Andreas Herrmann reported that 7d6b4670 ("x86, NUMA: Fix fakenuma
boot failure") causes certain physical NUMA topologies (for example
AMD Magny-Cours) to move sibling cpus to a single node when in reality
they are in separate domains.

This may result in some nodes being completely void of cpus, which
doesn't accurately represent the correct topology. The system will
boot, but will have suboptimal NUMA performance.

This commit was intended as a fix for NUMA emulation, but should
not cause a regression for real NUMA machines as a side effect.

( There will be a separate fix for the numa-debug code, which
  will not affect physical topologies. )
Reported-by: NAndreas Herrmann <herrmann.der.user@googlemail.com>
Signed-off-by: NDavid Rientjes <rientjes@google.com>
Acked-by: NKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.00.1104201918110.12634@chino.kir.corp.google.comSigned-off-by: NIngo Molnar <mingo@elte.hu>

Currently, numa=fake boot parameter is broken. If it's used,
kernel may panic due to devide by zero error depending on CPU
configuration

Call Trace:
 [<ffffffff8104ad4c>] find_busiest_group+0x38c/0xd30
 [<ffffffff81086aff>] ? local_clock+0x6f/0x80
 [<ffffffff81050533>] load_balance+0xa3/0x600
 [<ffffffff81050f53>] idle_balance+0xf3/0x180
 [<ffffffff81550092>] schedule+0x722/0x7d0
 [<ffffffff81550538>] ? wait_for_common+0x128/0x190
 [<ffffffff81550a65>] schedule_timeout+0x265/0x320
 [<ffffffff81095815>] ? lock_release_holdtime+0x35/0x1a0
 [<ffffffff81550538>] ? wait_for_common+0x128/0x190
 [<ffffffff8109bb6c>] ? __lock_release+0x9c/0x1d0
 [<ffffffff815534e0>] ? _raw_spin_unlock_irq+0x30/0x40
 [<ffffffff815534e0>] ? _raw_spin_unlock_irq+0x30/0x40
 [<ffffffff81550540>] wait_for_common+0x130/0x190
 [<ffffffff81051920>] ? try_to_wake_up+0x510/0x510
 [<ffffffff8155067d>] wait_for_completion+0x1d/0x20
 [<ffffffff8107f36c>] kthread_create_on_node+0xac/0x150
 [<ffffffff81077bb0>] ? process_scheduled_works+0x40/0x40
 [<ffffffff8155045f>] ? wait_for_common+0x4f/0x190
 [<ffffffff8107a283>] __alloc_workqueue_key+0x1a3/0x590
 [<ffffffff81e0cce2>] cpuset_init_smp+0x6b/0x7b
 [<ffffffff81df3d07>] kernel_init+0xc3/0x182
 [<ffffffff8155d5e4>] kernel_thread_helper+0x4/0x10
 [<ffffffff81553cd4>] ? retint_restore_args+0x13/0x13
 [<ffffffff81df3c44>] ? start_kernel+0x400/0x400
 [<ffffffff8155d5e0>] ? gs_change+0x13/0x13

The divede by zero is caused by the following line,
group->cpu_power==0:

 kernel/sched_fair.c::update_sg_lb_stats()
        /* Adjust by relative CPU power of the group */
        sgs->avg_load = (sgs->group_load * SCHED_LOAD_SCALE) / group->cpu_power;

This regression was caused by commit e23bba60 ("x86-64, NUMA: Unify
emulated distance mapping") because it changes cpu -> node
mapping in the process of dropping fake_physnodes().

  old) all cpus are assinged node 0
  now) cpus are assigned round robin
       (the logic is implemented by numa_init_array())

  Note: The change in behavior only happens if the system doesn't
        have neither ACPI SRAT table nor AMD northbridge NUMA
	information.

Round robin assignment doesn't work because init_numa_sched_groups_power()
assumes all logical cpus in the same physical cpu share the same node
(then it only accounts for group_first_cpu()), and the simple round robin
breaks the above assumption.

