 Jul 2018 13:31:52 +0000 (15:31 +0200)]
mtd: rawnand: allocate dynamically ONFI parameters during detection

Now that it is possible to do dynamic allocations during the
identification phase, convert the onfi_params structure (which is only
needed with ONFI compliant chips) into a pointer that will be allocated
only if needed.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: allocate model parameter dynamically
Miquel Raynal [Wed, 25 Jul 2018 13:31:51 +0000 (15:31 +0200)]
mtd: rawnand: allocate model parameter dynamically

Thanks to the migration of all drivers to use nand_scan() and the
related nand_controller_ops, we can now allocate data during the
detection phase. Let's do it first for the NAND model parameter which
is allocated in nand_detect().

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: do not export nand_scan_[ident|tail]() anymore
Miquel Raynal [Wed, 25 Jul 2018 13:31:50 +0000 (15:31 +0200)]
mtd: rawnand: do not export nand_scan_[ident|tail]() anymore

Both nand_scan_ident() and nand_scan_tail() helpers used to be called
directly from controller drivers that needed to tweak some ECC-related
parameters before nand_scan_tail(). This separation prevented dynamic
allocations during the phase of NAND identification, which was
inconvenient.

All controller drivers have been moved to use nand_scan(), in
conjunction with the chip->ecc.[attach|detach]_chip() hooks that
actually do the required tweaking sequence between both ident/tail
calls, allowing programmers to use dynamic allocation as they need all
across the scanning sequence.

Declare nand_scan_[ident|tail]() statically now.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: txx9ndfmc: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:49 +0000 (15:31 +0200)]
mtd: rawnand: txx9ndfmc: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: txx9ndfmc: clarify ECC parameters assignation
Miquel Raynal [Wed, 25 Jul 2018 13:31:48 +0000 (15:31 +0200)]
mtd: rawnand: txx9ndfmc: clarify ECC parameters assignation

A comment in the probe declares that values are assigned to ecc.size
and ecc.bytes, but these values will be overwritten. This is not
entirely right as they are overwritten only if
mtd->writesize >= 512. Let's clarify this by moving these assignations
to txx9ndfmc_nand_scan().

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: tegra: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:47 +0000 (15:31 +0200)]
mtd: rawnand: tegra: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: jz4740: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:46 +0000 (15:31 +0200)]
mtd: rawnand: jz4740: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: jz4740: group nand_scan_{ident, tail} calls
Miquel Raynal [Wed, 25 Jul 2018 13:31:45 +0000 (15:31 +0200)]
mtd: rawnand: jz4740: group nand_scan_{ident, tail} calls

Prepare the migration to nand_scan() by moving both calls to
nand_scan_ident() and nand_scan_tail() in a single spot.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: jz4740: fix probe function error path
Miquel Raynal [Wed, 25 Jul 2018 13:31:44 +0000 (15:31 +0200)]
mtd: rawnand: jz4740: fix probe function error path

An error after nand_scan_tail() should trigger a nand_cleanup(), not a
nand_release() as mtd_device_register() (or one of its variants) has not
been called and there is no need to deregister any MTD device yet.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: docg4: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:43 +0000 (15:31 +0200)]
mtd: rawnand: docg4: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: do not execute nand_scan_ident() if maxchips is zero
Miquel Raynal [Wed, 25 Jul 2018 13:31:42 +0000 (15:31 +0200)]
mtd: rawnand: do not execute nand_scan_ident() if maxchips is zero

Some driver (eg. docg4) will need to handle themselves the
identification phase. As part of the migration to use nand_scan()
everywhere (which will unconditionnaly call nand_scan_ident()), we add
a condition at the start of nand_scan_with_ids() to jump over
nand_scan_ident() if the maxchips parameters is zero, meaning that the
driver does not want the core to handle this phase.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: atmel: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:41 +0000 (15:31 +0200)]
mtd: rawnand: atmel: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: atmel: clarify NAND addition/removal paths
Miquel Raynal [Wed, 25 Jul 2018 13:31:40 +0000 (15:31 +0200)]
mtd: rawnand: atmel: clarify NAND addition/removal paths

No need for an atmel_nand_register() function, let's move the code in
it directly where the function was called: in
atmel_nand_controller_add_nand(). To make things consistent, also
rename atmel_nand_unregister() into
atmel_nand_controller_remove_nand().

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: omap2: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:39 +0000 (15:31 +0200)]
mtd: rawnand: omap2: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: lpc32xx_mlc: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:38 +0000 (15:31 +0200)]
mtd: rawnand: lpc32xx_mlc: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: cafe: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:37 +0000 (15:31 +0200)]
mtd: rawnand: cafe: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: brcmnand: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 13:31:36 +0000 (15:31 +0200)]
mtd: rawnand: brcmnand: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agodrivers/memory/Kconfig: Add CONFIG_OF dependency
Anders Roxell [Wed, 25 Jul 2018 09:19:28 +0000 (11:19 +0200)]
drivers/memory/Kconfig: Add CONFIG_OF dependency

JZ4780_NEMC doesn't depend on OF, and if OF isn't enabled we get this
error:
drivers/memory/jz4780-nemc.c: In function ‘jz4780_nemc_num_banks’:
drivers/memory/jz4780-nemc.c:72:10: error: implicit declaration of
 function ‘of_read_number’; did you mean ‘down_read_nested’?
 [-Werror=implicit-function-declaration]
   bank = of_read_number(prop, 1);
          ^~~~~~~~~~~~~~
          down_read_nested

Make JZ4780_NEMC depend on OF.

