mtd: introduce mtd_write_oob interface

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: introduce mtd_read_oob interface

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: introduce mtd_panic_write interface

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: introduce mtd_write interface

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: introduce mtd_read interface

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: introduce mtd_get_unmapped_area interface

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: introduce mtd_unpoint interface

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: introduce mtd_point interface

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: introduce mtd_erase interface

This patch is part of a patch-set which changes the MTD interface
from 'mtd->func()' form to 'mtd_func()' form. We need this because
we want to add common code to to all drivers in the mtd core level,
which is impossible with the current interface when MTD clients
call driver functions like 'read()' or 'write()' directly.

At this point we just introduce a new inline wrapper function, but
later some of them are expected to gain more code. E.g., the input
parameters check should be moved to the wrappers rather than be
duplicated at many drivers.

This particular patch introduced the 'mtd_erase()' interface. The
following patches add all the other interfaces one by one.

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: mtdchar: rename functions

We are going to re-work the MTD interface and change 'mtd->write()' to
'mtd_write()', 'mtd->read()' to 'mtd_read()' and so forth for all functions
in the 'struct mtd_info' structure.

However, mtdchar.c has its own 'mtd_read()', 'mtd_write()', etc functions
which collide with our changes. This patch renames these functions
to 'mtdchar_read()', 'mtdchar_write()', etc.

Additionally, to make the 'mtdchar.c' file look consistent, rename
similarly all the other functions starting with 'mtd_'.

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

logfs: rename functions starting with mtd_

We are going to re-work the MTD interface and change 'mtd->write()' to
'mtd_write()', 'mtd->read()' to 'mtd_read()' and so forth for all functions
in the 'struct mtd_info' structure.

However, logfs has its own 'mtd_read()', 'mtd_write()', etc functions
which collide with our changes. This patch renames these logfs functions
to 'logfs_mtd_read()', 'logfs_mtd_write()', etc.

Additionally, to make the 'fs/logfs/dev_mtd.c' file look consistent, rename
similarly all the other functions starting with 'mtd_'.

Cc: Jörn Engel <joern@logfs.org>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: lpddr: drop unnecessary zeroing

We allocate the "mtd" structure using kzalloc which means we do not have
to initialize unused MTD function pointers to NULL, since it is safe to
assume in Linux that NULL contains all zeroes.

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: map.h: fix arm cross-build failure

This patch fixes the following build failure:
In file included from include/linux/mtd/qinfo.h:4:0,
                 from include/linux/mtd/pfow.h:7,
                 from drivers/mtd/lpddr/lpddr_cmds.c:27:
include/linux/mtd/map.h: In function 'inline_map_read':
include/linux/mtd/map.h:409:3: error: implicit declaration of function 'BUILD_BUG_ON' [-Werror=implicit-function-declaration]

Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: txx9ndfmc: use devm_request_and_ioremap

Reimplement a call to devm_request_mem_region followed by a call to ioremap
or ioremap_nocache by a call to devm_request_and_ioremap.

The semantic patch that makes this transformation is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@nm@
expression myname;
identifier i;
@@

struct platform_driver i = { .driver = { .name = myname } };

@@
expression dev,res,size;
expression nm.myname;
@@

-if (!devm_request_mem_region(dev, res->start, size,
-                              \(res->name\|dev_name(dev)\|myname\))) {
-   ...
-   return ...;
-}
... when != res->start
(
-devm_ioremap(dev,res->start,size)
+devm_request_and_ioremap(dev,res)
|
-devm_ioremap_nocache(dev,res->start,size)
+devm_request_and_ioremap(dev,res)
)
... when any
    when != res->start
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: lantiq-flash: drop iounmap for devm_ allocated data

Data allocated with devm_ioremap or devm_ioremap_nocache should not be
freed using iounmap, because doing so causes a dangling pointer, and a
subsequent double free.

The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@r@
expression x;
@@
(
 x = devm_ioremap(...)
|
 x = devm_ioremap_nocache(...)
)

@@
expression r.x;
@@
* iounmap(x)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: gpio-nand: add device tree bindings

Add device tree bindings so that the gpio-nand driver may be
instantiated from the device tree.  This also allows the partitions
to be specified in the device tree.

v7: - restore runtime device tree/non device tree detection
v6: - convert to mtd_device_parse_register()
v5: - fold dt config helpers into a single gpio_nand_of_get_config()
v4: - get io sync address from gpio-control-nand,io-sync-reg
  property rather than a resource
- clarified a few details in the binding
v3: - remove redundant cast and a couple of whitespace/naming
  changes
v2: - add CONFIG_OF guards for non-dt platforms
- compatible becomes gpio-control-nand
- clarify some binding details

Signed-off-by: Jamie Iles <jamie@jamieiles.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: bcm63xxpart: check the image tag's crc32

Only use the values from the image tag if it is valid. Always create
the CFE, NVRAM and linux partitions, to allow flashing a new image even
if the old is invalid without overwriting CFE or NVRAM.

