smt_snooze_delay was designed to  delay idle loop's nap entry
in the native idle code before it got  ported over to use as part of
the cpuidle framework.

A -ve value  assigned to smt_snooze_delay should result in
busy looping, in other words disabling the entry to nap state.

	- https://lists.ozlabs.org/pipermail/linuxppc-dev/2010-May/082450.html

This particular functionality can be achieved currently by
echo 1 > /sys/devices/system/cpu/cpu*/state1/disable
but it is broken when one assigns -ve value to  the smt_snooze_delay
variable either via sysfs entry or ppc64_cpu util.

This patch aims to fix this, by disabling nap state when smt_snooze_delay
variable is set to -ve value.
Signed-off-by: NDeepthi Dharwar <deepthi@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

This is a preparatory patch for the introduction of NT_SIGINFO elf note.

Make the location of compat_siginfo_t uniform across eight architectures
which have it.  Now it can be pulled in by including asm/compat.h or
linux/compat.h.

Most of the copies are verbatim.  compat_uid[32]_t had to be replaced by
__compat_uid[32]_t.  compat_uptr_t had to be moved up before
compat_siginfo_t in asm/compat.h on a several architectures (tile already
had it moved up).  compat_sigval_t had to be relocated from linux/compat.h
to asm/compat.h.
Signed-off-by: NDenys Vlasenko <vda.linux@googlemail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Amerigo Wang <amwang@redhat.com>
Cc: "Jonathan M. Foote" <jmfoote@cert.org>
Cc: Roland McGrath <roland@hack.frob.com>
Cc: Pedro Alves <palves@redhat.com>
Cc: Fengguang Wu <fengguang.wu@intel.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

There are a number of issues in the recent IOMMU pools code:

- On a preempt kernel we might switch CPUs in the middle of building
  a scatter gather list. When this happens the handle hint passed in
  no longer falls within the local CPU's pool. Check for this and
  fall back to the pool hint.

- We were missing a spin_unlock/spin_lock in one spot where we
  switch pools.

- We need to provide locking around dart_tlb_invalidate_all and
  dart_tlb_invalidate_one now that the global lock is gone.
Reported-by: NAlexander Graf <agraf@suse.de>
Signed-off-by: NAnton Blanchard <anton@samba.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
CC: <stable@kernel.org> [v3.6]

This function is used by sparc, powerpc and arm64 for compat support.
The patch adds a generic implementation which calls do_sendfile()
directly and avoids set_fs().

The sparc architecture has wrappers for the sign extensions while
powerpc relies on the compiler to do the this. The patch adds wrappers
for powerpc to handle the u32->int type conversion.

compat_sys_sendfile64() can be replaced by a sys_sendfile() call since
compat_loff_t has the same size as off_t on a 64-bit system.

On powerpc, the patch also changes the 64-bit sendfile call from
sys_sendile64 to sys_sendfile.
Signed-off-by: NCatalin Marinas <catalin.marinas@arm.com>
Acked-by: NDavid S. Miller <davem@davemloft.net>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

This include is no longer needed.
(seems to be a leftover from try_to_freeze())
Signed-off-by: NRichard Weinberger <richard@nod.at>
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

the only non-obvious part is that current_pt_regs() is really needed
here - task_pt_regs() is NULL for kernel threads; it's OK for ptrace
uses (the thing task_pt_regs() is intended for), but not for us.
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

... and get rid of in-kernel syscalls in kernel_thread()
Signed-off-by: NAl Viro <viro@zeniv.linux.org.uk>

In commit 407821a3 we assigned a poison value to the paca->data_offset.

Unfortunately with CONFIG_LOCK_STAT=y lockdep will read & write to percpu
data very early in boot, prior to us initialising the percpu areas,
leading to a crash.

We have been getting away with this because the data_offset was previously
set to zero. This causes lockdep to read & write to the initial copy of
the percpu variables, which are discarded later in boot.

Although that is "fishy", it does work, and for lock statistics it is no
big deal to discard the counts from early boot.

So set the paca->data_offset = 0 for the boot cpu paca only.
Reported-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NMichael Ellerman <michael@ellerman.id.au>
Tested-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Move the code that finds out to which context we account the
cputime into generic layer.

Archs that consider the whole time spent in the idle task as idle
time (ia64, powerpc) can rely on the generic vtime_account()
and implement vtime_account_system() and vtime_account_idle(),
letting the generic code to decide when to call which API.