Fixes: ab99e11062c1 ("memory: jz4780-nemc: Allow selection of this driver when COMPILE_TEST=y")
Reported-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Anders Roxell <anders.roxell@linaro.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agosparc64: add reads{b,w,l}/writes{b,w,l}
Arnd Bergmann [Wed, 25 Jul 2018 09:18:34 +0000 (11:18 +0200)]
sparc64: add reads{b,w,l}/writes{b,w,l}

Some drivers need these for compile-testing. On most architectures
they come from asm-generic/io.h, but not on sparc64, which has its
own definitions.

Since we already have ioread*_rep()/iowrite*_rep() that have the
same behavior on sparc64 (i.e. all PCI I/O space is memory mapped),
we can rename the existing helpers and add macros to define them
to the same implementation.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: David S. Miller <davem@davemloft.net>
Tested-by: Boris Brezillon <boris.brezillon@bootlin>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agoia64: use asm-generic/io.h
Arnd Bergmann [Wed, 25 Jul 2018 09:17:20 +0000 (11:17 +0200)]
ia64: use asm-generic/io.h

asm-generic/io.h provides a generic implementation of all I/O accessors,
which the architectures can override.

Since ia64 does not provide readsl/writesl etc, any driver using those
fails to build, and including asm-generic/io.h will provide the
missing interfaces, as well as any other future interfaces that get
added there. We need to #define a couple of symbols to themselves
in the ia64 to ensure that we use the ia64 specific version of those
rather than the generic one.

There should be no other effect than adding {read,write}s{b,w,l}()
as well as {in,out}s{b,w,l}_p(), which were also not provided
by ia64 but are provided by the generic header for historic reasons.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Tested-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: qcom: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:22 +0000 (17:15 +0200)]
mtd: rawnand: qcom: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: sm_common: convert driver to nand_scan_with_ids()
Miquel Raynal [Thu, 26 Jul 2018 23:18:05 +0000 (01:18 +0200)]
mtd: rawnand: sm_common: convert driver to nand_scan_with_ids()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan_with_ids() (alternative to nand_scan() for passing
a flash IDs table) instead of the nand_scan_ident() + nand_scan_tail()
pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: sm_common: fix the probe function error path
Miquel Raynal [Sun, 25 Feb 2018 22:09:14 +0000 (23:09 +0100)]
mtd: rawnand: sm_common: fix the probe function error path

nand_cleanup() should be called upon error after a successful
nand_scan_tail().

Rework the error path to follow this rule .

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: vf610: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:17 +0000 (17:15 +0200)]
mtd: rawnand: vf610: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Reviewed-by: Stefan Agner <stefan@agner.ch>
6 years agomtd: rawnand: txx9ndfmc: rename nand controller internal structure
Miquel Raynal [Fri, 20 Jul 2018 15:15:15 +0000 (17:15 +0200)]
mtd: rawnand: txx9ndfmc: rename nand controller internal structure

As already done in the core, calling a struct nand_controller
'hw_control' is misleading. Use the same name as in nand_base.c:
'controller'.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: tango: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:14 +0000 (17:15 +0200)]
mtd: rawnand: tango: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: sunxi: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:13 +0000 (17:15 +0200)]
mtd: rawnand: sunxi: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: sh_flctl: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:11 +0000 (17:15 +0200)]
mtd: rawnand: sh_flctl: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: s3c2410: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:10 +0000 (17:15 +0200)]
mtd: rawnand: s3c2410: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: nandsim: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:08 +0000 (17:15 +0200)]
mtd: rawnand: nandsim: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: mxc: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:07 +0000 (17:15 +0200)]
mtd: rawnand: mxc: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: mtk: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:06 +0000 (17:15 +0200)]
mtd: rawnand: mtk: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Xiaolei Li <xiaolei.li@mediatek.com>
6 years agomtd: rawnand: marvell: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:05 +0000 (17:15 +0200)]
mtd: rawnand: marvell: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: lpc32xx_slc: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:04 +0000 (17:15 +0200)]
mtd: rawnand: lpc32xx_slc: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: jz4780: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:02 +0000 (17:15 +0200)]
mtd: rawnand: jz4780: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Acked-by: Harvey Hunt <harveyhuntnexus@gmail.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: hisi504: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:01 +0000 (17:15 +0200)]
mtd: rawnand: hisi504: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: gpmi: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:15:00 +0000 (17:15 +0200)]
mtd: rawnand: gpmi: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: fsmc: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:14:59 +0000 (17:14 +0200)]
mtd: rawnand: fsmc: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: fsl_ifc: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:14:58 +0000 (17:14 +0200)]
mtd: rawnand: fsl_ifc: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: fsl_elbc: convert driver to nand_scan()
Miquel Raynal [Wed, 25 Jul 2018 08:37:40 +0000 (10:37 +0200)]
mtd: rawnand: fsl_elbc: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Notes:
    "pw 947037"

6 years agomtd: rawnand: fsl_elbc: return meaningful values
Miquel Raynal [Wed, 25 Jul 2018 08:35:57 +0000 (10:35 +0200)]
mtd: rawnand: fsl_elbc: return meaningful values

Return -ENOTSUPP instead of -1 from ->chip_init_tail() before migrating
this driver to use nand_scan() and transform this function to be a
callback run by the core.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: denali: convert to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:14:56 +0000 (17:14 +0200)]
mtd: rawnand: denali: convert to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Masahiro Yamada <yamada.masahiro@socionext.com>
6 years agomtd: rawnand: davinci: convert driver to nand_scan()
Miquel Raynal [Fri, 20 Jul 2018 15:14:55 +0000 (17:14 +0200)]
mtd: rawnand: davinci: convert driver to nand_scan()

Two helpers have been added to the core to do all kind of controller
side configuration/initialization between the detection phase and the
final NAND scan. Implement these hooks so that we can convert the driver
to just use nand_scan() instead of the nand_scan_ident() +
nand_scan_tail() pair.