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

MIPS: BCM63XX: bcm963xx_tag.h: make crc fields integers

All CRC32 fields are 32 bit integers, so define them as such to prevent
unnecessary casts if we want to use them.

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: bcm63xxpart: don't assume NVRAM is always the fourth partition

Instead of referencing the sizes of fixed partitions, use the
precomputed CFE/NVRAM lengths.

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: bcm63xxpart: make sure CFE and NVRAM partitions are at least 64KiB

The CFE boot loader on BCM63XX platforms assumes itself and the NVRAM
partition to be 64 KiB (or erase block sized, if larger).
Ensure this assumption is also met when creating the partitions to
prevent accidential erasure of CFE or NVRAM.

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: bcm63xxpart: check version marker string for newer CFEs

Recent CFEs do not contain the CFE1CFE1 magic anymore, so check for the
"cfe-v" version marker string instead. As very old CFEs do not have
this string, leave the CFE1CFE1 magic as a fallback for detection.

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: sm_ftl: fix module parameter

You didn't mean this to be a bool.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: remove unused function

As the MTD api has no use for the number of erase cycles
each block has endured, remove the function which calculated
that value.

If one day MTD api finds it usefull for wear levelling
algorithms to have this information, the function should be
put back in place.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: nand: Add ONFI support for FSL NAND controller

- fix NAND_CMD_READID command for ONFI detect.
- add NAND_CMD_PARAM command to read the ONFI parameter page.

Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: nand: fixup for fmr initialization of Freescale NAND controller

There was a bug for fmr initialization, which lead to  fmr was always 0x100
in fsl_elbc_chip_init() and caused FCM command timeout before calling
fsl_elbc_chip_init_tail(), now we initialize CWTO to maximum timeout value
and not relying on the setting of bootloader.

Signed-off-by: Shengzhou Liu <Shengzhou.Liu@freescale.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: document that MEMWRITE ioctl is NAND-specific

Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: fsl_elbc_nand: set Nand flash page address to FBAR and FPAR correctly

If we use the Nand flash chip whose number of pages in a block is greater
than 64(for large page), we must treat the low bit of FBAR as being the
high bit of the page address due to the limitation of FCM, it simply uses
the low 6-bits (for large page) of the combined block/page address as the
FPAR component, rather than considering the actual block size.

Signed-off-by: Liu Shuo <b35362@freescale.com>
Signed-off-by: Jerry Huang <Chang-Ming.Huang@freescale.com>
Signed-off-by: Tang Yuantian <b29983@freescale.com>
Signed-off-by: Li Yang <leoli@freescale.com>
Acked-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtdoops: fix the oops_page_used array size

The array of unsigned long pointed by oops_page_used is allocated
by vmalloc which requires the size to be in bytes.

BITS_PER_LONG is equal to 32.
If we want to allocate memory for 32 pages with one bit per page then
32 / BITS_PER_LONG  is equal to 1 byte that is 8 bits.
To fix it we need to multiply the result by sizeof(unsigned long) equal to 4.

Cc: stable@kernel.org
Signed-off-by: Roman Tereshonkov <roman.tereshonkov@nokia.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

jffs2: fix up error handling for insert_inode_locked

after 250df6ed274d767da844a5d9f05720b804240197
(fs: protect inode->i_state with inode->i_lock), insert_inode_locked()
no longer returns the inode with I_NEW set on failure.  However,
the error handler still calls unlock_new_inode() on failure,
which does a WARN_ON if I_NEW is not set, so any failure spews
a lot of warnings.

We can just drop the unlock_new_inode() if insert_inode_locked()
fails here.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: maps: remove the now unused bcm963xx-flash

bcm963xx-flash does nothing meaningful anymore.

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

MIPS: BCM63XX: use the new bcm63xxpart parser

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: maps: physmap: allow partition parsers for physmap_flash_data

Arch setup code might want to use their own partition parsers, but still
use the generic physmap flash driver.

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: maps: bcm963xx-flash: make CFE partition parsing an mtd parser

Recent BCM63XX devices support a variety of flash types (parallel, SPI,
NAND) and share the partition layout. To prevent code duplication make
the CFE partition parsing code a stand alone mtd parser to allow SPI or
NAND flash drivers to use it.

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: maps: bcm963xx-flash: clean up printk usage

Replace raw printk's with their pr_XXX equivalent and unify broken up
strings so they become grepable.

Also replace the PFX definition with a pr_fmt().

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: maps: bcm963xx-flash: remove superfluous semicolons

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: maps: bcm963xx-flash: fix word order for spare partition

Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: davinci: if no amif timing is passed, don; t setup cscfg register

Signed-off-by: Heiko Schocher <hs@denx.de>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: nand: set correct length to FBCR for a non-full-page write

When we do a non-full-page write, the length be set to FBCR should
not be 'elbc_fcm_ctrl->index', it should be 'elbc_fcm_ctrl->index -
elbc_fcm_ctrl->column'.

