mtd: devices: elm: update DRIVER_NAME as "omap-elm"

use "omap-elm" as DRIVER_NAME

Signed-off-by: Pekon Gupta <pekon@ti.com>
Reviewed-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: devices: elm: configure parallel channels based on ecc_steps

ELM hardware can process up to maximum of 8 hannels in parallel for
ECC error detection. Currently the number of channels getting configured for
processing is static determined by macro ERROR_VECTOR_MAX. However, the actual
number of channels that need to be processed is the ECC step number.
This patch just avoids configuring extra unused channels.

Signed-off-by: Pekon Gupta <pekon@ti.com>
Reviewed-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: devices: elm: clean elm_load_syndrome

This patch refactors elm_load_syndrome() to make it scalable for newer
ECC schemes by removing scheme specific macros (like ECC_BYTES*xx),
and instead using ECC control information passed during elm_config.

Signed-off-by: Pekon Gupta <pekon@ti.com>
Reviewed-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: devices: elm: check for hardware engine's design constraints

ELM hardware engine is used by BCH ecc-schemes for detecting and locating ECC
errors. This patch adds the following checks for ELM hardware engine:

 - ELM internal buffers are of 1K,
   so it cannot process data with ecc-step-size > 1K.

 - ELM engine can execute upto maximum of 8 threads in parallel,
   so in *page-mode* (when complete page is processed in single iteration),
   ELM cannot support ecc-steps > 8.

Signed-off-by: Pekon Gupta <pekon@ti.com>
Reviewed-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Succinctly reorganise .remove()

Reported-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Allow loop to run at least once before giving up CPU

Reported-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Correct vendor name spelling issue - missing "M"

Reported-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Avoid duplicating MTD core code

Reported-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Remove useless consts from function arguments

Reported-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
[Brian: tweaked a bit]
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Convert ST SPI FSM (NOR) Flash driver to new DT partitions

The old API expected a "partitions" property provided a phandle to a
separate partitions node, which itself contained yet more nodes each
representing one partition. The new API rids the requirement for the
superfluous intermediary partitions node. This patch provides the
added information required for automatic parsing by the core.

Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Move runtime configurable msg sequences into device's struct

Until now the dynamically configurable message sequences for read, write
and enable 32bit addressing have been global. Brian makes a good point
why this should not be the case. If there are ever two FSM's located on
the same platform, we could be potentially introducing a race condition
on "needlessly shared data".

Suggested-by: Brian Norris <computersforpeace@gmail.com>
Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply the W25Qxxx chip specific configuration call-back

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply the S25FLxxx chip specific configuration call-back

This patch allows us to prepare some of the message sequences which will
be required to talk to the  S25FLxxx family of Serial Flash devices. It
also allows us to do some required extra operations after any busy wait
failures.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply the MX25xxx chip specific configuration call-back

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Add the ability to write to FSM's status register

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Add the ability to read the FSM's status

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Erase partly or as a whole a Serial Flash device

When an erase is requested by userspace the MTD framework calls back
into the driver to conduct the actual command issue. Here we provide the
routines which do exactly that. We can choose to either do an entire chip
erase or by sector.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply a busy wait for post-write status

When we write data to the Serial Flash chip we'll wait a predetermined
period of time before giving up. During that period of time we poll the
status register until completion.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Write to Flash via the FSM FIFO

When we write data to the FIFO the FSM Controller subsequently writes
that data out to the Serial Flash chip.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Add the ability to read from a Serial Flash device

When a read is issued by userspace the MTD framework calls back into
the driver to conduct the actual command issue and data extraction.
Here we provide the routines which do exactly that.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Prepare default sequences for read/write/erase

Most chips require a predefined set of FSM message sequences for read,
write and erase operations. This patch provides a way to set them up,
which it will do so if a chip specific initialisation routine isn't
been provided.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply the N25Qxxx chip specific configuration call-back

In the FSM driver we handle chip differences by providing the possibility
of calling back into a chip specific initialisation routine. In this patch
we provide one for the N25Qxxx series, which endeavours to setup things
like the read, write and erase sequences, as they differ from the
default. We also configure 32bit support and the amount of dummy cycles to
use.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply the N25Qxxx specific read configurations

The N25Qxxx Serial Flash devices required different sequence
configurations depending on whether they're running in 24bit (3Byte)
or 32bit (4Byte) mode. We provide those here.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Provide the default read/write configurations

Message sequences can vary depending on how many pads (lines) are
required to address the chip (mode & dummy), how many data pads (lines)
are required to write out to the chip which will determine speed
amongst other things which are detailed by the SFDP specification. We
are able to use multiple configurations for each chip, but they need
to me matched to a device's capabilities. These configurations are
listed in preference order - most preferred first.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Update the flash Volatile Configuration Register

The FSM Serial Flash Controller is driven by issuing a standard set of
register writes we call a message sequence. This patch supplies a method
to prepare the message sequence responsible for updating a chip's VCR.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Provide a method to put the chip into 32bit addressing mode

Most Serial Flash chips support 24bit addressing as a default but more
recent incarnations can support 32bit. Based on information provided
though platform specific data and capabilities we can determine whether
or not our current chip can. This patch provides a means to setup the
FSM message sequence to put the chip into 32bit mode.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Add a check to if the chip can handle an SoC reset