Also change the unused "struct device *dev" parameter of the driver
structure into a platform device to reuse it in the ->attach_chip()
hook.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: au1550nd: Remove unneeded gpio.h inclusion
Boris Brezillon [Fri, 20 Jul 2018 09:57:40 +0000 (11:57 +0200)]
mtd: rawnand: au1550nd: Remove unneeded gpio.h inclusion

We don't use the GPIO API in this driver, let's just remove the
<linux/gpio.h> inclusion.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: atmel: Stop including gpio.h
Boris Brezillon [Fri, 20 Jul 2018 09:57:39 +0000 (11:57 +0200)]
mtd: rawnand: atmel: Stop including gpio.h

gpio/consumer.h defines everything we need, and it's clearly stated in
gpio.h that GPIO consumers should directly stop including gpio.h if they
can.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: sunxi: Remove gpio.h and of_gpio.h inclusions
Boris Brezillon [Fri, 20 Jul 2018 09:57:38 +0000 (11:57 +0200)]
mtd: rawnand: sunxi: Remove gpio.h and of_gpio.h inclusions

Commit ddd5ed3a90e7 ("mtd: rawnand: sunxi: Remove support for
GPIO-based Ready/Busy polling") removed GPIO-based RB polling. We no
longer need to include gpio headers.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: jz4740: Include gpio/consumer.h instead of gpio.h
Boris Brezillon [Thu, 19 Jul 2018 22:38:13 +0000 (00:38 +0200)]
mtd: rawnand: jz4740: Include gpio/consumer.h instead of gpio.h

GPIO consumers should no longer include <linux/gpio.h>, and instead
include <linux/gpio/consumer.h>.

Also, explicitly include <include/io.h> since it seems to be missing
after switching to <linux/gpio/consumer.h>.

This fixes a build error when selecting the driver without selecting
GPIOLIB, which can happen when COMPILE_TEST=y.

Fixes: 6968e07e8169 ("mtd: rawnand: jz4740: Allow selection of this driver when COMPILE_TEST=y")
Reported-by: Miquel Raynal <miquel.raynal@bootlin.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: add hooks that may be called during nand_scan()
Miquel Raynal [Wed, 18 Jul 2018 23:05:46 +0000 (01:05 +0200)]
mtd: rawnand: add hooks that may be called during nand_scan()

In order to remove the limitation that forbids dynamic allocation in
nand_scan_ident(), we must create a path that will be the same for all
controller drivers. The idea is to use nand_scan() instead of the widely
used nand_scan_ident()/nand_scan_tail() couple. In order to achieve
this, controller drivers will need to adjust some parameters between
these two functions depending on the NAND chip wired on them.

This takes the form of two new hooks (->{attach,detach}_chip()) that are
placed in a new nand_controller_ops structure, which is then attached
to the nand_controller object at driver initialization time.
->attach_chip() is called between nand_scan_ident() and
nand_scan_tail(), and ->detach_chip() is called in the error path of
nand_scan() and in nand_cleanup().

Note that some NAND controller drivers don't have a dedicated
nand_controller object and instead rely on the default/dummy one
embedded in nand_chip. If you're in this case and still want to
initialize the controller ops, you'll have to manipulate
chip->dummy_controller directly.

Last but not least, it's worth mentioning that we plan to move some of
the controller related hooks placed in nand_chip into
nand_controller_ops to make the separation between NAND chip and NAND
controller methods clearer.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Acked-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: better name for the controller structure
Miquel Raynal [Tue, 17 Jul 2018 07:08:02 +0000 (09:08 +0200)]
mtd: rawnand: better name for the controller structure

In the raw NAND core, a NAND chip is described by a nand_chip structure,
while a NAND controller is described with a nand_hw_control structure
which is not very meaningful.

Rename this structure nand_controller.

As the structure gets renamed, it is logical to also rename the core
function initializing it from nand_hw_control_init() to
nand_controller_init().

Lastly, the 'hwcontrol' entry of the nand_chip structure is not
meaningful neither while it has the role of fallback when no controller
structure is provided by the driver (the controller driver is dumb and
can only control a single chip). Thus, it is renamed dummy_controller.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Acked-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: Remove unused caller_is_module() definition
Boris Brezillon [Thu, 19 Jul 2018 20:47:38 +0000 (22:47 +0200)]
mtd: rawnand: Remove unused caller_is_module() definition

Commit 260e89a6e0d6 ("mtd: core: tone down suggestion that dev.parent
should be set") removed the only user of caller_is_module() but forgot
to remove the definition itself. Let's remove it now.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: s3c2410: Error out when ->nrsets < 0 or ->sets == NULL
Boris Brezillon [Thu, 19 Jul 2018 20:53:50 +0000 (22:53 +0200)]
mtd: rawnand: s3c2410: Error out when ->nrsets < 0 or ->sets == NULL

All of the code in the probe path assumes ->sets != NULL and
->nrsets > 0. Error out if that's not the case to avoid dereferencing a
NULL pointer.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: make subop helpers return unsigned values
Miquel Raynal [Wed, 18 Jul 2018 22:09:12 +0000 (00:09 +0200)]
mtd: rawnand: make subop helpers return unsigned values

A report from Colin Ian King pointed a CoverityScan issue where error
values on these helpers where not checked in the drivers. These
helpers can error out only in case of a software bug in driver code,
not because of a runtime/hardware error. Hence, let's WARN_ON() in this
case and return 0 which is harmless anyway.

