Merge ... /linux/kernel/git/torvalds/linux-2.6.git

[PATCH] x86_64 fsnotify build fix

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] softdog build fix

drivers/char/watchdog/softdog.c:94: too many arguments to function `emergency_restart'

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] eurotechwdt build fix

drivers/char/watchdog/eurotechwdt.c:165: too many arguments to function `emergency_restart'

Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] i386: clean up user_mode macros

 - make the new user_mode() return 0 or 1 (same as x86_64)

 - remove conditional jump from user_mode_vm() it's called every timer
   tick on each CPU on SMP)

Signed-off-by: Chuck Ebbert <76306.1226@compuserve.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge /pub/scm/linux/kernel/git/davem/net-2.6

[PATCH] Change PowerPC MPC8xx maintainer

As Marcelo has been spending a great deal of time working on MPC8xx
systems of late (thanks!) and has more time than I do now for it, I'm
handing this over to him.

Signed-off-by: Tom Rini <trini@kernel.crashing.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[XFRM]: Fix possible overflow of sock->sk_policy

Spotted by, and original patch by, Balazs Scheidler.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] fix MAX_USER_RT_PRIO and MAX_RT_PRIO

Here's the patch again to fix the code to handle if the values between
MAX_USER_RT_PRIO and MAX_RT_PRIO are different.

Without this patch, an SMP system will crash if the values are
different.

Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Dean Nelson <dcn@sgi.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[NETFILTER]: Fix ip_conntrack_put() prototype.

The function is not inline.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] Fix RLIMIT_RTPRIO breakage

RLIMIT_RTPRIO is supposed to grant non privileged users the right to use
SCHED_FIFO/SCHED_RR scheduling policies with priorites bounded by the
RLIMIT_RTPRIO value via sched_setscheduler(). This is usually used by
audio users.

Unfortunately this is broken in 2.6.13rc3 as you can see in the excerpt
from sched_setscheduler below:

        /*
         * Allow unprivileged RT tasks to decrease priority:
         */
        if (!capable(CAP_SYS_NICE)) {
                /* can't change policy */
                if (policy != p->policy)
                        return -EPERM;

After the above unconditional test which causes sched_setscheduler to
fail with no regard to the RLIMIT_RTPRIO value the following check is made:

               /* can't increase priority */
                if (policy != SCHED_NORMAL &&
                    param->sched_priority > p->rt_priority &&
                    param->sched_priority >
                                p->signal->rlim[RLIMIT_RTPRIO].rlim_cur)
                        return -EPERM;

Thus I do believe that the RLIMIT_RTPRIO value must be taken into
account for the policy check, especially as the RLIMIT_RTPRIO limit is
of no use without this change.

The attached patch fixes this problem.

Signed-off-by: Andreas Steinmetz <ast@domdv.de>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge master.kernel.org:/home/rmk/linux-2.6-arm

Merge kernel.org:/home/rmk/linux-2.6-arm-smp

Fix compiler warning in qla_iocb.c

Remove bogus initialization that was re-done (correctly) later.

[PATCH] ARM: 2831/1: S3C2440 - split s3c2440 clocks from central clock code

Patch from Ben Dooks

Split the s3c2440 specific clocks from the arch clock support, to
make the code clearer.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] acpi: Don't call acpi_sleep_prepare from acpi_power_off

Now that all of the code paths that call acpi_power_off
have been modified to call either call kernel_power_off
(which calls apci_sleep_prepare by way of acpi_shutdown)
or to call acpi_sleep_prepare directly it is redundant to call
acpi_sleep_prepare from acpi_power_off.

So simplify the code and simply don't call acpi_sleep_prepare.

In addition there is a little error handling done so if we
can't register the acpi class we don't hook pm_power_off.

I think I have done the right thing with the CONFIG_PM define
but I'm not certain.  Can this code even be compiled if
CONFIG_PM is false?