Based on information we can obtain though platform specific data and/or
chip capabilities we are able to determine whether or not we can handle
a SoC reset or not. To find out why this is important please read the
comment provided in the patch.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Prepare read/write sequences according to configuration

Firstly we search for our preference read/write configuration based on a
given chip's capabilities. Then we actually set up the message sequence
accordingly.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Provide the sequence for enabling 32bit addressing mode

The FSM Serial Flash Controller is driven by issuing a standard set of
register writes we call a message sequence. This patch supplies a method
to prepare the message sequence responsible for setting 32bit addressing
mode on the Flash chip.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Provide the erase one sector sequence

The FSM Serial Flash Controller is driven by issuing a standard set of
register writes we call a message sequence. This patch supplies a method
to prepare the message sequence responsible for erasing a single sector.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Fetch boot-device from mode pins

It's important for us to determine which device was used to boot from in
order to make some correct decisions surrounding Power Management. On
each of the platforms which support the FSM this is communicated via
a set of mode pins held in the system configuration area. This patch
determine the boot device and stores the result.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Add device-tree binding documentation

Cc: devicetree@vger.kernel.org
Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Prepare the read/write FSM message sequence(s)

The FSM Serial Flash Controller is driven by issuing a standard set of
register writes we call a message sequence. This patch supplies a method
to prepare read/write FSM message sequence(s) based on chip capability
and configuration.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Use device size to determine address width

Take some known parameters, namely size and number of sectors and use
them to determine weather a device can support 32bit addressing or not.
If it can, set the associated flash capability flag for latter use.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Search for preferred FSM message sequence configurations

Here we provide a means to traverse though all supplied FSM message
sequence configurations and pick one based on our chip's capabilities.
The first one we match will be the preferred one, as they are
presented in order of preference.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Dynamically setup flash device based on JEDEC ID

Using previously added infrastructure we can now extract a device's JEDEC
ID, compare it to a list of known and supported devices and make assumptions
based on known characteristics of a given chip.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Provide device look-up table

Supply a lookup table of all the devices we intend to support. This table
is used to store device information such as; a human readable device name,
their JEDEC ID (plus the extended version), sector size and amount, a bit
store of a device's capabilities, its maximum running frequency and
possible use of a per-device configuration call-back.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: devices: Provide header for shared OPCODEs and SFDP commands

JEDEC have helped to standardise a great deal of the commands which
can be issued to a Serial Flash devices. Many of the Serial Flash
Discoverable Parameters (SFDP) commands are generic across devices.
This patch provides a shared point where these commands can be
defined.

Suggested-by: Mark Brown <broonie@kernel.org>
Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Add support for JEDEC ID extraction

Once we start supporting devices it will be handy go detect them
dynamically. This will be done using the chip's unique JEDEC ID. This
patch allows us to extract a device's JEDEC ID using the a predefined
FSM register write sequence.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply a method to read from the FSM's FIFO

When invoked the driver will attempt to read any available data from
the FSM's data register. Any data collected from this FIFO would have
originated from the flash chip.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply framework for device requests

The FSM hardware works by setting a predetermined sequence of register
writes. Rather than open coding them inside each functional block we're
going to define them in a series of formatted 'sequence structures'.
This patch provides the framework which shall be used for every action.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Initialise and configure the FSM for normal working conditions

This patch uses default values to initialise a connected flash chip. This
includes; a device soft reset, setting of a safe working frequency, a
switch into Fast Sequencing Mode, configuring of timing data and a purge
of the FIFO.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Supply all register address and bit logic defines

Here we provide the FSM's register addresses, register bit names/offsets
and some commands which will prove useful as we start bulk the FMS's
driver out with functionality.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: st_spi_fsm: Allocate resources and register with MTD framework

This is a new driver. It's used to communicate with a special type of
optimised Serial Flash Controller called the FSM. The FSM uses a subset
of the SPI protocol to communicate with supported NOR-Flash devices.

Acked-by Angus Clark <angus.clark@st.com>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: remove is_elm_present flag

'is_elm_present' flag is not used anywhere. This check is implicitely
taken care while selecting appropriate ecc-scheme via DT or board-file.

Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.hwctl: clean omap_enable_hwecc_bch for redundant ECC configs

This patch
 - refactors GPMC configurations based on ecc-scheme
 - removed dependency on is_elm_present() flag, which is implicitely
   taken care by selecting appropriate ecc-scheme

Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.hwctl: refactor omap_enable_hwecc_bch for ECC related GPMC configs

Lots of if..then..else conditions in omap_enable_hwecc_bch() can be avoided if
code is refactored based on ecc-scheme.

Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.hwctl: rename omap3_enable_hwecc_bch to omap_enable_hwecc_bch

This patch
 - renames omap3_enable_hwecc_bch -> omap_enable_hwecc_bch to keep
   nomenclature independent of any device family.
 - using '__maybe_unused' instead of `ifdef based conditional compilation
   to suppress warning for un-used functions

Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.calculate: merge omap3_calculate_ecc_bch8 in omap_calculate_ecc_bch

merge omap3_calculate_ecc_bch8() into omap_calculate_ecc_bch() so that
common callback can be used for both OMAP_ECC_BCH8_CODE_HW and
OMAP_ECC_BCH8_CODE_HW_DETECTION_SW
+---------------------+-------------------------------------------------------+
|ecc-scheme           | nand_chip->calculate() after this patch               |
+---------------------+-------------------------------------------------------+
|HAM1_ECC             | omap_calculate_ecc()                                  |
+---------------------+-------------------------------------------------------+
|BCH4_HW_DETECTION_SW | omap_calculate_ecc_bch()                              |
|BCH4_HW              | omap_calculate_ecc_bch()                              |
|BCH8_HW_DETECTION_SW | omap3_calculate_ecc_bch8() -> omap_calculate_ecc_bch()|
|BCH8_HW              | omap_calculate_ecc_bch()                              |
+---------------------+-------------------------------------------------------+