Fixes: 8878b126df76 ("mtd: nand: add ->exec_op() implementation")
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: orion: Handle cases where __LINUX_ARM_ARCH__ is not defined
Boris Brezillon [Wed, 18 Jul 2018 08:56:51 +0000 (10:56 +0200)]
mtd: rawnand: orion: Handle cases where __LINUX_ARM_ARCH__ is not defined

Make sure __LINUX_ARM_ARCH__ is defined before testing its value.
This is needed if we want to allow selection of this driver when
COMPILE_TEST=y.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: orion: Avoid direct inclusion of asm headers
Boris Brezillon [Wed, 18 Jul 2018 08:56:50 +0000 (10:56 +0200)]
mtd: rawnand: orion: Avoid direct inclusion of asm headers

Include linux/sizes.h instead of asm/sizes.h to make code completely
arch independent.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: partitions: use DT info for parsing partitions with "compatible" prop
Rafał Miłecki [Fri, 13 Jul 2018 14:32:21 +0000 (16:32 +0200)]
mtd: partitions: use DT info for parsing partitions with "compatible" prop

So far only flash devices could be described in DT regarding partitions
parsing. That could be done with "partitions" subnode and a proper
"compatible" string.

Some devices may use hierarchical (multi-level) layouts and may mix used
layouts (fixed and dynamic). Describing that in DT is done by specifying
"compatible" for DT-represented partition plus optionally more
properties and/or subnodes.

To support such layouts each DT partition has to be checked for
additional description.

Please note this implementation will work in parallel with support for
partition type specified for non-DT setups. That already works since
commit 1a0915be1926 ("mtd: partitions: add support for partition
parsers").

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agodt-bindings: mtd: explicitly document nesting partitions descriptions
Rafał Miłecki [Fri, 13 Jul 2018 14:32:20 +0000 (16:32 +0200)]
dt-bindings: mtd: explicitly document nesting partitions descriptions

Documentation was already saying that fixed and dynamic partitioning can
be mixed but was missing a clear description and examples. This commit
adds a proper documentation of how descriptions can be nested and how
layouts can be mixed.

This addition is important for partitions that contain subpartitions.
In such cases partitions have to be properly described in order to let
system handle them correctly.

Depending on situation, nesting descriptions may provide more accurate
logic/structure and/or allow mixing partitioning types (various
"compatible" values).

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Reviewed-by: Rob Herring <robh@kernel.org>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd/maps: fix solutionengine.c printk format warnings
Randy Dunlap [Tue, 24 Jul 2018 18:29:01 +0000 (11:29 -0700)]
mtd/maps: fix solutionengine.c printk format warnings

Fix 2 printk format warnings (this driver is currently only used by
arch/sh/) by using "%pap" instead of "%lx".

Fixes these build warnings:

../drivers/mtd/maps/solutionengine.c: In function 'init_soleng_maps':
../include/linux/kern_levels.h:5:18: warning: format '%lx' expects argument of type 'long unsigned int', but argument 2 has type 'resource_size_t' {aka 'unsigned int'} [-Wformat=]
../drivers/mtd/maps/solutionengine.c:62:54: note: format string is defined here
  printk(KERN_NOTICE "Solution Engine: Flash at 0x%08lx, EPROM at 0x%08lx\n",
                                                  ~~~~^
                                                  %08x
../include/linux/kern_levels.h:5:18: warning: format '%lx' expects argument of type 'long unsigned int', but argument 3 has type 'resource_size_t' {aka 'unsigned int'} [-Wformat=]
../drivers/mtd/maps/solutionengine.c:62:72: note: format string is defined here
  printk(KERN_NOTICE "Solution Engine: Flash at 0x%08lx, EPROM at 0x%08lx\n",
                                                                    ~~~~^
                                                                    %08x

Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Brian Norris <computersforpeace@gmail.com>
Cc: Boris Brezillon <boris.brezillon@bootlin.com>
Cc: Marek Vasut <marek.vasut@gmail.com>
Cc: Richard Weinberger <richard@nod.at>
Cc: linux-mtd@lists.infradead.org
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: Rich Felker <dalias@libc.org>
Cc: linux-sh@vger.kernel.org
Cc: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: lpddr: use mtd_device_register()
Rafał Miłecki [Sat, 21 Jul 2018 11:55:09 +0000 (13:55 +0200)]
mtd: lpddr: use mtd_device_register()

This driver doesn't specify parsers so it can use that little helper.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: micron: Make ECC activation stateful
Boris Brezillon [Wed, 18 Jul 2018 08:42:21 +0000 (10:42 +0200)]
mtd: rawnand: micron: Make ECC activation stateful

We currently don't store the on-die ECC state (enabled/disabled) which
might force us to re-disable the engine even if it's already been
disabled after we've read the page in raw mode to count the actual
number of bitflips.