Signed-off-by: Liu Shuo <b35362@freescale.com>
Signed-off-by: Li Yang <leoli@freescale.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: nand: use elbc_fcm_ctrl->oob to set FPAR_MS bit of FPAR

On both of large-page chip and small-page chip, we always should use
'elbc_fcm_ctrl->oob' to set the FPAR_LP_MS/FPAR_SP_MS bit of FPAR, don't
use a overflowed 'column' to set it.

Signed-off-by: Liu Shuo <b35362@freescale.com>
Signed-off-by: Li Yang <leoli@freescale.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: fixes and cleanups

This patch takes into account checkpatch, sparse and ECC
comments.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: mtdoops: skip reading initially bad blocks

Use block_isbad to check and skip the bad blocks reading.
This will allow to get rid of the read errors if bad blocks
are present initially.

Cc: stable@kernel.org
Signed-off-by: Roman Tereshonkov <roman.tereshonkov@nokia.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: tests: stresstest: bail out if device has not enough eraseblocks

stresstest needs at least two eraseblocks. Bail out gracefully if that
condition is not met. Fixes the following 'division by zero' OOPS:

[  619.100000] mtd_stresstest: MTD device size 131072, eraseblock size 131072, page size 2048, count of eraseblocks 1, pages per eraseblock 64, OOB size 64
[  619.120000] mtd_stresstest: scanning for bad eraseblocks
[  619.120000] mtd_stresstest: scanned 1 eraseblocks, 0 are bad
[  619.130000] mtd_stresstest: doing operations
[  619.130000] mtd_stresstest: 0 operations done
[  619.140000] Division by zero in kernel.
...

caused by

        /* Read or write up 2 eraseblocks at a time - hence 'ebcnt - 1' */
        eb %= (ebcnt - 1);

Cc: stable@kernel.org
Signed-off-by: Wolfram Sang <w.sang@pengutronix.de>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: convert drivers/mtd/* to use module_platform_driver()

This patch converts the drivers in drivers/mtd/* to use the
module_platform_driver() macro which makes the code smaller and a bit
simpler.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked by: Haojian Zhuang <haojian.zhuang@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

Documentation: add sysfs entries for mtd docg3 chips

Add documentation for MSystems disk-on-chip docg3 chips
sysfs entries, which enable and disable protection areas,
giving or disabling access to the chip's memory.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: dereferencing an ERR_PTR() in docg3_probe()

If doc_probe_device() returned an ERR_PTR, then we accidentally saved
that to docg3_floors[floor] = mtd; which gets derefenced in the error
handling when we call doc_release_device().

I've reworked the error handling to take care of that and hopefully
make it a little simpler.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: Remove redundant spi driver bus initialization

In ancient times it was necessary to manually initialize the bus field of an
spi_driver to spi_bus_type. These days this is done in spi_driver_register(),
so we can drop the manual assignment.

The patch was generated using the following coccinelle semantic patch:
// <smpl>
@@
identifier _driver;
@@
struct spi_driver _driver = {
.driver = {
- .bus = &spi_bus_type,
},
};
// </smpl>

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add protection areas sysfs access

As each docg3 chip has 2 protection areas (DPS0 and DPS1),
and because theses areas can prevent user access to the chip
data, add for each floor the sysfs entries which insert the
protection key into the right DPS.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add fast mode

Docg3 chips can work in 3 modes : normal MLC mode, fast
mode and reliable mode. Normally, as docg3 is a MLC chip, it
should be configured to work in normal mode.

In both normal mode, each page is distinct. This
means that writing to page 12 of blocks 14,15 writes only to
that page, and reading from page 12 of blocks 14,15 reads
only from that page.

In reliable and fast modes, pages are coupled by pairs, and
are clones one of each other. This means that the available
capacity of the chip is halved. Pages are coupled in each
block, and page of index 2*n contains the same data as page
2*n+1 of the same block.

In fast mode, the reads occur a bit faster, but are a bit
less reliable that in normal mode.

When reading from page 2*n, the chip reads bytes from both
page 2*n and page 2*n+1, makes a logical and for each byte,
and returns the result. As programming a page means
"clearing bits", even if a bit was not cleared on one page
because the flash is worn out, the other page has the bit
cleared, and the result of the "AND" gives a correct result.

When writing to page 2*n, the chip writes data to both page
2*n and page 2*n+1.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add suspend and resume

Add functions to powerdown and powerup from suspend, in
order to save power.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add ECC correction code

Credit for discovering the BCH algorith parameters, and bit
reversing algorithm is to be give to Mike Dunn and Ivan
Djelic.