Tested-by: Stefan Roese <sr@denx.de>
Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.calculate: merge omap3_calculate_ecc_bch4 in omap_calculate_ecc_bch

merges omap3_calculate_ecc_bch4() into omap_calculate_ecc_bch() so that
common callback can be used for both OMAP_ECC_BCH4_CODE_HW and
OMAP_ECC_BCH4_CODE_HW_DETECTION_SW ecc-schemes

+---------------------+-------------------------------------------------------+
|ecc-scheme           | nand_chip->calculate() after this patch               |
+---------------------+-------------------------------------------------------+
|HAM1_ECC             | omap_calculate_ecc()                                  |
+---------------------+-------------------------------------------------------+
|BCH4_HW_DETECTION_SW | omap3_calculate_ecc_bch4() -> omap_calculate_ecc_bch()|
|BCH4_HW              | omap_calculate_ecc_bch()                              |
|BCH8_HW_DETECTION_SW | omap3_calculate_ecc_bch8()                            |
|BCH8_HW              | omap_calculate_ecc_bch()                              |
+---------------------+-------------------------------------------------------+

Tested-by: Stefan Roese <sr@denx.de>
Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.calculate: refactor omap_calculate_ecc_bch for BCHx_HW ecc-scheme

OMAP NAND driver supports multiple flavours of BCH4 and BCH8 ECC algorithms.
+------+------------------------------------+---------------+---------------+
| Algo | ECC scheme                         |ECC calculation|Error detection|
+------+------------------------------------+---------------+---------------+
|      |OMAP_ECC_BCH4_CODE_HW_DETECTION_SW  |H/W (GPMC)     |S/W            |
| BCH4 |OMAP_ECC_BCH4_CODE_HW               |H/W (GPMC)     |H/W (ELM)      |
+------+------------------------------------+---------------+---------------+
|      |OMAP_ECC_BCH8_CODE_HW_DETECTION_SW  |H/W (GPMC)     |S/W            |
| BCH8 |OMAP_ECC_BCH8_CODE_HW               |H/W (GPMC)     |H/W (ELM)      |
+------+------------------------------------+---------------+---------------+

This patch refactors omap_calculate_ecc_bch() so that
 - separate out ecc-scheme specific code so that common-code can be reused
   between different implementations of same ECC algorithm.
 - new ecc-schemes can be added with ease in future.

Tested-by: Stefan Roese <sr@denx.de>
Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.calculate: rename omap3_calculate_ecc_bch to omap_calculate_ecc_bch

rename omap3_calculate_ecc_bch -> omap_calculate_ecc_bch to
keep nomenclature independent of any device family.

Tested-by: Stefan Roese <sr@denx.de>
Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.correct: omap_elm_correct_data: fix programmed-page bit-flip correction logic

This patch updates following checks when bit-flips are detected by ELM:

 - Do not evaluate bit-flips when un-correctable bit-flips is reported by ELM,
   because as per [1] when ELM reports an un-correctable bit-flips,
   'number of error' field in its ELM_LOCATION_STATUS register is also invalid.

 - Return with error-code '-EBADMSG' on detection of un-correctable bit-flip.

 - Return with error-code '-EBADMSG' when bit-flips position is outside current
   Sector and OOB area.

[1] ELM IP spec Table-25 ELM_LOCATION_STATUS Register.
    ELM_LOCATION_STATUS[8] = ECC_CORRECTABLE: Error location process exit status
        0x0: ECC error location process failed.
             Number of errors and error locations are invalid.
        0x1: all errors were successfully located.
             Number of errors and error locations are valid.

Tested-by: Stefan Roese <sr@denx.de>
Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.correct: omap_elm_correct_data: cleanup for future enhancements

Current omap_elm_correct_data() code is not scalable for future ecc-schemes
due to presence of tweaks and hard-coded macros for BCH4_ECC and BCH8_ECC
ecc-schemes at multiple places.

This patch:
 - replaces 'ecc_opt' with '(info->nand.ecc.strength == BCH8_MAX_ERROR)
   used to differentiate between BCH8_HW and BCH4_SW
 - replaces macros (defining magic number for specific ecc-scheme) with
   generic variables
 - removes dependency on macros defined in elm.h (like BCHx_ECC_OOB_BYTES)

Tested-by: Stefan Roese <sr@denx.de>
Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.correct: omap_elm_correct_data: fix erased-page detection for BCHx_HW ECC schemes

As erased-pages do not have ECC stored in their OOB area, so they need to be
seperated out from programmed-pages, before doing BCH ECC correction.

In current implementation of omap_elm_correct_data() which does ECC correction
for BCHx ECC schemes, this erased-pages are detected based on specific marker
byte (reserved as 0x00) in ecc-layout.
However, this approach has some limitation like;
 1) All ecc-scheme layouts do not have such Reserved byte marker to
    differentiate between erased-page v/s programmed-page. Thus this is a
    customized solution.
 2) Reserved marker byte can itself be subjected to bit-flips causing
    erased-page to be misunderstood as programmed-page.