Add an "enabled" field to struct micron_on_die_ecc to keep track of the
ECC state.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Tested-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: micron: Avoid enabling/disabling ECC when it can't be disabled
Boris Brezillon [Wed, 18 Jul 2018 08:42:20 +0000 (10:42 +0200)]
mtd: rawnand: micron: Avoid enabling/disabling ECC when it can't be disabled

Some chips have their on-die ECC forcibly enabled, there's no point in
trying to enable/disable the ECC engine in that case.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Tested-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: micron: Get the actual number of bitflips
Boris Brezillon [Wed, 18 Jul 2018 08:42:19 +0000 (10:42 +0200)]
mtd: rawnand: micron: Get the actual number of bitflips

The MT29F2Gxxx chips with 4bits/512byte on-die ECC let us know when
some bitflips were corrected by the on-die ECC, but they do not report
the actual number of bitflips that were present in the data+ECC chunk.

We initially decided to always return ecc->strength to avoid re-reading
the page in raw mode + comparing it to the corrected buffer to extract
the real number of bitflips, but this forces UBI to move data around as
soon as one bitflip is present in a page.

This not only wears the NAND out faster, but also degrades
performances, since reading a full PEB + writing it back to a different
PEB + erasing the old one is much more expensive than re-reading the
faulty page in raw mode and comparing it to the corrected buffer.
In most cases, the actual number of bitflips does not exceed the
bitflips threshold, and UBI won't have to move data around. Otherwise,
we can assume the time spent re-reading the page and doing the
comparison is negligible compared to the time UBI spends moving a full
PEB to another PEB.

Note that this logic is not applied to chips with 8bits/512byte on-die
ECC, because those chips provide fine-grained information (the maximum
error is 1 bit, and it will not force UBI to move blocks around at the
first bitflip).

Suggested-by: Bean Huo <beanhuo@micron.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: micron: allow forced on-die ECC
Chris Packham [Wed, 18 Jul 2018 08:42:18 +0000 (10:42 +0200)]
mtd: rawnand: micron: allow forced on-die ECC

Some Micron NAND chips have on-die ECC forceably enabled. Allow such
chips to be used as long as the controller has set chip->ecc.mode to
NAND_ECC_ON_DIE.

Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: Expose _notsupp() helpers for raw page accessors
Boris Brezillon [Wed, 18 Jul 2018 08:42:17 +0000 (10:42 +0200)]
mtd: rawnand: Expose _notsupp() helpers for raw page accessors

Some implementations simply can't disable their ECC engine. Expose
helpers returning -ENOTSUPP so that the caller knows that raw accesses
are not supported instead of silently falling back to non-raw
accessors.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: micron: support 8/512 on-die ECC
Chris Packham [Wed, 18 Jul 2018 08:42:16 +0000 (10:42 +0200)]
mtd: rawnand: micron: support 8/512 on-die ECC

Micron MT29F1G08ABAFAWP-ITE:F supports an on-die ECC with 8 bits
per 512 bytes. Add support for this combination.

Signed-off-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: micron: Fix on-die ECC detection logic
Boris Brezillon [Wed, 18 Jul 2018 08:42:15 +0000 (10:42 +0200)]
mtd: rawnand: micron: Fix on-die ECC detection logic

Basing the "mandatory on-die" detection on ID byte 2 does not work,
because Micron has plenty of NANDs using the same device ID code, and
not all of them have forcibly enabled on-die ECC.

Since the "Array Operation" feature does not provide the "ECC
enabled/disabled" bit when the ECC can't be disabled, let's try to use
the "ECC enabled/disabled" bit in the READ_ID bytes.

It seems that this bit is dynamically updated on NANDs where on-die ECC
can freely be enabled/disabled, so let's hope it stays at one when we
have a NAND with on-die ECC forcibly enabled.

Fixes: 51f3b3970a8c ("mtd: rawnand: micron: detect forced on-die ECC")
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Tested-by: Chris Packham <chris.packham@alliedtelesis.co.nz>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: fsmc: Stop using chip->read_buf()
Boris Brezillon [Wed, 18 Jul 2018 08:28:14 +0000 (10:28 +0200)]
mtd: rawnand: fsmc: Stop using chip->read_buf()

chip->read_buf is left unassigned since commit 4da712e70294 ("mtd: nand:
fsmc: use ->exec_op()"), leading to a NULL pointer dereference when it's
called from fsmc_read_page_hwecc(). Fix that by using the appropriate
helper to read data out of the NAND.

Fixes: 4da712e70294 ("mtd: nand: fsmc: use ->exec_op()")
Cc: <stable@vger.kernel.org>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: qcom: stop using phys_to_dma()
Arnd Bergmann [Tue, 17 Jul 2018 20:27:42 +0000 (22:27 +0200)]
mtd: rawnand: qcom: stop using phys_to_dma()

Compile-testing this driver on x86 caused a link error:

ERROR: "__phys_to_dma" [drivers/mtd/nand/raw/qcom_nandc.ko] undefined!

The problem here is that the driver attempts to convert the physical
address into the DMA controller as a dma_addr_t and calls phys_to_dma()
to do the conversion.

The correct way to do the conversion is using the dma mapping interfaces.