Thus, this patch implements a reassignment of node-ids if buggy firmware
or numa emulation makes wrong cpu node map. Tt enforce all logical cpus
in the same physical cpu share the same node.
Signed-off-by: NKOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: NTejun Heo <tj@kernel.org>
Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Brian Gerst <brgerst@gmail.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Cc: Shaohui Zheng <shaohui.zheng@intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: H. Peter Anvin <hpa@linux.intel.com>
Link: http://lkml.kernel.org/r/20110415203928.1303.A69D9226@jp.fujitsu.comSigned-off-by: NIngo Molnar <mingo@elte.hu>

Add this_cpu_has() which determines if the current cpu has a certain
ability using a segment prefix and a bit test operation.

For that we need to add bit operations to x86s percpu.h.

Many uses of cpu_has use a pointer passed to a function to determine
the current flags. That is no longer necessary after this patch.

However, this patch only converts the straightforward cases where
cpu_has is used with this_cpu_ptr. The rest is work for later.

-tj: Rolled up patch to add x86_ prefix and use percpu_read() instead
     of percpu_read_stable().
Signed-off-by: NChristoph Lameter <cl@linux.com>
Acked-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NTejun Heo <tj@kernel.org>

Currently arch_disable_smp_support() on x86 disables only the
support for the IOAPIC and is also compiled in if SMP-support is
not.

Therefore this function is renamed to disable_ioapic_support(),
which meets its purpose and is only compiled in the kernel
when IOAPIC support is also.

A new arch_disable_smp_support() is created in smpboot.c,
which calls disable_ioapic_support() and gets only compiled
in the kernel when SMP support is also.
Signed-off-by: NHenrik Kretzschmar <henne@nachtwindheim.de>
LKML-Reference: <1298385487-4708-3-git-send-email-henne@nachtwindheim.de>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Common infrastructure for low memory trampolines.  This code installs
the trampolines permanently in low memory very early.  It also permits
multiple pieces of code to be used for this purpose.

This code also introduces a standard infrastructure for computing
symbol addresses in the trampoline code.

The only change to the actual SMP trampolines themselves is that the
64-bit trampoline has been made reusable -- the previous version would
overwrite the code with a status variable; this moves the status
variable to a separate location.
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>
LKML-Reference: <4D5DFBE4.7090104@intel.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Matthieu Castet <castet.matthieu@free.fr>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>

Additionally doing things conditionally upon smp_processor_id()
being zero is generally a bad idea, as this means CPU 0 cannot
be offlined and brought back online later again.

While there may be other places where this is done, I think adding
more of those should be avoided so that some day SMP can really
become "symmetrical".
Signed-off-by: NJan Beulich <jbeulich@novell.com>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
LKML-Reference: <4D525C7E0200007800030EE1@vpn.id2.novell.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Since checkin ebba638a we call
verify_cpu even in 32-bit mode.  Unfortunately, calling a function
means using the stack, and the stack pointer was not initialized in
the 32-bit setup code!  This code initializes the stack pointer, and
simplifies the interface slightly since it is easier to rely on just a
pointer value rather than a descriptor; we need to have different
values for the segment register anyway.

This retains start_stack as a virtual address, even though a physical
address would be more convenient for 32 bits; the 64-bit code wants
the other way around...
Reported-by: NMatthieu Castet <castet.matthieu@free.fr>
LKML-Reference: <4D41E86D.8060205@free.fr>
Tested-by: NKees Cook <kees.cook@canonical.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

x86_32 has been managing node_to_cpumask_map explicitly from
map_cpu_to_node() and friends in a rather ugly way.  With
previous changes, it's now possible to share the code with
64bit.

* When CONFIG_NUMA_EMU is disabled, numa_add/remove_cpu() are
  implemented in numa.c and shared by 32 and 64bit.  CONFIG_NUMA_EMU
  versions still live in numa_64.c.

  NUMA_EMU's dependency on 64bit is planned to be removed and the
  above should go away together.