The BCH correction code relied upon the BCH library, where
all data and ECC is bit-reversed. The BCH library works
correctly when each input byte is bit-reversed, and
accordingly ECC output is also bit-reversed.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: map erase and write functions

Map the developped write and erase functions into the mtd
structure.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add erase functions

Add erase capability to the docg3 driver. The erase block is
made of 2 physical blocks, as both share all 64 pages. That
makes an erase block of at least 64 kBytes.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add write functions

Add write capability to the docg3 driver. The writes are
possible on a single page (512 bytes + 16 bytes), even if
that page is split on 2 physical pages on 2 blocks (each on
one plane).

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add OOB buffer to device structure

Add OOB buffer area to store the OOB data until the actual
page is written, so that it can be completed by hardware ECC
generator.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add registers for erasing and writing

Add the required registers and commands to erase and write
flash pages / blocks.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add OOB layout to mtdinfo

Add OOB layout description for docg3, so that userspace can
use this information to setup the data for write_oob().

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: add multiple floor support

Add support for multiple floors, ie. cascaded docg3
chips. There might be 4 docg3 chips cascaded, sharing the
same address space, and providing up to 4 times the storage
capacity of a unique chip.

Each floor will be seen as an independant mtd device.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: fix reading oob+data without correction

Fix the docg3 reads to be able to cope with all possible
data buffer / oob buffer / file mode combinations from
docg3_read_oob().
This especially ensures that raw reads do not use ECC
corrections, and AUTOOOB and PLACEOOB do use ECC
correction.

The approach is to empty docg3_read() and make it a wrapper
to docg3_read_oob(). As docg3_read_oob() handles all the
funny cases (no data buffer but oob buffer, data buffer but
no oob buffer, ...), docg3_read() is just a special use of
docg3_read_oob().

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: fix BCH registers

BCH registers are contiguous, not on every byte. Fix the
register definitions.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: fix protection areas reading

The protection areas boundaries were on 16bit registers, not
8bit. This is consistent with block numbers, which can
extend up to 4096 on bigger chips (and is 2048 on the
docg3).

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: fix tracing of IO in writeb

Writeb was incorrectly traced as a 16 bits write, instead of
a 8 bits write. Fix it by tracing the correct width.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: docg3: fix debug log verbosity

Change the NOP debug log verbosity to very verbose to
unburden log analysis.

Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Reviewed-by: Ivan Djelic <ivan.djelic@parrot.com>
Reviewed-by: Mike Dunn <mikedunn@newsguy.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: nand: Making MTD_NAND_OMAP2 depend on ARCH_OMAP2PLUS

Making  MTD_NAND_OMAP2 depend on ARCH_OMAP2PLUS instead of
oring with ARCH2/3/4.

Reported-by: Russell King <rmk+kernel@arm.linux.org.uk>
Signed-off-by: Shubhrajyoti D <shubhrajyoti@ti.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: cfi: Allow per-mapping CFI device endianness

This patch allows each CFI device map to use its own endianness. The
globally defined CFI endianness (CONFIG_MTD_CFI_NOSWAP,
CONFIG_MTD_CFI_BE_BYTE_SWAP or CONFIG_MTD_CFI_LE_BYTE_SWAP) becomes the
default value which can be overridden by a driver for a particular device.

Signed-off-by: Aaron Sierra <asierra@xes-inc.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: mtd_blkdevs: don't increase 'open' count on error path

Some error paths in mtd_blkdevs were fixed in the following commit:

    commit 94735ec4044a6d318b83ad3c5794e931ed168d10
    mtd: mtd_blkdevs: fix error path in blktrans_open

But on these error paths, the block device's `dev->open' count is
already incremented before we check for errors. This meant that, while
the error path was handled correctly on the first time through
blktrans_open(), the device is erroneously considered already open on
the second time through.

This problem can be seen, for instance, when a UBI volume is
simultaneously mounted as a UBIFS partition and read through its
corresponding gluebi mtdblockX device. This results in blktrans_open()
passing its error checks (with `dev->open > 0') without actually having
a handle on the device. Here's a summarized log of the actions and
results with nandsim:

    # modprobe nandsim
    # modprobe mtdblock
    # modprobe gluebi
    # modprobe ubifs
    # ubiattach /dev/ubi_ctrl -m 0
    ...
    # ubimkvol /dev/ubi0 -N test -s 16MiB
    ...
    # mount -t ubifs ubi0:test /mnt
    # ls /dev/mtdblock*
    /dev/mtdblock0  /dev/mtdblock1
    # cat /dev/mtdblock1 > /dev/null
    cat: can't open '/dev/mtdblock4': Device or resource busy
    # cat /dev/mtdblock1 > /dev/null

    CPU 0 Unable to handle kernel paging request at virtual address
    fffffff0, epc == 8031536c, ra == 8031f280
    Oops[#1]:
    ...
    Call Trace:
    [<8031536c>] ubi_leb_read+0x14/0x164
    [<8031f280>] gluebi_read+0xf0/0x148
    [<802edba8>] mtdblock_readsect+0x64/0x198
    [<802ecfe4>] mtd_blktrans_thread+0x330/0x3f4
    [<8005be98>] kthread+0x88/0x90
    [<8000bc04>] kernel_thread_helper+0x10/0x18

Cc: stable@kernel.org [3.0+]
Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: nand: scan 1st and 2nd page for Macronix SLC

Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: nand: add 512 Mbit device code (Macronix)

Macronix MX30LF1208AA is a 512 Mbit NAND with device code 0xF0.

Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

mtd: nand: add Macronix manufacturer

Macronix is produing SLC NAND MX30LF1208AA, so add their manufacturer
code to the manufacturer lists.

Signed-off-by: Brian Norris <computersforpeace@gmail.com>
Signed-off-by: Artem Bityutskiy <Artem.Bityutskiy@linux.intel.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>

Linux 3.2-rc7

Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
  VFS: Fix race between CPU hotplug and lglocks

Merge tag 'writeback' of git://git./linux/kernel/git/wfg/linux

for linus: writeback reason binary tracing format fix

* tag 'writeback' of git://git.kernel.org/pub/scm/linux/kernel/git/wfg/linux:
  writeback: show writeback reason with __print_symbolic

Merge branch 'rc-fixes' of git://git./linux/kernel/git/mmarek/kbuild

* 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:
  kconfig: adapt update-po-config to new UML layout

Merge branch 'v4l_for_linus' of git://git./linux/kernel/git/mchehab/linux-media

* 'v4l_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-media:
  [media] omap3isp: Fix crash caused by subdevs now having a pointer to devnodes

Merge branch 'for-linus' of git://git./linux/kernel/git/mason/linux-btrfs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  Btrfs: call d_instantiate after all ops are setup
  Btrfs: fix worker lock misuse in find_worker

Merge git://git./linux/kernel/git/davem/sparc

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
  sparc64: Fix MSIQ HV call ordering in pci_sun4v_msiq_build_irq().

Merge git://git./linux/kernel/git/davem/net

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  netfilter: xt_connbytes: handle negation correctly
  net: relax rcvbuf limits
  rps: fix insufficient bounds checking in store_rps_dev_flow_table_cnt()
  net: introduce DST_NOPEER dst flag
  mqprio: Avoid panic if no options are provided
  bridge: provide a mtu() method for fake_dst_ops

Merge branch 'nf' of git://1984.lsi.us.es/net

netfilter: xt_connbytes: handle negation correctly

"! --connbytes 23:42" should match if the packet/byte count is not in range.

As there is no explict "invert match" toggle in the match structure,
userspace swaps the from and to arguments
(i.e., as if "--connbytes 42:23" were given).

However, "what <= 23 && what >= 42" will always be false.

Change things so we use "||" in case "from" is larger than "to".

This change may look like it breaks backwards compatibility when "to" is 0.
However, older iptables binaries will refuse "connbytes 42:0",
and current releases treat it to mean "! --connbytes 0:42",
so we should be fine.

Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>

Btrfs: call d_instantiate after all ops are setup

This closes races where btrfs is calling d_instantiate too soon during
inode creation.  All of the callers of btrfs_add_nondir are updated to
instantiate after the inode is fully setup in memory.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Chris Mason <chris.mason@oracle.com>

Btrfs: fix worker lock misuse in find_worker

Dan Carpenter noticed that we were doing a double unlock on the worker
lock, and sometimes picking a worker thread without the lock held.

This fixes both errors.

Signed-off-by: Chris Mason <chris.mason@oracle.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>

net: relax rcvbuf limits

skb->truesize might be big even for a small packet.

Its even bigger after commit 87fb4b7b533 (net: more accurate skb
truesize) and big MTU.

We should allow queueing at least one packet per receiver, even with a
low RCVBUF setting.

Reported-by: Michal Simek <monstr@monstr.eu>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

rps: fix insufficient bounds checking in store_rps_dev_flow_table_cnt()

Setting a large rps_flow_cnt like (1 << 30) on 32-bit platform will
cause a kernel oops due to insufficient bounds checking.

if (count > 1<<30) {
/* Enforce a limit to prevent overflow */
return -EINVAL;
}
count = roundup_pow_of_two(count);
table = vmalloc(RPS_DEV_FLOW_TABLE_SIZE(count));

Note that the macro RPS_DEV_FLOW_TABLE_SIZE(count) is defined as:

... + (count * sizeof(struct rps_dev_flow))

where sizeof(struct rps_dev_flow) is 8.  (1 << 30) * 8 will overflow
32 bits.

This patch replaces the magic number (1 << 30) with a symbolic bound.