Fixes: c76b78d8ec05 ("mtd: nand: Qualcomm NAND controller driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: tegra: check bounds of die_nr properly
Stefan Agner [Tue, 17 Jul 2018 08:46:18 +0000 (10:46 +0200)]
mtd: rawnand: tegra: check bounds of die_nr properly

The Tegra driver currently only support a single chip select, hence
check boundaries accordingly. This fixes a off by one issue catched
with Smatch:
    drivers/mtd/nand/raw/tegra_nand.c:476 tegra_nand_select_chip()
    warn: array off by one? 'nand->cs[die_nr]'

Also warn in case the stack asks for a chip select we currently do
not support.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Stefan Agner <stefan@agner.ch>
Signed-off-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: add default values for dynamic timings
Miquel Raynal [Sat, 14 Jul 2018 10:23:54 +0000 (12:23 +0200)]
mtd: rawnand: add default values for dynamic timings

Some timings like tBERS (block erase time), tCCs (change column setup
time), tPROG (page program time) and tR (page read time) are derived
from the ONFI parameter page. They are set in the SDR interface only
if the chip is ONFI compliant.

It makes these timings unreliable and prevent the driver to use one of
these four values with a non-ONFI chip.

Fix this situation by taking the highest possible value (or a default
one) value for each missing timing (stored as unsigned 16-bit entries in
the parameter page).

This makes tBERS and tPROG being ~65ms while typical values are at most
a few milliseconds. As these are timeouts, it is not impacting at all
the performances in nominal use.

tR maximum time and tCCS minimum time (delay to wait after a change
column) are set, according to the ONFI specification, to default 'slow'
values; respectively 200us and 500ns.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: onenand: use mtd_device_register() where applicable
Rafał Miłecki [Fri, 13 Jul 2018 09:27:32 +0000 (11:27 +0200)]
mtd: onenand: use mtd_device_register() where applicable

If driver doesn't specify parsers it can use that little helper.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: use mtd_device_register() where applicable
Rafał Miłecki [Fri, 13 Jul 2018 09:27:31 +0000 (11:27 +0200)]
mtd: rawnand: use mtd_device_register() where applicable

If driver doesn't specify parsers it can use that little helper.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Acked-by: Xiaolei Li <xiaolei.li@mediatek.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: MTD_NAND_BCM47XXNFLASH needs CONFIG_BCMA
Arnd Bergmann [Wed, 11 Jul 2018 08:39:44 +0000 (10:39 +0200)]
mtd: rawnand: MTD_NAND_BCM47XXNFLASH needs CONFIG_BCMA

We already have a dependency on BCMA_NFLASH, which in turn depends on
BCMA, but since BCMA is a tristate option and BCMA_NFLASH is bool,
we can run into an invalid configuration with MTD_NAND_BCM47XXNFLASH=y
and BCMA=m:

drivers/mtd/nand/raw/bcm47xxnflash/ops_bcm4706.o: In function `bcm47xxnflash_ops_bcm4706_init':
ops_bcm4706.c:(.text+0x790): undefined reference to `bcma_chipco_pll_read'

Adding the dependency here forces MTD_NAND_BCM47XXNFLASH to only be
configured =m here so it can link against the BCMA driver.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomemory: jz4780-nemc: Allow selection of this driver when COMPILE_TEST=y
Boris Brezillon [Mon, 9 Jul 2018 20:09:45 +0000 (22:09 +0200)]
memory: jz4780-nemc: Allow selection of this driver when COMPILE_TEST=y

It just makes maintainers' life easier by allowing them to compile-test
this driver without having MACH_JZ4780 enabled.

We also need to add a dependency on HAS_IOMEM to make sure the
driver compiles correctly.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Paul Burton <paul.burton@mips.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: jz4740: Use the proper format specifier to print chipnr
Boris Brezillon [Thu, 19 Jul 2018 07:37:20 +0000 (09:37 +0200)]
mtd: rawnand: jz4740: Use the proper format specifier to print chipnr

In jz_nand_detect_bank(), chipnr is a size_t argument. Use %zu instead
of %i when printing it.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: adapt misleading comment in mtd_oob_ops structure
Miquel Raynal [Wed, 18 Jul 2018 20:33:02 +0000 (22:33 +0200)]
mtd: adapt misleading comment in mtd_oob_ops structure

A comment in the kernel doc of the mtd_oob_ops structure tells that it
is not possible to write more than one page with OOB. This is actually
true for only a few MTD devices like 'onenand' but it is definitely not
a general limitation. While this would benefit to be handled elsewhere
either by the MTD layer or by the limited drivers, let's update this
comment to reflect the reality.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: Make Kconfig formatting consistent
Boris Brezillon [Wed, 18 Jul 2018 15:09:52 +0000 (17:09 +0200)]
mtd: Make Kconfig formatting consistent

Fix indentation and replace '---help---' by 'help' to make things
consistent.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Reviewed-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtdchar: fix overflows in adjustment of `count`
Jann Horn [Sat, 7 Jul 2018 03:37:22 +0000 (05:37 +0200)]
mtdchar: fix overflows in adjustment of `count`

The first checks in mtdchar_read() and mtdchar_write() attempt to limit
`count` such that `*ppos + count <= mtd->size`. However, they ignore the
possibility of `*ppos > mtd->size`, allowing the calculation of `count` to
wrap around. `mtdchar_lseek()` prevents seeking beyond mtd->size, but the
pread/pwrite syscalls bypass this.

I haven't found any codepath on which this actually causes dangerous
behavior, but it seems like a sensible change anyway.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: Fallback to ->_read/write() when ->_read/write_oob() is missing
Miquel Raynal [Sat, 14 Jul 2018 12:42:00 +0000 (14:42 +0200)]
mtd: Fallback to ->_read/write() when ->_read/write_oob() is missing

Some MTD sublayers/drivers are implementing ->_read/write() and
not ->_read/write_oob().