Archs that have their own meaning of idle time, such as s390
that only considers the time spent in CPU low power mode as idle
time, can just override vtime_account().
Signed-off-by: NFrederic Weisbecker <fweisbec@gmail.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>

Use a naming based on vtime as a prefix for virtual based
cputime accounting APIs:

- account_system_vtime() -> vtime_account()
- account_switch_vtime() -> vtime_task_switch()

It makes it easier to allow for further declension such
as vtime_account_system(), vtime_account_idle(), ... if we
want to find out the context we account to from generic code.

This also make it better to know on which subsystem these APIs
refer to.
Signed-off-by: NFrederic Weisbecker <fweisbec@gmail.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Peter Zijlstra <peterz@infradead.org>

To help migrate archtectures over to the new update_vsyscall method,
redfine CONFIG_GENERIC_TIME_VSYSCALL as CONFIG_GENERIC_TIME_VSYSCALL_OLD

Cc: Tony Luck <tony.luck@intel.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Turner <pjt@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: NJohn Stultz <john.stultz@linaro.org>

Since users will need to include timekeeper_internal.h, move
update_vsyscall definitions to timekeeper_internal.h.

Cc: Tony Luck <tony.luck@intel.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Turner <pjt@google.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Richard Cochran <richardcochran@gmail.com>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: NJohn Stultz <john.stultz@linaro.org>

During suspend, all interrupts including IPI will be disabled. In this case,
the suspend process will hang in SMP. To prevent this, pass the flag
IRQF_NO_SUSPEND when requesting IPI irq.
Signed-off-by: NZhao Chenhui <chenhui.zhao@freescale.com>
Signed-off-by: NLi Yang <leoli@freescale.com>
Acked-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

We can't emulate stwu since that may corrupt current exception stack.
So we will have to do real store operation in the exception return code.

Firstly we'll allocate a trampoline exception frame below the kprobed
function stack and copy the current exception frame to the trampoline.
Then we can do this real store operation to implement 'stwu', and reroute
the trampoline frame to r1 to complete this exception migration.
Signed-off-by: NTiejun Chen <tiejun.chen@windriver.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

There are a few tracepoints in the interrupt code path, which is before
irq_enter(), or after irq_exit(), like
trace_irq_entry()/trace_irq_exit() in do_IRQ(),
trace_timer_interrupt_entry()/trace_timer_interrupt_exit() in
timer_interrupt().

If the interrupt is from idle(), and because tracepoint contains RCU
read-side critical section, we could see following suspicious RCU usage
reported:

[  145.127743] ===============================
[  145.127747] [ INFO: suspicious RCU usage. ]
[  145.127752] 3.6.0-rc3+ #1 Not tainted
[  145.127755] -------------------------------
[  145.127759] /root/.workdir/linux/arch/powerpc/include/asm/trace.h:33
suspicious rcu_dereference_check() usage!
[  145.127765]
[  145.127765] other info that might help us debug this:
[  145.127765]
[  145.127771]
[  145.127771] RCU used illegally from idle CPU!
[  145.127771] rcu_scheduler_active = 1, debug_locks = 0
[  145.127777] RCU used illegally from extended quiescent state!
[  145.127781] no locks held by swapper/0/0.
[  145.127785]
[  145.127785] stack backtrace:
[  145.127789] Call Trace:
[  145.127796] [c00000000108b530] [c000000000013c40] .show_stack
+0x70/0x1c0 (unreliable)
[  145.127806] [c00000000108b5e0]
[c0000000000f59d8] .lockdep_rcu_suspicious+0x118/0x150
[  145.127813] [c00000000108b680] [c00000000000fc58] .do_IRQ+0x498/0x500
[  145.127820] [c00000000108b750] [c000000000003950]
hardware_interrupt_common+0x150/0x180
[  145.127828] --- Exception: 501 at .plpar_hcall_norets+0x84/0xd4
[  145.127828]     LR = .check_and_cede_processor+0x38/0x70
[  145.127836] [c00000000108bab0] [c0000000000665dc] .shared_cede_loop
+0x5c/0x100
[  145.127844] [c00000000108bb70] [c000000000588ab0] .cpuidle_enter
+0x30/0x50
[  145.127850] [c00000000108bbe0]
[c000000000588b0c] .cpuidle_enter_state+0x3c/0xb0
[  145.127857] [c00000000108bc60] [c000000000589730] .cpuidle_idle_call
+0x150/0x6c0
[  145.127863] [c00000000108bd30] [c000000000058440] .pSeries_idle
+0x10/0x40
[  145.127870] [c00000000108bda0] [c00000000001683c] .cpu_idle
+0x18c/0x2d0
[  145.127876] [c00000000108be60] [c00000000000b434] .rest_init
+0x124/0x1b0
[  145.127884] [c00000000108bef0] [c0000000009d0d28] .start_kernel
+0x568/0x588
[  145.127890] [c00000000108bf90] [c000000000009660] .start_here_common
+0x20/0x40

This is because the RCU usage in interrupt context should be used in
area marked by rcu_irq_enter()/rcu_irq_exit(), called in
irq_enter()/irq_exit() respectively.