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] acpi_power_off: Don't switch to the boot cpu

machine_power_off on i386 and x86_64 now switch to the
boot cpu out of paranoia and because the MP Specification indicates it
is a good idea on reboot, so for those architectures it is a noop.
I can't see anything in the acpi spec that requires you to be on
the boot cpu to power off the system, so this should not be an issue
for ia64.  In addition ia64 has the altix a massive multi-node
system where switching to the boot cpu sounds insane as we may
hot removed the boot cpu.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86_64 sync machine_power_off with i386

i386 machine_power_off was disabling the local apic
and all of it's users wanted to be on the boot cpu.
So call machine_shutdown which places us on the boot
cpu and disables the apics.  This keeps us in sync
and reduces the number of cases we need to worry about in
the power management code.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] APM: Remove redundant call to set_cpus_allowed

machine_power_off now always switches to the boot cpu so there
is no reason for APM to also do that.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] i386 machine_power_off cleanup

Call machine_shutdown() to move to the boot cpu
and disable apics.  Both acpi_power_off and
apm_power_off want to move to the boot cpu.
and we are already disabling the local apics
so calling machine_shutdown simply reuses
code.

ia64 doesn't have a special path in power_off
for efi so there is no reason i386 should.  If
we really need to call the efi power off path
the efi driver can set pm_power_off like everyone
else.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] machine_shutdown: Typo fix to actually allow specifying which cpu to reboot on

This appears to be a typo I introduced when cleaning
this code up earlier. Ooops.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] pcwd.c: Call kernel_power_off not machine_power_off

The call appears to come from process context so kernel_power_off
should be safe.  And acpi_power_off won't necessarily work if you just
call machine_power_off.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] swpsuspend: Have suspend to disk use factors of sys_reboot

The suspend to disk code was a poor copy of the code in
sys_reboot now that we have kernel_power_off, kernel_restart
and kernel_halt use them instead of poorly duplicating them inline.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] 68328serial: sysrq should use emergency_reboot

The 68328serial.c driver has a weird local reimplementation of
magic sysrq.  The code is architecture specific enough that calling
machine_restart() is probably ok.  But there is no reason not to call
emergency_restart() so do so.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] In hangcheck-timer.c call emergency_restart()

If we've hung a clean reboot does not sound like a real
option.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Fix watchdog drivers to call emergency_reboot()

If a watchdog driver has decided it is time to reboot the system
we know something is wrong and we are in interrupt context
so emergency_reboot() is what we want.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Update sysrq-B to use emergency_restart()

sysrq calls into the reboot path from an interrupt handler
we can either push the code do into process context and
call kernel_restart and get a clean reboot or we can simply
reboot the machine, and increase our chances of actually
rebooting.  emergency_reboot() seems like the closest match
to what we have previously done, and what we want.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Call emergency_reboot from panic

We know the system is in trouble so there is no question if this
is an emergecy :)

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Use kernel_power_off in sysrq-o

We already do all of the gymnastics to run from process context
to call the power off code so call into the power off code cleanly.

This especially helps acpi as part of it's shutdown logic should
run acpi_shutdown called from device_shutdown which was not
being called from here.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86_64: Implemenent machine_emergency_restart

It is not safe to call set_cpus_allowed() in interrupt
context and disabling the apics is complicated code.
So unconditionally skip machine_shutdown in machine_emergency_reboot
on x86_64.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] x86_64: Fix reboot_force

We only want to shutdown the apics if reboot_force
is not specified.  Be we are doing this both
in machine_shutdown which is called unconditionally
and if (!reboot_force).  So simply call machine_shutdown
if (!reboot_force).  It looks like something
went weird with merging some of the kexec patches for
x86_64, and caused this.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] i386: Implement machine_emergency_reboot

set_cpus_allowed is not safe in interrupt context
and disabling apics is complicated code so don't
call machine_shutdown on i386 from emergency_restart().

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Don't export machine_restart, machine_halt, or machine_power_off.

machine_restart, machine_halt and machine_power_off are machine
specific hooks deep into the reboot logic, that modules
have no business messing with.  Usually code should be calling
kernel_restart, kernel_halt, kernel_power_off, or
emergency_restart. So don't export machine_restart,
machine_halt, and machine_power_off so we can catch buggy users.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Fix the arguments to machine_restart on cris

It appears machine_restart has been working cris just
by luck.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Add emergency_restart()

When the kernel is working well and we want to restart cleanly
kernel_restart is the function to use.   But in many instances
the kernel wants to reboot when thing are expected to be working
very badly such as from panic or a software watchdog handler.