While for NAND devices both are usually valid, for NOR devices, using
the _oob variant has no real meaning. But, as the MTD layer is supposed
to hide as much as possible the flash complexity to the user, there is
no reason to error out while it is just a matter of rewritting things
internally.

Add a fallback on mtd->_read() (resp. mtd->_write()) when the user calls
mtd_read_oob() (resp. mtd_write_oob()) while mtd->_read_oob() (resp.
mtd->_write_oob) is not implemented. There is already a fallback on the
_oob variant if the former is used.

Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agojffs2: use unsigned 32-bit timstamps consistently
Arnd Bergmann [Fri, 13 Jul 2018 14:47:17 +0000 (16:47 +0200)]
jffs2: use unsigned 32-bit timstamps consistently

Most users of jffs2 are 32-bit systems that traditionally only support
timestamps using a 32-bit signed time_t, in the range from years 1902 to
2038. On 64-bit systems, jffs2 however interpreted the same timestamps
as unsigned values, reading back negative times (before 1970) as times
between 2038 and 2106.

Now that Linux supports 64-bit inode timestamps even on 32-bit systems,
let's use the second interpretation everywhere to allow jffs2 to be
used on 32-bit systems beyond 2038 without a fundamental change to the
inode format.

This has a slight risk of regressions, when existing files with timestamps
before 1970 are present in file system images and are now interpreted
as future time stamps. I considered moving the wraparound point a bit,
e.g. to 1960, in order to deal with timestamps that ended up on Dec 31,
1969 due to incorrect timezone handling. However, this would complicate
the implementation unnecessarily, so I went with the simplest possible
method of extending the timestamps.

Writing files with timestamps before 1970 or after 2106 now results
in those times being clamped in the file system.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agojffs2: use 64-bit intermediate timestamps
Arnd Bergmann [Fri, 13 Jul 2018 14:47:16 +0000 (16:47 +0200)]
jffs2: use 64-bit intermediate timestamps

The VFS now uses timespec64 timestamps consistently, but jffs2 still
converts them to 32-bit numbers on the storage medium. As the helper
functions for the conversion (get_seconds() and timespec_to_timespec64())
are now deprecated, let's change them over to the more modern
replacements.

This keeps the traditional interpretation of those values, where
the on-disk 32-bit numbers are taken to be negative numbers, i.e.
dates before 1970, on 32-bit machines, but future numbers past 2038
on 64-bit machines.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: sst25l: use mtd_device_register()
Rafał Miłecki [Fri, 13 Jul 2018 09:27:34 +0000 (11:27 +0200)]
mtd: sst25l: use mtd_device_register()

This driver doesn't specify parsers so it can use that little helper.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: maps: use mtd_device_register() where applicable
Rafał Miłecki [Fri, 13 Jul 2018 09:27:33 +0000 (11:27 +0200)]
mtd: maps: use mtd_device_register() where applicable

If driver doesn't specify parsers it can use that little helper.

Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: powernv_flash: set of_node in mtd's dev
Rafał Miłecki [Fri, 13 Jul 2018 08:15:59 +0000 (10:15 +0200)]
mtd: powernv_flash: set of_node in mtd's dev

This enables some features implemented in mtd subsystem like reading
label and partitioning info from DT.

Reported-by: Timothy Pearson <tpearson@raptorengineering.com>
Signed-off-by: Rafał Miłecki <rafal@milecki.pl>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
6 years agomtd: rawnand: jz4780: Drop the dependency on MACH_JZ4780
Boris Brezillon [Mon, 9 Jul 2018 20:09:44 +0000 (22:09 +0200)]
mtd: rawnand: jz4780: Drop the dependency on MACH_JZ4780

This MACH_JZ4780 dependency is taken care of by JZ4780_NEMC, no need
to repeat it here.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Paul Burton <paul.burton@mips.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: jz4740: Allow selection of this driver when COMPILE_TEST=y
Boris Brezillon [Mon, 9 Jul 2018 20:09:43 +0000 (22:09 +0200)]
mtd: rawnand: jz4740: Allow selection of this driver when COMPILE_TEST=y

It just makes NAND maintainers' life easier by allowing them to
compile-test this driver without having MACH_JZ4740 enabled.

We also need to add a dependency on HAS_IOMEM to make sure the driver
compiles correctly.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Paul Burton <paul.burton@mips.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agoMIPS: jz4740: Move jz4740_nand.h header to include/linux/platform_data/jz4740
Boris Brezillon [Mon, 9 Jul 2018 20:09:42 +0000 (22:09 +0200)]
MIPS: jz4740: Move jz4740_nand.h header to include/linux/platform_data/jz4740

This way we will be able to compile the jz4740_nand driver when
COMPILE_TEST=y.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Paul Burton <paul.burton@mips.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: txx9ndfmc: Allow selection of this driver when COMPILE_TEST=y
Boris Brezillon [Mon, 9 Jul 2018 20:09:41 +0000 (22:09 +0200)]
mtd: rawnand: txx9ndfmc: Allow selection of this driver when COMPILE_TEST=y

It just makes NAND maintainers' life easier by allowing them to
compile-test this driver without having SOC_TX4938 or SOC_TX4939
enabled.

