The problem fixed in commit 014b164e
(space leak with in-band cleanmarkers) would have been caught a lot
quicker if our paranoid debugging mode had included adding up the size
counts from all the eraseblocks and comparing the totals with the counts
in the superblock. Add that.

Make jffs2_mark_erased_block() file the newly-erased block on the
free_list before calling the debug function, to make it happy.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

We were accounting for the cleanmarker by calling jffs2_link_node_ref()
(without locking!), which adjusted both superblock and per-eraseblock
accounting, subtracting the size of the cleanmarker from {jeb,c}->free_size
and adding it to {jeb,c}->used_size.

But only _then_ were we adding the size of the newly-erased block back
to the superblock counts, and we were adding each of jeb->{free,used}_size
to the corresponding superblock counts. Thus, the size of the cleanmarker
was effectively subtracted from the superblock's free_size _twice_.

Fix this, by always adding a full eraseblock size to c->free_size when
we've erased a block. And call jffs2_link_node_ref() under the proper
lock, while we're at it.

Thanks to Alexander Yurchenko and/or Damir Shayhutdinov for (almost)
pinpointing the problem.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

... instead of <linux/semaphore.h> which we don't need any more anyway.
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

With modern systems using bus-hold instead of bus pull-up, it can
often lead to erroneous reporting of NAND devices where there are
none. Do a double probe to ensure that the result we got the first
time is repeatable, and if it is not then return that there is no
chip there.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Add support to disable ECC checking for a given chip
when passed by the board via the platform data.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Add support for the ECC layout to be passed via the
platform data specified by the board.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

If a block's ecc field is all 0xff, then ignore the ECC
correction. This is for systems where some of the blocks,
such as the initial cramfs are written without ECC and
need to be loaded on start.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This adds support for using large page NAND devices
with the S3C24XX NAND controller. This also adds the
file Documentation/arm/Samsung-S3C24XX/NAND.txt to
describe the differences.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Commit 03680b1e incorrectly
was assuming S3C2410_NFCONF was being used to select the
NAND chip. Fix this error by ising the sel_reg.
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

A power loss while writing can result in a page becoming unreadable.  
When the device is mounted again, reading that page gives controller 
errors. Upper level software like JFFS2 treat -EIO as fatal, refusing to 
mount at all.  That means it is necessary to treat the error as an ECC 
error to allow recovery. Note that typically in this case, the 
eraseblock can still be erased and rewritten i.e. it has not become a 
bad block.
Signed-off-by: Adrian Hunter <ext-adrian.hunter@nokia.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Minor tweaks to omap_nor ... as with most platform drivers, its probe
and remove logic can (and should!) safely vanish in most configs.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This is a known erratum confirmed by Spansion.  I have an errata document,
but I can't find a link to it anywhere on their site to include here.

Some of the S29GL064N chips report 64 sectors when they should report 128,
and some of S29GL032N chips report 127 sectors when they should report 63.

Note that when the chip dies are fixed by Spansion, they will still have
the same id.  The fix is done in such a way that it won't affect corrected
chips.

The fixups use the extended id made available by a previous patch.  Without
that, virtually all newer AMD/Spansion chips will have the same ID (0x227e)
and it's not possible to apply the fixup to the correct chips.
Signed-off-by: Trent Piepho <tpiepho@freescale.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

AMD/Spansion use a device id of 0x7e to indicate an extended device is
present at offset 0xe and 0xf in the query data.

I've verified with Spansion that all their chips (mfr == 0x01) with an id
of 0x7e use it to indicate an extended id is present.  What's more, there
are no chips with a NON-extended id that is the same as a different chip's
extended id.  In other words, when the extended ID is present, one can
replace the normal id with the extended id without losing any information.
Which is what I've done.
Signed-off-by: Trent Piepho <tpiepho@freescale.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Using current driver elbc sometimes hangs during nand write. Reading back
last byte helps though (thanks to Scott Wood for the idea).
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch sets mtd->name to the platform bus ID in the plat_nand
driver, so that you can specify partitions readily with mtdparts=.

Currently it relies on nand_base filling in the name from the device,
which results in names like "NAND 256MiB 3,3V 8-bit", that you can't
use with cmdlineparts.
Signed-off-by: Hamish Moffatt <hamish@cloud.net.au>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Fix a race condition in fsl_elbc_run_command
Fix incorrect usage of clearbits32 that bashed option register
Remove work around for bashed register
Signed-off-by: Mike Hench <mhench@elutions.com>
Acked-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Currently fsl_elbc_nand doesn't initialize mtd->name, and this causes
nand_get_flash_type() to assign name that is equal to chip type, like
this:

   root@b1:~# cat /proc/mtd
   dev:    size   erasesize  name
   mtd0: 00800000 00010000 "fe000000.flash"
   mtd1: 02000000 00004000 "NAND 32MiB 3,3V 8-bit"

mtd0 is physmap_of flash (normal name), and mtd1 is fsl_elbc_nand.