Suggested-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Xi Wang <xi.wang@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: introduce DST_NOPEER dst flag

Chris Boot reported crashes occurring in ipv6_select_ident().

[  461.457562] RIP: 0010:[<ffffffff812dde61>]  [<ffffffff812dde61>]
ipv6_select_ident+0x31/0xa7

[  461.578229] Call Trace:
[  461.580742] <IRQ>
[  461.582870]  [<ffffffff812efa7f>] ? udp6_ufo_fragment+0x124/0x1a2
[  461.589054]  [<ffffffff812dbfe0>] ? ipv6_gso_segment+0xc0/0x155
[  461.595140]  [<ffffffff812700c6>] ? skb_gso_segment+0x208/0x28b
[  461.601198]  [<ffffffffa03f236b>] ? ipv6_confirm+0x146/0x15e
[nf_conntrack_ipv6]
[  461.608786]  [<ffffffff81291c4d>] ? nf_iterate+0x41/0x77
[  461.614227]  [<ffffffff81271d64>] ? dev_hard_start_xmit+0x357/0x543
[  461.620659]  [<ffffffff81291cf6>] ? nf_hook_slow+0x73/0x111
[  461.626440]  [<ffffffffa0379745>] ? br_parse_ip_options+0x19a/0x19a
[bridge]
[  461.633581]  [<ffffffff812722ff>] ? dev_queue_xmit+0x3af/0x459
[  461.639577]  [<ffffffffa03747d2>] ? br_dev_queue_push_xmit+0x72/0x76
[bridge]
[  461.646887]  [<ffffffffa03791e3>] ? br_nf_post_routing+0x17d/0x18f
[bridge]
[  461.653997]  [<ffffffff81291c4d>] ? nf_iterate+0x41/0x77
[  461.659473]  [<ffffffffa0374760>] ? br_flood+0xfa/0xfa [bridge]
[  461.665485]  [<ffffffff81291cf6>] ? nf_hook_slow+0x73/0x111
[  461.671234]  [<ffffffffa0374760>] ? br_flood+0xfa/0xfa [bridge]
[  461.677299]  [<ffffffffa0379215>] ?
nf_bridge_update_protocol+0x20/0x20 [bridge]
[  461.684891]  [<ffffffffa03bb0e5>] ? nf_ct_zone+0xa/0x17 [nf_conntrack]
[  461.691520]  [<ffffffffa0374760>] ? br_flood+0xfa/0xfa [bridge]
[  461.697572]  [<ffffffffa0374812>] ? NF_HOOK.constprop.8+0x3c/0x56
[bridge]
[  461.704616]  [<ffffffffa0379031>] ?
nf_bridge_push_encap_header+0x1c/0x26 [bridge]
[  461.712329]  [<ffffffffa037929f>] ? br_nf_forward_finish+0x8a/0x95
[bridge]
[  461.719490]  [<ffffffffa037900a>] ?
nf_bridge_pull_encap_header+0x1c/0x27 [bridge]
[  461.727223]  [<ffffffffa0379974>] ? br_nf_forward_ip+0x1c0/0x1d4 [bridge]
[  461.734292]  [<ffffffff81291c4d>] ? nf_iterate+0x41/0x77
[  461.739758]  [<ffffffffa03748cc>] ? __br_deliver+0xa0/0xa0 [bridge]
[  461.746203]  [<ffffffff81291cf6>] ? nf_hook_slow+0x73/0x111
[  461.751950]  [<ffffffffa03748cc>] ? __br_deliver+0xa0/0xa0 [bridge]
[  461.758378]  [<ffffffffa037533a>] ? NF_HOOK.constprop.4+0x56/0x56
[bridge]

This is caused by bridge netfilter special dst_entry (fake_rtable), a
special shared entry, where attaching an inetpeer makes no sense.

Problem is present since commit 87c48fa3b46 (ipv6: make fragment
identifications less predictable)

Introduce DST_NOPEER dst flag and make sure ipv6_select_ident() and
__ip_select_ident() fallback to the 'no peer attached' handling.

Reported-by: Chris Boot <bootc@bootc.net>
Tested-by: Chris Boot <bootc@bootc.net>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

mqprio: Avoid panic if no options are provided

Userspace may not provide TCA_OPTIONS, in fact tc currently does
so not do so if no arguments are specified on the command line.
Return EINVAL instead of panicing.

Signed-off-by: Thomas Graf <tgraf@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

bridge: provide a mtu() method for fake_dst_ops

Commit 618f9bc74a039da76 (net: Move mtu handling down to the protocol
depended handlers) forgot the bridge netfilter case, adding a NULL
dereference in ip_fragment().