We also need to add a dependency on HAS_IOMEM to make sure the driver
compiles correctly.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Paul Burton <paul.burton@mips.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agoMIPS: txx9: Move the ndfc.h header to include/linux/platform_data/txx9
Boris Brezillon [Mon, 9 Jul 2018 20:09:40 +0000 (22:09 +0200)]
MIPS: txx9: Move the ndfc.h header to include/linux/platform_data/txx9

This way we will be able to compile the ndfmc driver when
COMPILE_TEST=y.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Acked-by: Paul Burton <paul.burton@mips.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: fsl_ifc: Allow selection of this driver when COMPILE_TEST=y
Boris Brezillon [Mon, 9 Jul 2018 20:09:39 +0000 (22:09 +0200)]
mtd: rawnand: fsl_ifc: Allow selection of this driver when COMPILE_TEST=y

It just makes maintainers' life easier by allowing them to compile-test
this driver without having FSL_SOC, ARCH_LAYERSCAPE or SOC_LS1021A
enabled.

We also need to add a dependency on HAS_IOMEM to make sure the
driver compiles correctly.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: fsl_ifc: Add an __iomem specifier on eccstat_regs
Boris Brezillon [Mon, 9 Jul 2018 20:09:38 +0000 (22:09 +0200)]
mtd: rawnand: fsl_ifc: Add an __iomem specifier on eccstat_regs

The local eccstat_regs variable in fsl_ifc_run_command() is missing an
__iomem specifier, and sparce complains about that.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomemory: fsl_ifc: Allow selection of this driver when COMPILE_TEST=y
Boris Brezillon [Mon, 9 Jul 2018 20:09:37 +0000 (22:09 +0200)]
memory: fsl_ifc: Allow selection of this driver when COMPILE_TEST=y

It just makes maintainers' life easier by allowing them to compile-test
this driver without having FSL_SOC, ARCH_LAYERSCAPE or SOC_LS1021A
enabled.

We also need to add a dependency on HAS_IOMEM to make sure the
driver compiles correctly.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: fsmc: Allow selection of this driver when COMPILE_TEST=y
Boris Brezillon [Mon, 9 Jul 2018 20:09:36 +0000 (22:09 +0200)]
mtd: rawnand: fsmc: Allow selection of this driver when COMPILE_TEST=y

It just makes NAND maintainers' life easier by allowing them to
compile-test this driver without having PLAT_SPEAR, ARCH_NOMADIK,
ARCH_U8500 or MACH_U300 enabled.

We also need to add a dependency on HAS_IOMEM to make sure the driver
compiles correctly.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: fsmc: Use uintptr_t casts instead of unsigned ones
Boris Brezillon [Mon, 9 Jul 2018 20:09:35 +0000 (22:09 +0200)]
mtd: rawnand: fsmc: Use uintptr_t casts instead of unsigned ones

uintptr_t should be used when casting a pointer to an unsigned int so
that the code compiles without warnings even on 64-bit architectures.

This is needed if we want to allow selection of this driver when
COMPILE_TEST=y.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: fscm: Avoid collision on PC def when compiling for MIPS
Boris Brezillon [Mon, 9 Jul 2018 20:09:34 +0000 (22:09 +0200)]
mtd: rawnand: fscm: Avoid collision on PC def when compiling for MIPS

We want to allow this driver to be selected when COMPILE_TEST=y, this
means the driver can be compiled for any arch, including MIPS. When
compiling this driver for MIPS, we end up with a collision on the 'PC'
macro definition (also defined in arch/mips/include/asm/ptrace.h).

Prefix the fsmc one with FSMC_ to avoid this problem.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: sunxi: Allow selection of this driver when COMPILE_TEST=y
Boris Brezillon [Mon, 9 Jul 2018 20:09:33 +0000 (22:09 +0200)]
mtd: rawnand: sunxi: Allow selection of this driver when COMPILE_TEST=y

It just makes NAND maintainers' life easier by allowing them to
compile-test this driver without having ARCH_SUNXI enabled.

We also need to add a dependency on HAS_IOMEM to make sure the driver
compiles correctly.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: sunxi: Make sure ret is initialized in sunxi_nfc_read_byte()
Boris Brezillon [Mon, 9 Jul 2018 20:09:32 +0000 (22:09 +0200)]
mtd: rawnand: sunxi: Make sure ret is initialized in sunxi_nfc_read_byte()

Fixes the following smatch warning:

drivers/mtd/nand/raw/sunxi_nand.c:551 sunxi_nfc_read_byte() error: uninitialized symbol 'ret'.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: sunxi: Add an U suffix to NFC_PAGE_OP definition
Boris Brezillon [Mon, 9 Jul 2018 20:09:31 +0000 (22:09 +0200)]
mtd: rawnand: sunxi: Add an U suffix to NFC_PAGE_OP definition

Fixes the "warning: large integer implicitly truncated to unsigned type
[-Woverflow]" warning when compiled for x86.

This is needed in order to allow compiling this driver when
COMPILE_TEST=y.

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>
6 years agomtd: rawnand: davinci: Allow selection of this driver when COMPILE_TEST=y
Boris Brezillon [Mon, 9 Jul 2018 20:09:30 +0000 (22:09 +0200)]
mtd: rawnand: davinci: Allow selection of this driver when COMPILE_TEST=y

It just makes NAND maintainers' life easier by allowing them to
compile-test this driver without having ARCH_DAVINCI or ARCH_KEYSTONE
enabled.

We also need to add a dependency on HAS_IOMEM to make sure the
driver compiles correctly.

Signed-off-by: Boris Brezillon <boris.brezillon@bootlin.com>
Signed-off-by: Miquel Raynal <miquel.raynal@bootlin.com>