This patch removes dependency on any marker byte in ecc-layout, instead it
compares calc_ecc[] with pattern of ECC-of-all(0xff). This implicitely
means that both 'data + oob == all(0xff).

Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: ecc.correct: omap_elm_correct_data: rename ambiguous variable 'eccsize' and 'ecc_vector_size'

renaming following variables as they cause confusion due to resemblence to
another similar field in 'struct nand_ecc_ctrl' (nand_chip->ecc.size).
renaming: ecc_vector_size --> ecc->bytes (info->nand.ecc.bytes)
renaming: eccsize         --> actual_eccbytes (info->nand.ecc.bytes - 1) for BCH4 and BCH8

Tested-by: Stefan Roese <sr@denx.de>
Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: omap: add field to indicate current ecc-scheme in 'struct omap_nand_info'

Information of currently selected ECC scheme 'enum omap_ecc ecc_opt' should
available outside platform-data, so that single nand_chip->ecc callback can
support multiple ecc-scheme configurations.

Tested-by: Stefan Roese <sr@denx.de>
Signed-off-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: Fix the behavior of OTP write if there is not enough room for data

If a write to one time programmable memory (OTP) hits the end of this
memory area, no more data can be written. The count variable in
mtdchar_write() in drivers/mtd/mtdchar.c is not decreased anymore.
We are trapped in the loop forever, mtdchar_write() will never return
in this case.

The desired behavior of a write in such a case is described in [1]:
- Try to write as much data as possible, truncate the write to fit into
  the available memory and return the number of bytes that actually
  have been written.
- If no data could be written at all, return -ENOSPC.

This patch fixes the behavior of OTP write if there is not enough space
for all data:

1) mtd_write_user_prot_reg() in drivers/mtd/mtdcore.c is modified to
   return -ENOSPC if no data could be written at all.
2) mtdchar_write() is modified to handle -ENOSPC correctly. Exit if a
   write returned -ENOSPC and yield the correct return value, either
   then number of bytes that could be written, or -ENOSPC, if no data
   could be written at all.

Furthermore the patch harmonizes the behavior of the OTP memory write
in drivers/mtd/devices/mtd_dataflash.c with the other implementations
and the requirements from [1]. Instead of returning -EINVAL if the data
does not fit into the OTP memory, we try to write as much data as
possible/truncate the write.

[1] http://pubs.opengroup.org/onlinepubs/9699919799/functions/write.html

Signed-off-by: Christian Riesch <christian.riesch@omicron.at>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: block2mtd: check device size

fixme applied : check device size is a multiple of erasesize.

Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: atmel_nand: Disable subpage NAND write when using Atmel PMECC

Crash detected on sam5d35 and its pmecc nand ecc controller.

The problem was a call to chip->ecc.hwctl from nand_write_subpage_hwecc
(nand_base.c) when we write a sub page.
chip->ecc.hwctl function is not set when we are using PMECC controller.
As a workaround, set NAND_NO_SUBPAGE_WRITE for PMECC controller in
order to disable sub page access in nand_write_page.

Signed-off-by: Herve Codina <Herve.CODINA@celad.com>
Acked-by: Josh Wu <josh.wu@atmel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: gpmi: add subpage read support

1) Why add the subpage read support?
  The page size of the nand chip becomes larger and larger, the imx6 has to
  supports the 16K page or even bigger page. But sometimes, the upper layer only
  needs a small part of the page, such as 512 bytes or less.

  For example, ubiattach may only read 64 bytes per page.

2) We only enable the subpage read support when it meets the conditions:
   <1> the chip is imx6 (or later chips) which can supports large nand page.
   <2> the size of ECC parity is byte aligned.
       If the size of ECC parity is not byte aligned, the calling of NAND_CMD_RNDOUT
       will fail.

3) What does this patch do?
   This patch will fake a virtual small page for the subpage read, and call the
   gpmi_ecc_read_page() to do the real work.

   In order to fake a virtual small page, the patch changes the BCH registers and
   the bch_geometry{}. After the subpage read finished, we will restore them back.

4) Performace:
    4.1) Tested with Toshiba TC58NVG2S0F(4096 + 224) with the following command:
         #ubiattach /dev/ubi_ctrl -m 4

       The detail information of /dev/mtd4 shows below:
       --------------------------------------------------------------
       #mtdinfo /dev/mtd4
        mtd4
        Name:                           test
        Type:                           nand
        Eraseblock size:                262144 bytes, 256.0 KiB
        Amount of eraseblocks:          1856 (486539264 bytes, 464.0 MiB)
        Minimum input/output unit size: 4096 bytes
        Sub-page size:                  4096 bytes
        OOB size:                       224 bytes
        Character device major/minor:   90:8
        Bad blocks are allowed:         true
        Device is writable:             true
       --------------------------------------------------------------