Reported-by: Chris Boot <bootc@bootc.net>
CC: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'for-linus' of git://neil.brown.name/md

* 'for-linus' of git://neil.brown.name/md:
  md/bitmap: It is OK to clear bits during recovery.
  md: don't give up looking for spares on first failure-to-add
  md/raid5: ensure correct assessment of drives during degraded reshape.
  md/linear: fix hot-add of devices to linear arrays.

md/bitmap: It is OK to clear bits during recovery.

commit d0a4bb492772ce5c4bdfba3744a99ed6f6fb238f introduced a
regression which is annoying but fairly harmless.

When writing to an array that is undergoing recovery (a spare
in being integrated into the array), writing to the array will
set bits in the bitmap, but they will not be cleared when the
write completes.

For bits covering areas that have not been recovered yet this is not a
problem as the recovery will clear the bits.  However bits set in
already-recovered region will stay set and never be cleared.
This doesn't risk data integrity.  The only negatives are:
 - next time there is a crash, more resyncing than necessary will
   be done.
 - the bitmap doesn't look clean, which is confusing.

While an array is recovering we don't want to update the
'events_cleared' setting in the bitmap but we do still want to clear
bits that have very recently been set - providing they were written to
the recovering device.

So split those two needs - which previously both depended on 'success'
and always clear the bit of the write went to all devices.

Signed-off-by: NeilBrown <neilb@suse.de>

md: don't give up looking for spares on first failure-to-add

Before performing a recovery we try to remove any spares that
might not be working, then add any that might have become relevant.

Currently we abort on the first spare that cannot be added.
This is a false optimisation.
It is conceivable that - depending on rules in the personality - a
subsequent spare might be accepted.
Also the loop does other things like count the available spares and
reset the 'recovery_offset' value.

If we abort early these might not happen properly.

So remove the early abort.

In particular if you have an array what is undergoing recovery and
which has extra spares, then the recovery may not restart after as
reboot as the could of 'spares' might end up as zero.

Reported-by: Anssi Hannula <anssi.hannula@iki.fi>
Signed-off-by: NeilBrown <neilb@suse.de>

md/raid5: ensure correct assessment of drives during degraded reshape.

While reshaping a degraded array (as when reshaping a RAID0 by first
converting it to a degraded RAID4) we currently get confused about
which devices are in_sync.  In most cases we get it right, but in the
region that is being reshaped we need to treat non-failed devices as
in-sync when we have the data but haven't actually written it out yet.

Reported-by: Adam Kwolek <adam.kwolek@intel.com>
Signed-off-by: NeilBrown <neilb@suse.de>

md/linear: fix hot-add of devices to linear arrays.

commit d70ed2e4fafdbef0800e73942482bb075c21578b
broke hot-add to a linear array.
After that commit, metadata if not written to devices until they
have been fully integrated into the array as determined by
saved_raid_disk.  That patch arranged to clear that field after
a recovery completed.

However for linear arrays, there is no recovery - the integration is
instantaneous.  So we need to explicitly clear the saved_raid_disk
field.

Signed-off-by: NeilBrown <neilb@suse.de>

sparc64: Fix MSIQ HV call ordering in pci_sun4v_msiq_build_irq().

This silently was working for many years and stopped working on
Niagara-T3 machines.

We need to set the MSIQ to VALID before we can set it's state to IDLE.

On Niagara-T3, setting the state to IDLE first was causing HV_EINVAL
errors.  The hypervisor documentation says, rather ambiguously, that
the MSIQ must be "initialized" before one can set the state.

I previously understood this to mean merely that a successful setconf()
operation has been performed on the MSIQ, which we have done at this
point.  But it seems to also mean that it has been set VALID too.

Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'usb-linus' of git://git./linux/kernel/git/gregkh/usb

* 'usb-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  USB: Fix usb/isp1760 build on sparc
  usb: gadget: epautoconf: do not change number of streams
  usb: dwc3: core: fix cached revision on our structure
  usb: musb: fix reset issue with full speed device

Merge branch 'upstream-linus' of git://github.com/jgarzik/libata-dev

* 'upstream-linus' of git://github.com/jgarzik/libata-dev:
  pata_of_platform: Add missing CONFIG_OF_IRQ dependency.

pata_of_platform: Add missing CONFIG_OF_IRQ dependency.

Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

ipv4: using prefetch requires including prefetch.h

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: David Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

VFS: Fix race between CPU hotplug and lglocks

Currently, the *_global_[un]lock_online() routines are not at all synchronized
with CPU hotplug. Soft-lockups detected as a consequence of this race was
reported earlier at https://lkml.org/lkml/2011/8/24/185. (Thanks to Cong Meng
for finding out that the root-cause of this issue is the race condition
between br_write_[un]lock() and CPU hotplug, which results in the lock states
getting messed up).

Fixing this race by just adding {get,put}_online_cpus() at appropriate places
in *_global_[un]lock_online() is not a good option, because, then suddenly
br_write_[un]lock() would become blocking, whereas they have been kept as
non-blocking all this time, and we would want to keep them that way.