Move them into the irq_enter()/irq_exit() area to avoid the reporting.
Signed-off-by: NLi Zhong <zhong@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Increase max addressable range to 64TB. This is not tested on
real hardware yet.
Reviewed-by: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NAneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

On POWER6 and POWER7 if the input operand to an instruction is a
denormalised single precision binary floating point value we can take
a denormalisation exception where it's expected that the hypervisor
(HV=1) will fix up the inputs before the instruction is run.

This adds code to handle this denormalisation exception for POWER6 and
POWER7.

It also add a CONFIG_PPC_DENORMALISATION option and sets it in
pseries/ppc64_defconfig.

This is useful on bare metal systems only.  Based on patch from Milton
Miller.
Signed-off-by: NMichael Neuling <mikey@neuling.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Remove the dependency on PCI initialization for SWIOTLB initialization.
So that PCI can be initialized at proper time.

SWIOTLB is partly determined by PCI inbound/outbound map which is assigned
in PCI initialization. But swiotlb_init() should be done at the stage of
mem_init() which is much earlier than PCI initialization. So we reserve the
memory for SWIOTLB first and free it if not necessary.

All boards are converted to fit this change.
Signed-off-by: NJia Hongtao <B38951@freescale.com>
Signed-off-by: NLi Yang <leoli@freescale.com>
Acked-by: NTony Breeds <tony@bakeyournoodle.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

Added CPU_FTR_EMB_HV feature check for e5500.
Signed-off-by: NVarun Sethi <Varun.Sethi@freescale.com>
Signed-off-by: NMihai Caraman <mihai.caraman@freescale.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

For the 64 bit case separate out e5500 cpu_setup and cpu_restore functions.
The cpu_setup function (for the primary core) is passed the cpu_spec
pointer, which is not there in case of the cpu_restore function. Also, in
our case we will have to manipulate the CPU_FTR_EMB_HV flag on the primary
core.
Signed-off-by: NVarun Sethi <Varun.Sethi@freescale.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

Merge the 32 bit cpu setup code for e500mc/e5500 and define the
"cpu_restore" routine (for e5500/e6500) only for the 64 bit case. The
cpu_restore routine is used in the 64 bit case for setting up the secondary
cores.
Signed-off-by: NVarun Sethi <Varun.Sethi@freescale.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

Move the E.HV check and CPU_FTR_EMB_HV flag manipulation to the cpu setup
code.  Create a separate routine for E.HV ivors setup.
Signed-off-by: NVarun Sethi <Varun.Sethi@freescale.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

Add support to disable and re-enable individual cores at runtime on
MPC85xx/QorIQ SMP machines. Currently support e500v1/e500v2 core.

MPC85xx machines use ePAPR spin-table in boot page for CPU kick-off.  This
patch uses the boot page from bootloader to boot core at runtime.  It
supports 32-bit and 36-bit physical address.
Signed-off-by: NLi Yang <leoli@freescale.com>
Signed-off-by: NJin Qing <b24347@freescale.com>
Signed-off-by: NZhao Chenhui <chenhui.zhao@freescale.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

In the case of cpu hotplug, the cpu_state should be set to CPU_UP_PREPARE
when kicking cpu.  Otherwise, the cpu_state is always CPU_DEAD after
calling generic_set_cpu_dead(), which makes the delay in generic_cpu_die()
not happen.
Signed-off-by: NZhao Chenhui <chenhui.zhao@freescale.com>
Signed-off-by: NKumar Gala <galak@kernel.crashing.org>

This patch implements pcibios_window_alignment() so powerpc platforms can
force P2P bridge windows to be at larger alignments than the PCI spec
requires.

[bhelgaas: changelog]
Signed-off-by: NGavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: NBjorn Helgaas <bhelgaas@google.com>

Currently we mark the DABRX to interrupt on all matches
(hypervisor/kernel/user and then filter in software.  We can be a lot
smarter now that we can set the DABRX dynamically.

This sets the DABRX based on the flags passed by the user.
Signed-off-by: NMichael Neuling <mikey@neuling.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Rework set_dabr to take a DABRX value as well.