    4.2) Before this patch:
       --------------------------------------------------------------
       [   94.530495] UBI: attaching mtd4 to ubi0
       [   98.928850] UBI: scanning is finished
       [   98.953594] UBI: attached mtd4 (name "test", size 464 MiB) to ubi0
       [   98.958562] UBI: PEB size: 262144 bytes (256 KiB), LEB size: 253952 bytes
       [   98.964076] UBI: min./max. I/O unit sizes: 4096/4096, sub-page size 4096
       [   98.969518] UBI: VID header offset: 4096 (aligned 4096), data offset: 8192
       [   98.975128] UBI: good PEBs: 1856, bad PEBs: 0, corrupted PEBs: 0
       [   98.979843] UBI: user volume: 1, internal volumes: 1, max. volumes count: 128
       [   98.985878] UBI: max/mean erase counter: 2/1, WL threshold: 4096, image sequence number: 2024916145
       [   98.993635] UBI: available PEBs: 0, total reserved PEBs: 1856, PEBs reserved for bad PEB handling: 40
       [   99.001807] UBI: background thread "ubi_bgt0d" started, PID 831
       --------------------------------------------------------------
       The attach time is about 98.9 - 94.5 = 4.4s

    4.3) After this patch:
       --------------------------------------------------------------
       [  286.464906] UBI: attaching mtd4 to ubi0
       [  289.186129] UBI: scanning is finished
       [  289.211416] UBI: attached mtd4 (name "test", size 464 MiB) to ubi0
       [  289.216360] UBI: PEB size: 262144 bytes (256 KiB), LEB size: 253952 bytes
       [  289.221858] UBI: min./max. I/O unit sizes: 4096/4096, sub-page size 4096
       [  289.227293] UBI: VID header offset: 4096 (aligned 4096), data offset: 8192
       [  289.232878] UBI: good PEBs: 1856, bad PEBs: 0, corrupted PEBs: 0
       [  289.237628] UBI: user volume: 0, internal volumes: 1, max. volumes count: 128
       [  289.243553] UBI: max/mean erase counter: 1/1, WL threshold: 4096, image sequence number: 2024916145
       [  289.251348] UBI: available PEBs: 1812, total reserved PEBs: 44, PEBs reserved for bad PEB handling: 40
       [  289.259417] UBI: background thread "ubi_bgt0d" started, PID 847
       --------------------------------------------------------------
       The attach time is about 289.18 - 286.46 = 2.7s

     4.4) The conclusion:
       We achieve (4.4 - 2.7) / 4.4 = 38.6% faster in the ubiattach.

Signed-off-by: Huang Shijie <b32955@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: gpmi: do not use the mtd->writesize

The nfc_geo->payload_size is equal to the mtd->writesize now,
use the nfc_geo->payload_size to replace the mtd->writesize.

This patch makes preparation for the gpmi's subpage read support.
In the subpage support, the nfc_geo->payload_size maybe smaller then
the mtd->writesize.

Signed-off-by: Huang Shijie <b32955@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: add "page" argument for read_subpage hook

Add the "page" argument for the read_subpage hook. With this argument,
the implementation of this hook could prints out more accurate information
for debugging.

Signed-off-by: Huang Shijie <b32955@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: remove unused function input parameter

The nand_get_flash_type parameter "busw" input value is not used by any
branch, and it is updated before use it in the function, so remove it,
define the "busw" as an internal variable.

Signed-off-by: Cai Zhiyong <caizhiyong@huawei.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: pxa3xx: Print actual ECC strength in error message

The actual ECC strength used to select the ECC scheme is 'ecc_strength'.
Use it in the error message.

Signed-off-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: pxa3xx: Remove unused macro

This macro is not used so it's safe to remove it.

Signed-off-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: bcm47xxpart: allow enabling on ARCH_BCM_5301X

Home routers based on SoCs like BCM53010 (AKA BCM4708) use flashes
which can be nicely partitioned with bcm47xxpart. Header bcm47xx_nvram.h
is not available on bcm53xx, so don't include it.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: bcm47xxpart: avoid overflowing when registering trx

Our code parsing "trx" header registers few partitions at once (in one
loop iteration). Add extra check in that place.

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: bcm47xxpart: fix off by one in partitions limit

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: allow CONFIG_MTD_PHYSMAP_OF also for CONFIG_MTD_RAM

Up to now mtd-ram devices described in device trees were only accessible
if mtd-flash or mtd-rom were also configured at linux configuration
time, because MTD_PHYSMAP_OF was only available if (MTD_CFI ||
MTD_JEDECPROBE || MTD_ROM).  Allow MTD_PHYSMAP_OF selection also
when only MTD_RAM is set.

Signed-off-by: Philippe De Muyter <phdm@macqel.be>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: m25p80: add Macronix mx66l1g55g 1Gbit SPI flash

Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Acked-by: Marek Vasut <marex@denx.de>

mtd: nand: Add a devicetree binding for ECC strength and ECC step size

Some flashes can only be properly accessed when the ECC mode is
specified, so a way to describe such mode is required.

Together, the ECC strength and step size define the correction capability,
so that we say we will correct "{strength} bit errors per {size} bytes".

The interpretation of these parameters is implementation-defined, but they
often have ramifications on the formation, interpretation, and placement of
correction metadata on the flash. Not all implementations must support all
possible combinations. Implementations are encouraged to further define the
value(s) they support.

Acked-by: Boris BREZILLON <b.brezillon.dev@gmail.com>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

of_mtd: Add helpers to get ECC strength and ECC step size

This commit adds simple helpers to obtain the devicetree properties
that specify the ECC strength and ECC step size to use on a given
NAND controller.

