Fix for the problem detected by Ingo Molnar:
enabling CONFIG_PCWATCHDOG=y crashes bzImage bootup.

The reason for this can be found in drivers/makefile
We first do:
obj-y                           += char/
and later we do:
obj-y                           += base/ block/ misc/ mfd/ net/ media/

So if we put a platform or isa or usb bus driver in char/watchdog
(which is called from the Makefile in drivers/char/Makefile)
then we didn't have the different device drivers initialized yet
(they are in drivers/base and drivers/usb and ...)

This fix makes sure that we compile the watchdog drivers after
drivers/base, drivers/misc, drivers/pci and drivers/usb.
We also do the compile after hwmon because in the future the
watchdog temperature support will use the hwmon system.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Add support for the ICH9 I/O chipsets to iTCO_wdt.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Add all LPC bridges for the 631xESB/632xESB I/O chipset.
The datasheet says:
* Device Function = B0:D31:FO
* Function Description = LPC interface
* DEV ID = 267xh
* Comment = 2670h-267Fh
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Add 631xESB/632xESB support to the iTCO_wdt driver.
Signed-off-by: NSergey Kononenko <sergk@sergk.org.ua>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

The default value for an unknown ioctl call is -ENOTTY.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Return the value of the nonseekable_open function and not 0.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Consolidate the timeout config register modification into a single
function.  Also, use the enabled flag in the config register to
determine whether the timer is enabled instead of a separately
maintained flag, MV64x60_WDOG_FLAG_ENABLED.

Add spinlock protection around enabling/disabling the watchdog timer.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Make sure that we disable the watchdog at start-up.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Disallow disabling of the watchdog timer unless a particular
character ('V') was recently written to the watchdog device.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Also, use the WATCHDOG_NOWAYOUT macro, rather than #ifdefs,
and use __module_get to prevent module unloading if WATCHDOG_NOWAYOUT
is set.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Allow the watchdog timer to be enabled or disabled via the
WDIOC_SETOPTIONS ioctl.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Add the ability to modify the watchdog timer timeout interval.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

WDIOC_GETTIMEOUT returns seconds, not jiffies.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Return the value of the nonseekable_open function and not 0.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Add support for arch/powerpc, specifically for the prpmc2800 platform.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Previously, the address of the watchdog timer registers was
retrieved by calling a global function, mv64x60_get_bridge_vbase().
That function doesn't exist in arch/powerpc.  Instead, we now get
the register address from a platform data resource and ioremap
the registers within the driver.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

The driver previously registered its platform device data in its own
init function--that's bogus.  Move that code to platform-specific
code in arch/ppc.  This is being done so that the platform code can
decide at runtime whether to initialize this driver or not.
Signed-off-by: NDale Farnsworth <dale@farnsworth.org>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

1. Ensure that the mouse and keyboard do not ping the watchdog.
This is the default operation of the w83627, but some BIOSes change this.

2. Increase the max timeout from 63 seconds to 255 seconds
as supported by the w83627 chip

3. Comment that the watchdog supports the w83627hg version of the chip
Signed-Off-By: NPádraig Brady <P@draigBrady.com>
Tested-by: NTomas Hodek <tomas.hodek@volny.cz>
Signed-Off-By: NWim Van Sebroeck <wim@iguana.be>

Fixup the s3c2410 watchdog driver after moving some
of the arch specific includes it has been relying on.
Signed-off-by: NBen Dooks <ben-linux@fluff.org>

From: Andrew Morton <akpm@linux-foundation.org>

This driver isn't very coding-style friendly.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>

Clean-up of the watchdog's Kconfig and makefile files.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Add mandatory WDIOC_GETSTATUS and WDIOC_GETBOOTSTATUS ioctl's for
drivers that don't have them yet.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

All watchdog device drivers are VFSs (Virtual File Systems).
We thus return a nonseekable_open(inode, file) when we open the VFS.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

* Remove the redundant check for pwrite(), given that the open() routine
  already invokes nonseekable_open().
* The WDIOF_CARDRESET flag can only be used when you can read this status
  via the WDIOC_GETSTATUS ioctl call.
* Add the mandatory WDIOC_GETBOOTSTATUS ioctl call.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

* Add MODULE_ALIAS_MISCDEV(WATCHDOG_MINOR);
* Add mandatory WDIOC_GETSTATUS and WDIOC_GETBOOTSTATUS ioctl's.
* If unknown ioctl is used we should return -ENOTTY.
* All watchdog device drivers are VFSs (Virtual File Systems).
  We thus return a nonseekable_open(inode, file) when we open the VFS.
* Make sure that /dev/watchdog can be opened by 1 parent
* Add spin-locking to prevent that forked children can disturb
  each other's operations.
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Add watchdog support for TI Davinci DM644x/DM646x processors.

Signed-off-by: Vladimir Barinov <vbarinov@ru.mvista.com> 
Signed-off-by: Kevin Hilman <khilman@mvista.com> 
Signed-off-by: Wim Van Sebroeck <wim@iguana.be> 

This patch implements the driver necessary use the Analog Devices
Blackfin processor's on-chip watchdog controller, supports
BF53[123]/BF53[467]/BF54[2489]/BF561.
Signed-off-by: NMike Frysinger <michael.frysinger@analog.com>
Signed-off-by: NBryan Wu <bryan.wu@analog.com>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Driver for internal mpc5200 watchdog on general purpose timer 0.
For IPB clock of 132 MHz the maximum timeout is about 32 seconds.
Signed-off-by: NDomen Puncer <domen.puncer@telargo.com>
Signed-off-by: NWim Van Sebroeck <wim@iguana.be>

Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This avoids a conflict with sparse builds.