Despite inconsistency, with mtd name like this specifying paritions
from the kernel command line becomes a torture (though, I didn't tried
and not sure if mtdparts= can handle spaces at all). Plus, this causes
real bugs when multiple fsl_elbc_nand chips registered.

With this patch applied fsl_elbc_nand chip will have proper name:

   root@b1:~# cat /proc/mtd
   dev:    size   erasesize  name
   mtd0: 00800000 00010000 "fe000000.flash"
   mtd1: 02000000 00004000 "e0600000.flash"

p.s. We can't use priv->dev->bus_id as in physmap_of, because
fsl_elbc_nand pretends to be a localbus controller, so its bus_id is
"address.localbus", which is incorrect and thus will also not work
for multiple chips.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Signed-off-by: Matteo Croce <technoboy85@gmail.com>
Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Eugene Konev <ejka@imfi.kspu.ru>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This is very simple driver, NAND is connected through localbus,
and User-Programmable Machine is doing various adjustments to
speak NAND. No special efforts needed to do read and write cycles,
though to control ALE and CLE phases, we ask UPM to generate exact
pre-programmed signals on the localbus lines.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Add support for the SST 36VF3203 flash chip. It is used on Emerson 
KSI8560 board.
Signed-off-by: Andrei Dolnikov <adolnikov@ru.mvista.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This enhances plat-ram to take a map_probes argument in
the platform_data structure which allow plat-ram to support
any direct-mapped device that MTD supports (jedec, cfi, amd ..)

A few items are also fixed:
- Don't panic if probes is 0
- Actually use the partition list that is passed in
Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Jason Gunthorpe <jgunthorpe@obsidianresearch.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Untested, but shouldn't break anything... Makes MTD_XIP arch
independent. I guess this is why xip_iprefetch() was made for.
Signed-off-by: Paulius Zaleckas <paulius.zaleckas@teltonika.lt>
Acked-by: Nicolas Pitre <nico@cam.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This is preliminary since:

1. It supports only _one_ chip select at the moment. As there is no
   existing platforms available using two chip selects of the NAND
   controller, it shall really not include code for supporting the
   2nd chip select for now, as such code cannot be verified.

2. It resorts to the default and simpliest memory based badblock
   table

3. Only limited types of nand flash are currently supported. Most
   PXA3xx processors come with on-chip NAND flash dies, so there
   isn't much flexibility for other types of NAND.

4. The NAND controller should be configured to detect the device's
   ID, thus making it difficult to use nand_scan_ident() to assist
   the detection process (though it's not impossible)

TODO: fix all the above limitations of cuz :-)
Signed-off-by: eric miao <eric.miao@marvell.com>
Cc: Sergey Podstavin <spodstavin@ru.mvista.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Acked-By: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Return 0 partitions instead of -EINVAL on no mtdpart= argument in kernel
cmdline or missing partition info for device.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Acked-by: Stefan Roese <sr@denx.de>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch extends the existing MAPS driver for the Nvidia CK804 chipset
(ck804xrom.c) to also work on the Nvidia MCP55 chipset. As both chipsets
are rather similar, suporting them both with the same driver is easy.
Signed-off-by: Carl-Daniel Hailfinger <c-d.hailfinger.devel.2006@gmx.net>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch makes the needlessly global struct rfd_ftl_tr static.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch adds a proper prototype for onenand_bbt_read_oob() in
include/linux/mtd/onenand.h
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch makes the needlessly global cfi_staa_erase_varsize() static.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch makes the following needlessly global code static:
- ftl_freepart()
- struct ftl_tr
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Sparse spotted that 0 was compared to pointers.

While I was at it, I also moved the assignments out of the if's.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Every file should include the headers containing the externs for its
global functions (in this case for mtdram_init_device()).
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch adds proper prototypes for nftl_{read,write}_oob() in
include/linux/mtd/nftl.h
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch adds proper prototypes for inftl_{read,write}_oob() in
include/linux/mtd/inftl.h
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

This patch adds the missing MODULE_LICENSE("GPL").
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Previously, a READ command was erroneously issued rather than SEQIN.
Signed-off-by: Scott Wood <scottwood@freescale.com>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Signed-off-by: Mark Hindley <mark@hindley.org.uk>
Signed-off-by: David Woodhouse <dwmw2@infradead.org>

Signed-off-by: David Woodhouse <dwmw2@infradead.org>