* identify_cpu() now calls numa_add_cpu() for 32bit too.  This
  makes the explicit mask management from map_cpu_to_node() unnecessary.

* The whole x86_32 specific map_cpu_to_node() chunk is no longer
  necessary.  Dropped.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NPekka Enberg <penberg@kernel.org>
Cc: eric.dumazet@gmail.com
Cc: yinghai@kernel.org
Cc: brgerst@gmail.com
Cc: gorcunov@gmail.com
Cc: shaohui.zheng@intel.com
Cc: rientjes@google.com
LKML-Reference: <1295789862-25482-16-git-send-email-tj@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Cc: David Rientjes <rientjes@google.com>
Cc: Shaohui Zheng <shaohui.zheng@intel.com>

Unlike 64bit, 32bit has been using its own cpu_to_node_map[] for
CPU -> NUMA node mapping.  Replace it with early_percpu variable
x86_cpu_to_node_map and share the mapping code with 64bit.

* USE_PERCPU_NUMA_NODE_ID is now enabled for 32bit too.

* x86_cpu_to_node_map and numa_set/clear_node() are moved from
  numa_64 to numa.  For now, on 32bit, x86_cpu_to_node_map is initialized
  with 0 instead of NUMA_NO_NODE.  This is to avoid introducing unexpected
  behavior change and will be updated once init path is unified.

* srat_detect_node() is now enabled for x86_32 too.  It calls
  numa_set_node() and initializes the mapping making explicit
  cpu_to_node_map[] updates from map/unmap_cpu_to_node() unnecessary.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: eric.dumazet@gmail.com
Cc: yinghai@kernel.org
Cc: brgerst@gmail.com
Cc: gorcunov@gmail.com
Cc: penberg@kernel.org
Cc: shaohui.zheng@intel.com
Cc: rientjes@google.com
LKML-Reference: <1295789862-25482-15-git-send-email-tj@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Cc: David Rientjes <rientjes@google.com>

The mapping between cpu/apicid and node is done via
apicid_to_node[] on 64bit and apicid_2_node[] +
apic->x86_32_numa_cpu_node() on 32bit. This difference makes it
difficult to further unify 32 and 64bit NUMA handling.

This patch unifies it by replacing both apicid_to_node[] and
apicid_2_node[] with __apicid_to_node[] array, which is accessed
by two accessors - set_apicid_to_node() and numa_cpu_node().  On
64bit, numa_cpu_node() always consults __apicid_to_node[]
directly while 32bit goes through apic->numa_cpu_node() method
to allow apic implementations to override it.

srat_detect_node() for amd cpus contains workaround for broken
NUMA configuration which assumes relationship between APIC ID,
HT node ID and NUMA topology.  Leave it to access
__apicid_to_node[] directly as mapping through CPU might result
in undesirable behavior change.  The comment is reformatted and
updated to note the ugliness.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NPekka Enberg <penberg@kernel.org>
Cc: eric.dumazet@gmail.com
Cc: yinghai@kernel.org
Cc: brgerst@gmail.com
Cc: gorcunov@gmail.com
Cc: shaohui.zheng@intel.com
Cc: rientjes@google.com
LKML-Reference: <1295789862-25482-14-git-send-email-tj@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>
Cc: David Rientjes <rientjes@google.com>

apic->apicid_to_node() is 32bit specific apic operation which
determines NUMA node for a CPU.  Depending on the APIC
implementation, it can be easier to determine NUMA node from
either physical or logical apicid.  Currently,
->apicid_to_node() takes @logical_apicid and calls
hard_smp_processor_id() if the physical apicid is needed.

This prevents NUMA mapping from being queried from a different
CPU, which in turn makes it impossible to initialize NUMA
mapping before SMP bringup.