This patch adds the function emergency_restart() so that
callers can be clear what semantics they expect when calling
restart.  emergency_restart() is expected to be callable
from interrupt context and possibly reliable in even more
trying circumstances.

This is an initial generic implementation for all architectures.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Make ctrl_alt_del call kernel_restart to get a proper reboot.

It is obvious we wanted to call kernel_restart here
but since we don't have it the code was expanded inline and hasn't
been correct since sometime in 2.4.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Refactor sys_reboot into reusable parts

Because the factors of sys_reboot don't exist people calling
into the reboot path duplicate the code badly, leading to
inconsistent expectations of code in the reboot path.

This patch should is just code motion.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] Add missing device_suspsend(PMSG_FREEZE) calls.

In the recent addition of device_suspend calls into
sys_reboot two code paths were missed.

Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] qla2xxx: Kconfig dependency fix

*** Warning: "fc_remote_port_block" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!
*** Warning: "scsi_is_fc_rport" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!
*** Warning: "fc_remote_port_unblock" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!
*** Warning: "fc_remote_port_rolechg" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!
*** Warning: "fc_release_transport" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!
*** Warning: "fc_remove_host" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!
*** Warning: "fc_remote_port_add" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!
*** Warning: "fc_attach_transport" [drivers/scsi/qla2xxx/qla2xxx.ko] undefined!

Cc: James Bottomley <James.Bottomley@steeleye.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] qla: remove anonymous union

Older gcc's dont support anonymous unions, so this driver gets hundreds of
error.

Fortunately the fix is easy...

Cc: James Bottomley <James.Bottomley@steeleye.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: fix oops fix

Cc: Robert Love <rml@novell.com>
Cc: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: add x86-64 syscall entries

Add inotify syscall entries to x86-64.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: add missing hook to sys32_open

Add missing fsnotify_open() hook to sys32_open().

Add fsnotify_open() hook to sys32_open() on x86-64.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: check retval in init

Check for (unlikely) errors in the filesystem initialization stuff in
our module_init() function.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: change default limits

Change default inotify limits: Maximum instances per user to 128 and
maximum events per queue to 16k.  The max instances used to be 128; the
change to 8 was a mistake.  Memory consumption is fine.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: exit path cleanups

Handle error out paths better.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: oops fix

Bug fix: Ensure that the fd passed to inotify_add_watch() and
inotify_rm_watch() belongs to inotify.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: use fget_light

As an optimization, use fget_light() and fput_light() where possible.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] inotify: misc. cleanup

Miscellaneous invariant clean up, comment fixes, and so on.  Trivial
stuff.

Signed-off-by: Robert Love <rml@novell.com>
Signed-off-by: John McCutchan <ttb@tentacle.dhs.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

Merge /linux/kernel/git/jejb/scsi-for-linus-2.6

[PATCH] ARM SMP: Mark device mappings as "device" in ARMv6 parlance

ARMv6 introduces memory types into the page tables.  Mark devices
mappings with the "shared device" memory type.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM SMP: Add ARMv6 memory barriers

Convert explicit gcc asm-based memory barriers into smp_mb() calls.
These change between barrier() and the ARMv6 data memory barrier
instruction depending on whether ARMv6 SMP is enabled.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM SMP: Use exclusive load/store for __xchg

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2830/1: Fix Jornada 720 PCMCIA-support

Patch from Michael Gernoth

This patch lets the Jornada 720 PCMCIA-driver compile again. The
resulting driver has been tested on a Jornada with a CF-card, which
was mounted and accessed successfully.

Signed-off-by: Michael Gernoth <michael@gernoth.net>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2829/1: S3C2410 - split s3c2440 irq specifics from core irq.c

Patch from Ben Dooks

Remove the need for the #ifdefs and place the IRQ handling code for
the s3c2440 into a new file, which is only compiled when the
s3c2440 cpu support is enabled.

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2828/1: BAST - remove static map of ASIX area

Patch from Ben Dooks

There is no point in mapping this staticaly, the driver is going
to ioremap() the area as it sees fit. Also correct the dates on
the changelog comments

Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[PATCH] ARM: 2819/1: Fix several S3C24x0 IIS defines

Patch from Dimitry Andric

- Change S3C2440_IISCON_MPLL to S3C2440_IISMOD_MPLL:
  The S3C2440 IISCON register doesn\'t control the master clock selection, this is done with the IISMOD register.
- Correct S3C2410_IISMOD_256FS and S3C2410_IISMOD_384FS:
  This is set via bit 2 of IISMOD, not bit 1.
- Add S3C2410_IISCON_PSCEN (prescaler enable), for completeness\' sake.