Reported by Alexey Dobriyan, fix suggested by Al Viro
Signed-off-by: NAndi Kleen <ak@suse.de>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The recent 9p commit: bd238fb4 that
supposedly only moved files also introduced a new 9p sysctl interface
that did not properly register it's sysctl binary numbers.

And since it was only for debugging clearly did not need a binary fast
path in any case.  So this patch just remove the binary numbers.

See Documentation/sysctl/ctl_unnumbered.txt for more details.

While I was at it I cleaned up the sysctl initializers a little as
well so there is less to read.

Cc: Latchesar Ionkov <lucho@ionkov.net>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Signed-off-by: NEric W. Biederman <ebiederm@xmission.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Fix a typo in SubmittingPatches where "probably" was spelt "probabally".
Signed-off-by: NLinus Nilsson <lajnold@gmail.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Change a headline to reflect that there are three main types of kernel
locking, not two.
Signed-off-by: NLinus Nilsson <lajnold@gmail.com>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
  IB/ehca: Support small QP queues
  IB/ehca: Make internal_create/destroy_qp() static
  IB/ehca: Move ehca2ib_return_code() out of line
  IB/ehca: Generate async event when SRQ limit reached
  IB/ehca: Support large page MRs
  IB/mlx4: Fix error path in create_qp_common()
  mlx4_core: Change command token on timeout
  IB/mthca: Change command token on timeout
  IB/ipath: Remove ipath_layer dead code
  IB/mlx4: Fix leaks in __mlx4_ib_modify_qp

Now that the last inlined instances are gone, all that is left to do
is turning disable_irq_nosync on arm26 and m68k from defines to aliases
and we are all set - we can make these externs in linux/interrupt.h
uncoditional and kill remaining instances in asm/irq.h
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

* master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6: (60 commits)
  [SCSI] libsas: make ATA functions selectable by a config option
  [SCSI] bsg: unexport sg v3 helper functions
  [SCSI] bsg: fix bsg_unregister_queue
  [SCSI] bsg: make class backlinks
  [SCSI] 3w-9xxx: add support for 9690SA
  [SCSI] bsg: fix bsg_register_queue error path
  [SCSI] ESP: Increase ESP_BUS_TIMEOUT to 275.
  [SCSI] libsas: fix scr_read/write users and update the libata documentation
  [SCSI] mpt fusion: update Kconfig help
  [SCSI] scsi_transport_sas: add destructor for bsg
  [SCSI] iscsi_tcp: buggered kmalloc()
  [SCSI] qla2xxx: Update version number to 8.02.00-k2.
  [SCSI] qla2xxx: Add ISP25XX support.
  [SCSI] qla2xxx: Use pci_try_set_mwi().
  [SCSI] qla2xxx: Use PCI-X/PCI-Express read control interfaces.
  [SCSI] qla2xxx: Re-factor isp_operations to static structures.
  [SCSI] qla2xxx: Validate mid-layer 'underflow' during check-condition handling.
  [SCSI] qla2xxx: Correct setting of 'current' and 'supported' speeds during FDMI registration.
  [SCSI] qla2xxx: Generalize iIDMA support.
  [SCSI] qla2xxx: Generalize FW-Interface-2 support.
  ...

Not everyone wants libsas automatically to pull in libata.  This patch
makes the behaviour configurable, so you can build libsas with or
without ATA support.
Signed-off-by: NJames Bottomley <James.Bottomley@SteelEye.com>

* 'devel' of master.kernel.org:/home/rmk/linux-2.6-arm: (60 commits)
  [ARM] 4524/1: S3C: Move register out of include/asm-arm/arch-s3c2410
  [ARM] 4523/1: S3C: Remove FIFO_MAX from uncompression headers
  [ARM] 4522/1: S3C: split include/asm-arm/arch/memory.h
  [ARM] 4521/2: S3C: Reorganise VA mapping headers
  [ARM] 4520/1: S3C: Remove old VA values from static map
  [ARM] 4519/1: S3C: split S3C2400 values out of S3C24XX map.h
  [ARM] 4518/1: S3C: Rename watchdog configuration options
  [ARM] 4517/1: S3C: Fix debug macros for ARM926 output
  [ARM] 4516/1: S3C: Fix uncompressor serial output for ARM926
  [ARM] 4515/1: S3C: Move uncompress code to plat-s3c
  [ARM] 4514/1: S3C: Rename DEBUG_S3C2410_PORT and DEBUG_S3C_UART
  [ARM] 4513/1: S3C: Rename CONFIG_S3C2410_LOWLEVEL_UART_PORT
  [ARM] 4512/1: S3C: rename the debug macros for per-cpu updates
  [ARM] 4511/1: S3C: updated LLSERIAL Kconfig defines for CPU support
  [ARM] 4510/1: S3C: split debug-macro support into plat-s3c
  [ARM] 4509/1: S3C: Create initial arch/arm/plat-s3c
  [ARM] 4508/1: S3C: Move items to include/asm-arm/plat-s3c
  [ARM] 4461/1: MXC platform and i.MX31ADS core support
  [ARM] 4507/1: pxa2xx clock_event_device
  [ARM] 4497/1: Only allow safe cache configurations on ARMv6 and later
  ...

* 'for-linus' of git://git.o-hand.com/linux-rpurdie-leds:
  leds: Convert from struct class_device to struct device
  leds: leds-gpio for ngw100
  leds: Add warning printks in error paths
  leds: Fix trigger unregister_simple if register_simple fails
  leds: Use menuconfig objects II - LED
  leds: Teach leds-gpio to handle timer-unsafe GPIOs
  leds: Add generic GPIO LED driver

* 'for-linus' of git://git.o-hand.com/linux-rpurdie-backlight:
  leds: cr_bllcd.c: build fix
  backlight: Convert from struct class_device to struct device
  backlight: Fix order of Kconfig entries