This patch replaces apic->apicid_to_node() with
->x86_32_numa_cpu_node() which takes @cpu, from which both
logical and physical apicids can easily be determined.  While at
it, drop duplicate implementations from bigsmp_32 and summit_32,
and use the default one.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NPekka Enberg <penberg@kernel.org>
Cc: eric.dumazet@gmail.com
Cc: yinghai@kernel.org
Cc: brgerst@gmail.com
Cc: gorcunov@gmail.com
Cc: shaohui.zheng@intel.com
Cc: rientjes@google.com
LKML-Reference: <1295789862-25482-13-git-send-email-tj@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Currently, cpu -> logical apic id translation is done by
apic->cpu_to_logical_apicid() callback which may or may not use
x86_cpu_to_logical_apicid.  This is unnecessary as it should
always equal logical_smp_processor_id() which is known early
during CPU bring up.

Initialize x86_cpu_to_logical_apicid after apic->init_apic_ldr()
in setup_local_APIC() and always use x86_cpu_to_logical_apicid
for cpu -> logical apic id mapping.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: eric.dumazet@gmail.com
Cc: yinghai@kernel.org
Cc: brgerst@gmail.com
Cc: gorcunov@gmail.com
Cc: penberg@kernel.org
Cc: shaohui.zheng@intel.com
Cc: rientjes@google.com
LKML-Reference: <1295789862-25482-6-git-send-email-tj@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Unlike x86_64, on x86_32, the mapping from cpu to logical apicid
may vary depending on apic in use.  cpu_2_logical_apicid[] array
is used for this mapping.  Replace it with early percpu variable
x86_cpu_to_logical_apicid to make it better aligned with other
mappings.
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: eric.dumazet@gmail.com
Cc: yinghai@kernel.org
Cc: brgerst@gmail.com
Cc: gorcunov@gmail.com
Cc: penberg@kernel.org
Cc: shaohui.zheng@intel.com
Cc: rientjes@google.com
LKML-Reference: <1295789862-25482-5-git-send-email-tj@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Commit 56d91f13 (x86, acpi: Add MAX_LOCAL_APIC for 32bit) added
MAX_LOCAL_APIC for x86_32 but didn't replace MAX_APICID users
with it. Convert MAX_APICID users to MAX_LOCAL_APIC and drop
MAX_APICID.
Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NPekka Enberg <penberg@kernel.org>
Acked-by: NYinghai Lu <yinghai@kernel.org>
Cc: eric.dumazet@gmail.com
Cc: yinghai@kernel.org
Cc: brgerst@gmail.com
Cc: gorcunov@gmail.com
Cc: shaohui.zheng@intel.com
Cc: rientjes@google.com
LKML-Reference: <1295789862-25482-3-git-send-email-tj@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Signed-off-by: NTejun Heo <tj@kernel.org>
Reviewed-by: NPekka Enberg <penberg@kernel.org>
Acked-by: NYinghai Lu <yinghai@kernel.org>
Cc: eric.dumazet@gmail.com
Cc: yinghai@kernel.org
Cc: brgerst@gmail.com
Cc: gorcunov@gmail.com
Cc: shaohui.zheng@intel.com
Cc: rientjes@google.com
LKML-Reference: <1295789862-25482-2-git-send-email-tj@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

smp_store_cpu_info(0) will do that.
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
LKML-Reference: <4D3A16F2.5090902@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

cpu_info is already with per_cpu, We can take llc_shared_map out
of cpu_info, and declare it as per_cpu variable directly.

So later referencing could be simple and directly instead of
diving to find cpu_info at first.

Also could make smp_store_cpu_info() much simple to avoid to do
save and restore trick.
Signed-off-by: NYinghai Lu <yinghai@kernel.org>
Cc: Hans Rosenfeld <hans.rosenfeld@amd.com>
Cc: Alok N Kataria <akataria@vmware.com>
Cc: Stephen Hemminger <shemminger@vyatta.com>
Cc: Hans J. Koch <hjk@linutronix.de>
Cc: Tejun Heo <tj@kernel.org>
Cc: Borislav Petkov <borislav.petkov@amd.com>
Cc: Andreas Herrmann <andreas.herrmann3@amd.com>
Cc: Robert Richter <robert.richter@amd.com>
Cc: Suresh Siddha <suresh.b.siddha@intel.com>
LKML-Reference: <4D3A16E8.5020608@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

On multi-node CPUs we don't need the socket wide compute unit ID
but the node-wide compute unit ID. Thus we need to normalize the
value. This is similar to what we do with cpu_core_id.