Signed-off-by: Dimitry Andric <dimitry.andric@tomtom.com>
Signed-off-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[ATM]: speedtch: Revert 86cf42e4e029b83110cf98692420239103363dbf

It was already fixed more sufficiently by Andrew Morton's
change 843c944fb86e5e31ee7b319172e657ea22301322.

Noted by Duncan Sands.

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

[TG3]: Update driver version and reldate.

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

[TG3]: add 5780 fiber support

Add 5780S support by adding a new tg3_setup_fiber_mii_phy() function and
a timer function for parallel link detection. 5780S uses standard MII
registers for 1000BaseX and runs in GMII mode as opposed to TBI mode on
older serdes chips.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: disallow jumbo TSO on 5780

Disallow jumbo TSO on 5780 due to hardware restrictions.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: consolidate all DMA water mark settings

Consolidate all DMA watermark settings for standard and jumbo frames on
all chips in tg3_init_bufmgr_config() and add new settings for 5780.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: add variable buffer size for standard ring

Add a new rx_pkt_buf_sz to the tg3 structure to support variable buffer
sizes on the standard ring.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: add 5780 basic jumbo frame support

Add basic jumbo frames support for 5780. This chip supports jumbo frames
on the standard receive ring without the jumbo ring. The
TG3_FLAG_JUMBO_ENABLE is changed to TG3_FLAG_JUMBO_RING_ENABLE to
indicate using the jumbo ring on 5704 and older chips. A new
TG3_FLG2_JUMBO_CAPABLE flag is added to indicate jumbo frames support
with or without the jumbo ring.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

[TG3]: add 5780 basic support

Add 5780 PCI IDs, chip IDs, and other basic support.

Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge /pub/scm/linux/kernel/git/davem/net-2.6

Merge /pub/scm/linux/kernel/git/davem/sparc-2.6

[EMATCH]: Remove feature ifdefs in meta ematch.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Thomas Graf <tgraf@suug.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC64]: Move syscall success and newchild state out of thread flags.

These two bits were accesses non-atomically from assembler
code.  So, in order to eliminate any potential races resulting
from that, move these pieces of state into two bytes elsewhere
in struct thread_info.

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

[SPARC64]: Privatize sun5_timer.

It is only used by some localized code in irq.c, and also
delete enable_prom_timer() as that is totally unused.

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

[SPARC64]: Kill totally unused inline functions from asm/spitfire.h

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

[SPARC64]: Simplify asm/rwsem.h slightly.

rwsem_atomic_update and rwsem_atomic_add can be implemented
straightly using atomic_*() routines.

Also, rwsem_cmpxchgw() is totally unused, kill it.

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

[SPARC64]: Non-atomic bitops do not need volatile operations

Noticed this while comparing sparc64's bitops.h to ppc64's.
We can cast the volatile memory argument to be non-volatile.

While we're here, __inline__ --> inline.

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

[SPARC]: Fix __KERNEL_SYSCALLS__ defining in envctrl.c and bbc_envctrl.c

It needs to happen before any header includes because nowadays
some things implicitly include asm/unistd.h which ends up being
before the __KERNEL_SYSCALLS__ define gets done.

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

[SPARC]: sbus/vfc_i2c: remove sleep_on() usage

Use wait-queue directly instead of the deprecated sleep_on()
function. This required adding a local wait-queue. Also use new (added in
separate patch to K-J) usecs to jiffies function to convert value.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Domen Puncer <domen@coderock.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: sbus/envctrl: replace schedule_timeout() with msleep_interruptible()

Use msleep_interruptible() instead of schedule_timeout() to guarantee
the task delays as expected. Change the units of poll_interval to
msecs as it is only used in this delay.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Domen Puncer <domen@coderock.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[SPARC]: sbus/aurora: replace schedule_timeout() with msleep_interruptible()

Use msleep_interruptible() instead of
schedule_timeout() to guarantee the task delays as expected.

Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Maximilian Attems <janitor@sternwelten.at>
Signed-off-by: Domen Puncer <domen@coderock.org>

[SPARC]: Remvoe APM_RTC_IS_GMT from config.

I can't see any effect of this option outside the i386-specific APM code.

Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

[W1]: Do not use NFLOG netlink number.

Use the reserved by never used NETLINK_SKIP value instead.

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

[IPV6]: fix implicit declaration of function `xfrm6_tunnel_unregister'

Signed-off-by: Cal Peake <cp@absolutedigital.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[PATCH] ARM SMP: Fix ARMv6 spinlock and semaphore implementations

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

Merge master.kernel.org:/home/rmk/linux-2.6-serial

Merge master.kernel.org:/home/rmk/linux-2.6-arm

[PATCH] ARM: 2825/1: S3C2410: turns %d into %ld on DMA printk

Patch from Lucas Correia Villa Real

This patch replaces the sizeof()'s %d specifier by %ld on a S3C2410 DMA
printk.

Signed-off-by: Lucas Correia Villa Real <lucasvr@gobolinux.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

/home/lenb/linux-2.6 branch 'acpi-2.6.12'

ACPI: delete unnecessary EC console messages

http://bugzilla.kernel.org/show_bug.cgi?id=4534

Signed-off-by: Luming Yu <luming.yu@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>

Remove "noreplacement" kernel command line option.

It is no longer valid to not replace instructions, since we depend on
different behaviour depending on CPU capabilities.

If you need to limit the capabilities of the replacements (because the
boot CPU has features that non-boot CPU's do not have, for example), you
need to explicitly disable those capabilities that are not shared across
all CPU's.

For example, if your boot CPU has FXSR, but other CPU's in your system
do not, you need to use the "nofxsr" kernel command line, not disable
instruction replacement per se.

x86: use alternative instructions for fnsave/fxsave too

This one ends up using an inline asm format that claims to read memory
and then clobber it (rather than just write it directly), which made it
easier to use the existing "alternative_input()" infrastructure support.

Now the fxsave code matches the fxrstor.

Merge /pub/scm/linux/kernel/git/davem/net-2.6

[PKT_SCHED]: em_meta: Kill TCF_META_ID_{INDEV,SECURITY,TCVERDICT}

More unusable TCF_META_* match types that need to get eliminated
before 2.6.13 goes out the door.

Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Thomas Graf <tgraf@suug.ch>

x86: make restore_fpu() use alternative assembler instructions

It's really just a single instruction, conditional on whether the CPU
supports FXSR or not, so implement it as such instead of making it a
function that queries FXSR dynamically.

This means that the instruction just gets automatically rewritten to the
correct one at boot-time.

[NETFILTER]: Fix ip6t_LOG MAC format

I broke this in the patch that consolidated MAC logging.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Use correct byteorder in ICMP NAT

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Wait until all references to ip_conntrack_untracked are dropped on unload

Fixes a crash when unloading ip_conntrack.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Fix potential memory corruption in NAT code (aka memory NAT)

The portptr pointing to the port in the conntrack tuple is declared static,
which could result in memory corruption when two packets of the same
protocol are NATed at the same time and one conntrack goes away.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

[NETFILTER]: Fix deadlock in ip6_queue

Already fixed in ip_queue, ip6_queue was missed.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Fix up incorrect "unlikely()" on %gs reload in x86 __switch_to

These days %gs is normally the TLS segment, so it's no longer zero.  As
a result, we shouldn't just assume that %fs/%gs tend to be zero
together, but test them independently instead.

Also, fix setting of debug registers to use the "next" pointer instead
of "current".  It so happens that the scheduler will have set the new
current pointer before calling __switch_to(), but that's just an
implementation detail.

[PKT_SCHED]: Kill TCF_META_ID_REALDEV from meta ematch.

It won't exist any longer when we shrink the SKB in 2.6.14,
and we should kill this off before anyone in userspace starts
using it.

Signed-off-by: David S. Miller <davem@davemloft.net>
Acked-by: Thomas Graf <tgraf@suug.ch>

[PATCH] Serial: No need to check for priv != NULL in remove_one

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

[EQL]: Proper num_slaves decrement

Signed-off-by: Loic Le Loarer <loic.le-loarer+lk@polytechnique.org>
Signed-off-by: David S. Miller <davem@davemloft.net>