Both the pseries and PS3 hypervisors do some checks on the DABRX
values that are passed in the hcall.  This patch stops bogus values
from being passed to hypervisor.  Also, in the case where we are
clearing the breakpoint, where DABR and DABRX are zero, we modify the
DABRX value to make it valid so that the hcall won't fail.
Signed-off-by: NMichael Neuling <mikey@neuling.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

The idea comes from Benjamin Herrenschmidt. The eeh cache helps
fetching the pci device according to the given I/O address. Since
the eeh cache is serving for eeh, it's reasonable for eeh cache
to trace eeh device except pci device.

The patch make eeh cache to trace eeh device. Also, the major
eeh entry function eeh_dn_check_failure has been renamed to
eeh_dev_check_failure since it will take eeh device as input
parameter.
Signed-off-by: NGavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Currently, we have 3 phases for EEH initialization on pSeries platform.
All of them are done through builtin functions: platform initialization,
EEH device creation, and EEH subsystem enablement. All of them are done
no later than ppc_md.setup_arch. That means that the slab/slub isn't ready
yet, so we have to allocate memory chunks on basis of PAGE_SIZE for those
dynamically created EEH devices. That's pretty expensive.

In order to utilize slab/slub for memory allocation, we have to move the EEH
initialization functions around, but all of them should be called after slab
is ready.
Signed-off-by: NGavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

It's possible for the cpu_possible_mask to change between the time we
initialise the pacas and the time we setup per_cpu areas.

Obviously impossible cpus shouldn't ever be running, but stranger things
have happened. So be paranoid and initialise data_offset with a poison
value in case we don't set it up later.

Based on a patch from Anton Blanchard.
Signed-off-by: NMichael Ellerman <michael@ellerman.id.au>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Change bp_info to info to be consistent with the rest of this file.
Signed-off-by: NMichael Neuling <mikey@neuling.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

The powerpc kernel doesn't export the memory limit enforced by 'mem='
kernel parameter. This is required for building the ELF header in
kexec-tools to limit the vmcore to capture only the used memory. On
powerpc the kexec-tools depends on the device-tree for memory related
information, unlike /proc/iomem on the x86.

Without this information, the kexec-tools assumes the entire System
RAM and vmcore creates an unnecessarily larger dump.

This patch exports the memory limit, if present, via
chosen/linux,memory-limit
property, so that the vmcore can be limited to the memory limit.

The prom_init seems to export this value in the same node. But doesn't
really
appear there.  Also the memory_limit gets adjusted with the processing of
crashkernel= parameter. This patch makes sure we get the actual limit.

The kexec-tools will use the value to limit the 'end' of the memory
regions.

Tested this patch on ppc64 and ppc32(ppc440) with a kexec-tools
patch by Mahesh.
Signed-off-by: NSuzuki K. Poulose <suzuki@in.ibm.com>
Tested-by: NMahesh J. Salgaonkar <mahesh@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

There are some device-tree nodes, whose values are of type phys_addr_t.
The phys_addr_t is variable sized based on the CONFIG_PHSY_T_64BIT.

Change these to a fixed unsigned long long for consistency.

This patch does the change only for memory_limit.

The following is a list of such variables which need the change:

 1) kernel_end, crashk_size - in arch/powerpc/kernel/machine_kexec.c

 2) (struct resource *)crashk_res.start - We could export a local static
    variable from machine_kexec.c.

Changing the above values might break the kexec-tools. So, I will
fix kexec-tools first to handle the different sized values and then change
 the above.
Suggested-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: NSuzuki K. Poulose <suzuki@in.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

When PCI probe flag PCI_REASSIGN_ALL_RSRC has been passed into PCI
core, it's hoped that all resources to be reassigned by PCI core.
As to particular P2P (PCI-to-PCI) bridge, the size of the corresponding
BAR (I/O, MMIO, prefetchable MMIO) is calculated by the resources
required by the PCI devices behind the P2P bridge. That means that
the information like start/end address retrieved from the hardware
registers of the P2P bridge is meainingless in the case. However,
we still count that in and the BARs might have been configured by
firmware with non-zero size. That leads to space waste.

The patch explicitly sets the size of P2P bridge BARs to zero in
case that resource reassignment is expected with PCI probe flag
PCI_REASSIGN_ALL_RSRC. In the result, it will save overall resource
required by the system without waste.
Signed-off-by: NGavin Shan <shangw@linux.vnet.ibm.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Critical exception on 64-bit booke uses user-visible SPRG3 as scratch.
Restore VDSO information in SPRG3 on exception prolog.