A compute unit is then identified by physical_package_id,
node_id, and compute_unit_id.
Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
LKML-Reference: <1295881543-572552-2-git-send-email-hans.rosenfeld@amd.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

ea530692 made a CPU use monitor/mwait
when offline. This is not the optimal choice for AMD wrt to powersavings
and we'd prefer our cores to halt (i.e. enter C1) instead. For this, the
same selection whether to use monitor/mwait has to be used as when we
select the idle routine for the machine.

With this patch, offlining cores 1-5 on a X6 machine allows core0 to
boost again.

[ hpa: putting this in urgent since it is a (power) regression fix ]
Reported-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
Cc: stable@kernel.org # 37.x
Cc: H. Peter Anvin <hpa@linux.intel.com>
Cc: Arjan van de Ven <arjan@linux.intel.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Venkatesh Pallipadi <venki@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.hl>
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
LKML-Reference: <1295534572-10730-1-git-send-email-bp@amd64.org>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Fix sparse warning for non-ANSI function declaration:

  arch/x86/kernel/smpboot.c:100:30: warning: non-ANSI function declaration of function 'cpu_hotplug_driver_lock'
  arch/x86/kernel/smpboot.c:105:32: warning: non-ANSI function declaration of function 'cpu_hotplug_driver_unlock'
Signed-off-by: NRandy Dunlap <randy.dunlap@oracle.com>
LKML-Reference: <20110108195914.95d366ea.randy.dunlap@oracle.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Replace all uses of current_cpu_data with this_cpu operations on the
per cpu structure cpu_info.  The scala accesses are replaced with the
matching this_cpu ops which results in smaller and more efficient
code.

In the long run, it might be a good idea to remove cpu_data() macro
too and use per_cpu macro directly.

tj: updated description

Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Acked-by: NH. Peter Anvin <hpa@zytor.com>
Acked-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NChristoph Lameter <cl@linux.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

Go through x86 code and replace __get_cpu_var and get_cpu_var
instances that refer to a scalar and are not used for address
determinations.

Cc: Yinghai Lu <yinghai@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Acked-by: NTejun Heo <tj@kernel.org>
Acked-by: N"H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: NChristoph Lameter <cl@linux.com>
Signed-off-by: NTejun Heo <tj@kernel.org>

During suspend, we disable all the non boot cpus. And during resume we bring
them all back again. So no need to do alternatives_smp_switch() in between.

On my core 2 based laptop, this speeds up the suspend path by 15msec and the
resume path by 5 msec (suspend/resume speed up differences can be attributed
to the different P-states that the cpu is in during suspend/resume).
Signed-off-by: NSuresh Siddha <suresh.b.siddha@intel.com>
LKML-Reference: <1290557500.4946.8.camel@sbsiddha-MOBL3.sc.intel.com>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

When booting up a CPU set the various topology masks before
calling the CPU_STARTING notifier. This way the notifier
can actually use the masks.

This is needed for a perf change.
Signed-off-by: NAndi Kleen <ak@linux.intel.com>
Signed-off-by: NPeter Zijlstra <a.p.zijlstra@chello.nl>
LKML-Reference: <1290077254-12165-2-git-send-email-andi@firstfloor.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Now that the bulk of the old nmi_watchdog is gone, remove all
the stub variables and hooks associated with it.

This touches lots of files mainly because of how the io_apic
nmi_watchdog was implemented.  Now that the io_apic nmi_watchdog
is forever gone, remove all its fingers.