Acked-by: Boris BREZILLON <b.brezillon.dev@gmail.com>
Signed-off-by: Ezequiel Garcia <ezequiel.garcia@free-electrons.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: Add a retlen parameter to _get_{fact,user}_prot_info

Signed-off-by: Christian Riesch <christian.riesch@omicron.at>
Cc: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

jffs2: Fix crash due to truncation of csize

mounting JFFS2 partition sometimes crashes with this call trace:

[ 1322.240000] Kernel bug detected[#1]:
[ 1322.244000] Cpu 2
[ 1322.244000] $ 0   : 0000000000000000 0000000000000018 000000003ff00070 0000000000000001
[ 1322.252000] $ 4   : 0000000000000000 c0000000f3980150 0000000000000000 0000000000010000
[ 1322.260000] $ 8   : ffffffffc09cd5f8 0000000000000001 0000000000000088 c0000000ed300de8
[ 1322.268000] $12   : e5e19d9c5f613a45 ffffffffc046d464 0000000000000000 66227ba5ea67b74e
[ 1322.276000] $16   : c0000000f1769c00 c0000000ed1e0200 c0000000f3980150 0000000000000000
[ 1322.284000] $20   : c0000000f3a80000 00000000fffffffc c0000000ed2cfbd8 c0000000f39818f0
[ 1322.292000] $24   : 0000000000000004 0000000000000000
[ 1322.300000] $28   : c0000000ed2c0000 c0000000ed2cfab8 0000000000010000 ffffffffc039c0b0
[ 1322.308000] Hi    : 000000000000023c
[ 1322.312000] Lo    : 000000000003f802
[ 1322.316000] epc   : ffffffffc039a9f8 check_tn_node+0x88/0x3b0
[ 1322.320000]     Not tainted
[ 1322.324000] ra    : ffffffffc039c0b0 jffs2_do_read_inode_internal+0x1250/0x1e48
[ 1322.332000] Status: 5400f8e3    KX SX UX KERNEL EXL IE
[ 1322.336000] Cause : 00800034
[ 1322.340000] PrId  : 000c1004 (Netlogic XLP)
[ 1322.344000] Modules linked in:
[ 1322.348000] Process jffs2_gcd_mtd7 (pid: 264, threadinfo=c0000000ed2c0000, task=c0000000f0e68dd8, tls=0000000000000000)
[ 1322.356000] Stack : c0000000f1769e30 c0000000ed010780 c0000000ed010780 c0000000ed300000
        c0000000f1769c00 c0000000f3980150 c0000000f3a80000 00000000fffffffc
        c0000000ed2cfbd8 ffffffffc039c0b0 ffffffffc09c6340 0000000000001000
        0000000000000dec ffffffffc016c9d8 c0000000f39805a0 c0000000f3980180
        0000008600000000 0000000000000000 0000000000000000 0000000000000000
        0001000000000dec c0000000f1769d98 c0000000ed2cfb18 0000000000010000
        0000000000010000 0000000000000044 c0000000f3a80000 c0000000f1769c00
        c0000000f3d207a8 c0000000f1769d98 c0000000f1769de0 ffffffffc076f9c0
        0000000000000009 0000000000000000 0000000000000000 ffffffffc039cf90
        0000000000000017 ffffffffc013fbdc 0000000000000001 000000010003e61c
        ...
[ 1322.424000] Call Trace:
[ 1322.428000] [<ffffffffc039a9f8>] check_tn_node+0x88/0x3b0
[ 1322.432000] [<ffffffffc039c0b0>] jffs2_do_read_inode_internal+0x1250/0x1e48
[ 1322.440000] [<ffffffffc039cf90>] jffs2_do_crccheck_inode+0x70/0xd0
[ 1322.448000] [<ffffffffc03a1b80>] jffs2_garbage_collect_pass+0x160/0x870
[ 1322.452000] [<ffffffffc03a392c>] jffs2_garbage_collect_thread+0xdc/0x1f0
[ 1322.460000] [<ffffffffc01541c8>] kthread+0xb8/0xc0
[ 1322.464000] [<ffffffffc0106d18>] kernel_thread_helper+0x10/0x18
[ 1322.472000]
[ 1322.472000]
Code: 67bd0050  94a4002c  2c830001 <00038036> de050218  2403fffc  0080a82d  00431824  24630044
[ 1322.480000] ---[ end trace b052bb90e97dfbf5 ]---

The variable csize in structure jffs2_tmp_dnode_info is of type uint16_t, but it
is used to hold the compressed data length(csize) which is declared as uint32_t.
So, when the value of csize exceeds 16bits, it gets truncated when assigned to
tn->csize. This is causing a kernel BUG.
Changing the definition of csize in jffs2_tmp_dnode_info to uint32_t fixes the issue.

Signed-off-by: Ajesh Kunhipurayil Vijayan <ajesh@broadcom.com>
Signed-off-by: Kamlakant Patel <kamlakant.patel@broadcom.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

jffs2: Fix segmentation fault found in stress test

Creating a large file on a JFFS2 partition sometimes crashes with this call
trace:

[  306.476000] CPU 13 Unable to handle kernel paging request at virtual address c0000000dfff8002, epc == ffffffffc03a80a8, ra == ffffffffc03a8044
[  306.488000] Oops[#1]:
[  306.488000] Cpu 13
[  306.492000] $ 0   : 0000000000000000 0000000000000000 0000000000008008 0000000000008007
[  306.500000] $ 4   : c0000000dfff8002 000000000000009f c0000000e0007cde c0000000ee95fa58
[  306.508000] $ 8   : 0000000000000001 0000000000008008 0000000000010000 ffffffffffff8002
[  306.516000] $12   : 0000000000007fa9 000000000000ff0e 000000000000ff0f 80e55930aebb92bb
[  306.524000] $16   : c0000000e0000000 c0000000ee95fa5c c0000000efc80000 ffffffffc09edd70
[  306.532000] $20   : ffffffffc2b60000 c0000000ee95fa58 0000000000000000 c0000000efc80000
[  306.540000] $24   : 0000000000000000 0000000000000004
[  306.548000] $28   : c0000000ee950000 c0000000ee95f738 0000000000000000 ffffffffc03a8044
[  306.556000] Hi    : 00000000000574a5
[  306.560000] Lo    : 6193b7a7e903d8c9
[  306.564000] epc   : ffffffffc03a80a8 jffs2_rtime_compress+0x98/0x198
[  306.568000]     Tainted: G        W
[  306.572000] ra    : ffffffffc03a8044 jffs2_rtime_compress+0x34/0x198
[  306.580000] Status: 5000f8e3    KX SX UX KERNEL EXL IE
[  306.584000] Cause : 00800008
[  306.588000] BadVA : c0000000dfff8002
[  306.592000] PrId  : 000c1100 (Netlogic XLP)
[  306.596000] Modules linked in:
[  306.596000] Process dd (pid: 170, threadinfo=c0000000ee950000, task=c0000000ee6e0858, tls=0000000000c47490)
[  306.608000] Stack : 7c547f377ddc7ee4 7ffc7f967f5d7fae 7f617f507fc37ff4 7e7d7f817f487f5f
        7d8e7fec7ee87eb3 7e977ff27eec7f9e 7d677ec67f917f67 7f3d7e457f017ed7
        7fd37f517f867eb2 7fed7fd17ca57e1d 7e5f7fe87f257f77 7fd77f0d7ede7fdb
        7fba7fef7e197f99 7fde7fe07ee37eb5 7f5c7f8c7fc67f65 7f457fb87f847e93
        7f737f3e7d137cd9 7f8e7e9c7fc47d25 7dbb7fac7fb67e52 7ff17f627da97f64
        7f6b7df77ffa7ec5 80057ef17f357fb3 7f767fa27dfc7fd5 7fe37e8e7fd07e53
        7e227fcf7efb7fa1 7f547e787fa87fcc 7fcb7fc57f5a7ffb 7fc07f6c7ea97e80
        7e2d7ed17e587ee0 7fb17f9d7feb7f31 7f607e797e887faa 7f757fdd7c607ff3
        7e877e657ef37fbd 7ec17fd67fe67ff7 7ff67f797ff87dc4 7eef7f3a7c337fa6
        7fe57fc97ed87f4b 7ebe7f097f0b8003 7fe97e2a7d997cba 7f587f987f3c7fa9
        ...
[  306.676000] Call Trace:
[  306.680000] [<ffffffffc03a80a8>] jffs2_rtime_compress+0x98/0x198
[  306.684000] [<ffffffffc0394f10>] jffs2_selected_compress+0x110/0x230
[  306.692000] [<ffffffffc039508c>] jffs2_compress+0x5c/0x388
[  306.696000] [<ffffffffc039dc58>] jffs2_write_inode_range+0xd8/0x388
[  306.704000] [<ffffffffc03971bc>] jffs2_write_end+0x16c/0x2d0
[  306.708000] [<ffffffffc01d3d90>] generic_file_buffered_write+0xf8/0x2b8
[  306.716000] [<ffffffffc01d4e7c>] __generic_file_aio_write+0x1ac/0x350
[  306.720000] [<ffffffffc01d50a0>] generic_file_aio_write+0x80/0x168
[  306.728000] [<ffffffffc021f7dc>] do_sync_write+0x94/0xf8
[  306.732000] [<ffffffffc021ff6c>] vfs_write+0xa4/0x1a0
[  306.736000] [<ffffffffc02202e8>] SyS_write+0x50/0x90
[  306.744000] [<ffffffffc0116cc0>] handle_sys+0x180/0x1a0
[  306.748000]
[  306.748000]
Code: 020b202d  0205282d  90a50000 <90840000> 14a40038  00000000  0060602d  0000282d  016c5823
[  306.760000] ---[ end trace 79dd088435be02d0 ]---
Segmentation fault

This crash is caused because the 'positions' is declared as an array of signed
short. The value of position is in the range 0..65535, and will be converted
to a negative number when the position is greater than 32767 and causes a
corruption and crash. Changing the definition to 'unsigned short' fixes this
issue

Signed-off-by: Jayachandran C <jchandra@broadcom.com>
Signed-off-by: Kamlakant Patel <kamlakant.patel@broadcom.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

jffs2: unlock f->sem on error in jffs2_new_inode()

If jffs2_new_inode() succeeds, it returns with f->sem held, and the caller
is responsible for releasing the lock.  If it fails, it still returns with
the lock held, but the caller won't release the lock, which will lead to
deadlock.

Fix it by releasing the lock in jffs2_new_inode() on error.

Signed-off-by: Wang Guoli <andy.wangguoli@huawei.com>
Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Wang Guoli <andy.wangguoli@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[Brian: not marked for stable; no one observed deadlock, and I don't
        think it can happen here]
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

jffs2: avoid soft-lockup in jffs2_reserve_space_gc()

We triggered soft-lockup under stress test on 2.6.34 kernel.