Use a common sprg3 field in PACA for all powerpc64 architectures.
Signed-off-by: NMihai Caraman <mihai.caraman@freescale.com>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

We have been observing hangs, both of KVM guest vcpu tasks and more
generally, where a process that is woken doesn't properly wake up and
continue to run, but instead sticks in TASK_WAKING state.  This
happens because the update of rq->wake_list in ttwu_queue_remote()
is not ordered with the update of ipi_message in
smp_muxed_ipi_message_pass(), and the reading of rq->wake_list in
scheduler_ipi() is not ordered with the reading of ipi_message in
smp_ipi_demux().  Thus it is possible for the IPI receiver not to see
the updated rq->wake_list and therefore conclude that there is nothing
for it to do.

In order to make sure that anything done before smp_send_reschedule()
is ordered before anything done in the resulting call to scheduler_ipi(),
this adds barriers in smp_muxed_message_pass() and smp_ipi_demux().
The barrier in smp_muxed_message_pass() is a full barrier to ensure that
there is a full ordering between the smp_send_reschedule() caller and
scheduler_ipi().  In smp_ipi_demux(), we use xchg() rather than
xchg_local() because xchg() includes release and acquire barriers.
Using xchg() rather than xchg_local() makes sense given that
ipi_message is not just accessed locally.

This moves the barrier between setting the message and calling the
cause_ipi() function into the individual cause_ipi implementations.
Most of them -- those that used outb, out_8 or similar -- already had
a full barrier because out_8 etc. include a sync before the MMIO
store.  This adds an explicit barrier in the two remaining cases.

These changes made no measurable difference to the speed of IPIs as
measured using a simple ping-pong latency test across two CPUs on
different cores of a POWER7 machine.

The analysis of the reason why processes were not waking up properly
is due to Milton Miller.

Cc: stable@vger.kernel.org # v3.0+
Reported-by: NMilton Miller <miltonm@bga.com>
Signed-off-by: NPaul Mackerras <paulus@samba.org>
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

During a context switch we always restore the per thread DSCR value.
If we aren't doing explicit DSCR management
(ie thread.dscr_inherit == 0) and the default DSCR changed while
the process has been sleeping we end up with the wrong value.

Check thread.dscr_inherit and select the default DSCR or per thread
DSCR as required.

This was found with the following test case, when running with
more threads than CPUs (ie forcing context switching):

http://ozlabs.org/~anton/junkcode/dscr_default_test.c

With the four patches applied I can run a combination of all
test cases successfully at the same time:

http://ozlabs.org/~anton/junkcode/dscr_default_test.c
http://ozlabs.org/~anton/junkcode/dscr_explicit_test.c
http://ozlabs.org/~anton/junkcode/dscr_inherit_test.cSigned-off-by: NAnton Blanchard <anton@samba.org>
Cc: <stable@kernel.org> # 3.0+
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

If the default DSCR is non zero we set thread.dscr_inherit in
copy_thread() meaning the new thread and all its children will ignore
future updates to the default DSCR. This is not intended and is
a change in behaviour that a number of our users have hit.

We just need to inherit thread.dscr and thread.dscr_inherit from
the parent which ends up being much simpler.

This was found with the following test case:

http://ozlabs.org/~anton/junkcode/dscr_default_test.cSigned-off-by: NAnton Blanchard <anton@samba.org>
Cc: <stable@kernel.org> # 3.0+
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

When we update the DSCR either via emulation of mtspr(DSCR) or via
a change to dscr_default in sysfs we don't update thread.dscr.
We will eventually update it at context switch time but there is
a period where thread.dscr is incorrect.

If we fork at this point we will copy the old value of thread.dscr
into the child. To avoid this, always keep thread.dscr in sync with
reality.

This issue was found with the following testcase:

http://ozlabs.org/~anton/junkcode/dscr_inherit_test.cSigned-off-by: NAnton Blanchard <anton@samba.org>
Cc: <stable@kernel.org> # 3.0+
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>

Writing to dscr_default in sysfs doesn't actually change the DSCR -
we rely on a context switch on each CPU to do the work. There is no
guarantee we will get a context switch in a reasonable amount of time
so fire off an IPI to force an immediate change.

This issue was found with the following test case:

http://ozlabs.org/~anton/junkcode/dscr_explicit_test.cSigned-off-by: NAnton Blanchard <anton@samba.org>
Cc: <stable@kernel.org> # 3.0+
Signed-off-by: NBenjamin Herrenschmidt <benh@kernel.crashing.org>