So, overall, we want to ensure 3 things:
1. br_write_lock() and br_write_unlock() must remain as non-blocking.
2. The corresponding lock and unlock of the per-cpu spinlocks must not happen
   for different sets of CPUs.
3. Either prevent any new CPU online operation in between this lock-unlock, or
   ensure that the newly onlined CPU does not proceed with its corresponding
   per-cpu spinlock unlocked.

To achieve all this:
(a) We introduce a new spinlock that is taken by the *_global_lock_online()
    routine and released by the *_global_unlock_online() routine.
(b) We register a callback for CPU hotplug notifications, and this callback
    takes the same spinlock as above.
(c) We maintain a bitmap which is close to the cpu_online_mask, and once it is
    initialized in the lock_init() code, all future updates to it are done in
    the callback, under the above spinlock.
(d) The above bitmap is used (instead of cpu_online_mask) while locking and
    unlocking the per-cpu locks.

The callback takes the spinlock upon the CPU_UP_PREPARE event. So, if the
br_write_lock-unlock sequence is in progress, the callback keeps spinning,
thus preventing the CPU online operation till the lock-unlock sequence is
complete. This takes care of requirement (3).

The bitmap that we maintain remains unmodified throughout the lock-unlock
sequence, since all updates to it are managed by the callback, which takes
the same spinlock as the one taken by the lock code and released only by the
unlock routine. Combining this with (d) above, satisfies requirement (2).

Overall, since we use a spinlock (mentioned in (a)) to prevent CPU hotplug
operations from racing with br_write_lock-unlock, requirement (1) is also
taken care of.

By the way, it is to be noted that a CPU offline operation can actually run
in parallel with our lock-unlock sequence, because our callback doesn't react
to notifications earlier than CPU_DEAD (in order to maintain our bitmap
properly). And this means, since we use our own bitmap (which is stale, on
purpose) during the lock-unlock sequence, we could end up unlocking the
per-cpu lock of an offline CPU (because we had locked it earlier, when the
CPU was online), in order to satisfy requirement (2). But this is harmless,
though it looks a bit awkward.

Debugged-by: Cong Meng <mc@linux.vnet.ibm.com>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@vger.kernel.org

Merge git://git./linux/kernel/git/davem/net

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  net: Add a flow_cache_flush_deferred function
  ipv4: reintroduce route cache garbage collector
  net: have ipconfig not wait if no dev is available
  sctp: Do not account for sizeof(struct sk_buff) in estimated rwnd
  asix: new device id
  davinci-cpdma: fix locking issue in cpdma_chan_stop
  sctp: fix incorrect overflow check on autoclose
  r8169: fix Config2 MSIEnable bit setting.
  llc: llc_cmsg_rcv was getting called after sk_eat_skb.
  net: bpf_jit: fix an off-one bug in x86_64 cond jump target
  iwlwifi: update SCD BC table for all SCD queues
  Revert "Bluetooth: Revert: Fix L2CAP connection establishment"
  Bluetooth: Clear RFCOMM session timer when disconnecting last channel
  Bluetooth: Prevent uninitialized data access in L2CAP configuration
  iwlwifi: allow to switch to HT40 if not associated
  iwlwifi: tx_sync only on PAN context
  mwifiex: avoid double list_del in command cancel path
  ath9k: fix max phy rate at rate control init
  nfc: signedness bug in __nci_request()
  iwlwifi: do not set the sequence control bit is not needed

Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: atmel/ac97c: using software reset instead hardware reset if not available

Merge branch 'for-linus' of git://git./linux/kernel/git/sameo/mfd-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sameo/mfd-2.6:
  mfd: Include linux/io.h to jz4740-adc
  mfd: Use request_threaded_irq for twl4030-irq instead of irq_set_chained_handler
  mfd: Base interrupt for twl4030-irq must be one-shot
  mfd: Handle tps65910 clear-mask correctly
  mfd: add #ifdef CONFIG_DEBUG_FS guard for ab8500_debug_resources
  mfd: Fix twl-core oops while calling twl_i2c_* for unbound driver
  mfd: include linux/module.h for ab5500-debugfs
  mfd: Update wm8994 active device checks for WM1811
  mfd: Set tps6586x bits if new value is different from the old one
  mfd: Set da903x bits if new value is different from the old one
  mfd: Set adp5520 bits if new value is different from the old one
  mfd: Add missed free_irq in da903x_remove

vfs: __read_cache_page should use gfp argument rather than GFP_KERNEL

lockdep reports a deadlock in jfs because a special inode's rw semaphore
is taken recursively.  The mapping's gfp mask is GFP_NOFS, but is not
used when __read_cache_page() calls add_to_page_cache_lru().

Signed-off-by: Dave Kleikamp <dave.kleikamp@oracle.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>