BUG: soft lockup - CPU#1 stuck for 60009ms! [lockf2.test:14488]
...
[<bf09a4d4>] (jffs2_do_reserve_space+0x420/0x440 [jffs2])
[<bf09a528>] (jffs2_reserve_space_gc+0x34/0x78 [jffs2])
[<bf0a1350>] (jffs2_garbage_collect_dnode.isra.3+0x264/0x478 [jffs2])
[<bf0a2078>] (jffs2_garbage_collect_pass+0x9c0/0xe4c [jffs2])
[<bf09a670>] (jffs2_reserve_space+0x104/0x2a8 [jffs2])
[<bf09dc48>] (jffs2_write_inode_range+0x5c/0x4d4 [jffs2])
[<bf097d8c>] (jffs2_write_end+0x198/0x2c0 [jffs2])
[<c00e00a4>] (generic_file_buffered_write+0x158/0x200)
[<c00e14f4>] (__generic_file_aio_write+0x3a4/0x414)
[<c00e15c0>] (generic_file_aio_write+0x5c/0xbc)
[<c012334c>] (do_sync_write+0x98/0xd4)
[<c0123a84>] (vfs_write+0xa8/0x150)
[<c0123d74>] (sys_write+0x3c/0xc0)]

Fix this by adding a cond_resched() in the while loop.

[akpm@linux-foundation.org: don't initialize `ret']
Signed-off-by: Li Zefan <lizefan@huawei.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

jffs2: remove from wait queue after schedule()

@wait is a local variable, so if we don't remove it from the wait queue
list, later wake_up() may end up accessing invalid memory.

This was spotted by eyes.

Signed-off-by: Li Zefan <lizefan@huawei.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: ts5500: Add dependency

There is no point in displaying the TS5500-specific driver entries if
TS5500 board support itself isn't enabled.

Signed-off-by: Jean Delvare <jdelvare@suse.de>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: Vivien Didelot <vivien.didelot@savoirfairelinux.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nuc900_nand: NULL dereference in nuc900_nand_enable()

Instead of writing to "nand->reg + REG_FMICSR" we write to "REG_FMICSR"
which is NULL and not a valid register.

Fixes: 8bff82cbc308 ('mtd: add nand support for w90p910 (v2)')
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: print out the right information for JEDEC compliant NAND

Check the chip->jedec_version, and print out the right information
for JEDEC compliant NAND.

Signed-off-by: Huang Shijie <b32955@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: parse out the JEDEC compliant NAND

This patch adds the parsing code for the JEDEC compliant NAND.

Since we need the 0x40 as the column address, this patch also
makes the NAND_CMD_PARAM to use the 8-bit address only.

Signed-off-by: Huang Shijie <b32955@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: add a helper to get the supported features for JEDEC

Add a helper to get the supported features for JEDEC compliant NAND.
Also add a macro JEDEC_FEATURE_16_BIT_BUS.

Signed-off-by: Huang Shijie <b32955@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: add fields for JEDEC in nand_chip

Add the jedec_version field, and add an anonymous union which
contains the nand_onfi_params and nand_jedec_params.

Signed-off-by: Huang Shijie <b32955@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: add the data structures for JEDEC parameter page

Create the nand_jedec_params{} and jedec_ecc_info{} according to
the JESD230A (Revision of JESD230, October 2012).

Signed-off-by: Huang Shijie <b32955@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: nand: s3c2410: Trivial cleanup in header file

Commit 436d42c61c3e ("ARM: samsung: move platform_data definitions")
moved the files to the current location but forgot to remove the pointer
to its previous location. Clean it up. While at it also add the header
file protection macros appropriately.

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: mxc_nand: Propagate the error if platform_get_irq() fails

Check the return value from platform_get_irq() and propagate it in the case of
error.

Signed-off-by: Fabio Estevam <fabio.estevam@freescale.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: remove some duplicative checks

"rc" is an error code here, no need to check it a second time.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: atmel_nand: change log level

PIO fall back is not an issue, so don't make this much noise.

Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: m25p80: add support for the Spansion s25fl008k chip

Signed-off-by: Yusuke Goda <yusuke.goda.sx@renesas.com>
Signed-off-by: Kuninori Morimoto <kuninori.morimoto.gx@renesas.com>
Signed-off-by: Geert Uytterhoeven <geert+renesas@linux-m68k.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: gen_probe: Remove unnecessary OOM messages

The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: cfi: Remove unnecessary OOM messages

The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: onenand: Remove unnecessary OOM messages

The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: lpddr: Remove unnecessary OOM messages

The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: plat-ram: Remove unnecessary OOM messages

The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: pmc551: Remove unnecessary OOM messages

The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: spear_smi: Remove unnecessary OOM messages

The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: devices: elm: Remove unnecessary OOM messages

The site-specific OOM messages are unnecessary, because they
duplicate the MM subsystem generic OOM message.

Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Acked-by: Pekon Gupta <pekon@ti.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>

mtd: elm: Use correct check on return value of pm_runtime_get_sync

The ELM driver incorrectly reagard any non-zero return value from
pm_runtime_get_sync as an error, but it may return 1 if the device
was already active. Fix to only error when return value is negative.

Signed-off-by: Stefan Sørensen <stefan.sorensen@spectralink.com>
Signed-off-by: Brian Norris <computersforpeace@gmail.com>