Most of this code was not being exercised by virtue of
nmi_watchdog != NMI_IO_APIC, so there shouldn't be anything to
risky here.
Signed-off-by: NDon Zickus <dzickus@redhat.com>
Cc: fweisbec@gmail.com
Cc: gorcunov@openvz.org
LKML-Reference: <1289578944-28564-3-git-send-email-dzickus@redhat.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

The percpu allocator cannot handle alignments larger than one
page. Allocate the irq stacks seperately, and only keep the
pointers as percpu data.
Signed-off-by: NBrian Gerst <brgerst@gmail.com>
Acked-by: NLinus Torvalds <torvalds@linux-foundation.org>
Cc: tj@kernel.org
LKML-Reference: <1288158182-1753-1-git-send-email-brgerst@gmail.com>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

Silly though it is, completions and wait_queue_heads use foo_ONSTACK
(COMPLETION_INITIALIZER_ONSTACK, DECLARE_COMPLETION_ONSTACK,
__WAIT_QUEUE_HEAD_INIT_ONSTACK and DECLARE_WAIT_QUEUE_HEAD_ONSTACK) so I
guess workqueues should do the same thing.

s/INIT_WORK_ON_STACK/INIT_WORK_ONSTACK/
s/INIT_DELAYED_WORK_ON_STACK/INIT_DELAYED_WORK_ONSTACK/

Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: NTejun Heo <tj@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This patch adds an initial page table with low mappings used exclusively
for booting APs/resuming after ACPI suspend/machine restart. After this,
there's no need to add low mappings to swapper_pg_dir and zap them later
or create own swsusp PGD page solely for ACPI sleep needs - we have
initial_page_table for that.
Signed-off-by: NBorislav Petkov <bp@alien8.de>
LKML-Reference: <20101020070526.GA9588@liondog.tnic>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Signed-off-by: NThomas Gleixner <tglx@linutronix.de>
Reviewed-by: NIngo Molnar <mingo@elte.hu>

This information is vital for different load balancing policies.
Signed-off-by: NAndreas Herrmann <andreas.herrmann3@amd.com>
LKML-Reference: <20100930124156.GF20545@loge.amd.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

When we're using MWAIT for play_dead, explicitly CLFLUSH the cache
line before executing MONITOR.  This is a potential workaround for the
Xeon 7400 erratum AAI65 after having a spurious wakeup and returning
around the loop.  "Potential" here because it is not certain that that
erratum could actually trigger; however, the CLFLUSH should be
harmless.
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Acked-by: NVenkatesh Pallipadi <venki@google.com>
Cc: Asit Mallick <asit.k.mallick@intel.com>
Cc: Arjan van de Ven <arjan@linux.kernel.org>
Cc: Len Brown <lenb@kernel.org>

On processors with hyperthreading, when only one thread is offlined
the other thread can cause a spurious wakeup on the idled thread.  We
do not want to re-WBINVD when that happens.

Ideally, we should simply skip WBINVD unless we're the last thread on
a particular core to shut down, but there might be similar issues
elsewhere in the system.

Thus, revert to previous behavior of only WBINVD outside the loop.
Partly as a result, remove the mb()'s around it: they are not
necessary since wbinvd() is a serializing instruction, but they were
intended to make sure the compiler didn't do any funny loop
optimizations.
Reported-by: NAsit Mallick <asit.k.mallick@intel.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Cc: Arjan van de Ven <arjan@linux.kernel.org>
Cc: Len Brown <lenb@kernel.org>
Cc: Venkatesh Pallipadi <venki@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.hl>
LKML-Reference: <tip-ea530692@git.kernel.org>

The code in native_play_dead() has a number of problems:

1. We should use MWAIT when available, to put ourselves into a deeper
   sleep state.
2. We use the existence of CLFLUSH to determine if WBINVD is safe, but
   that is totally bogus -- WBINVD is 486+, whereas CLFLUSH is a much
   later addition.
3. We should do WBINVD inside the loop, just in case of something like
   setting an A bit on page tables.  Pointed out by Arjan van de Ven.

This code is based in part of a previous patch by Venki Pallipadi, but
unlike that patch this one keeps all the detection code local instead
of pre-caching a bunch of information.  We're shutting down the CPU;
there is absolutely no hurry.

This patch moves all the code to C and deletes the global
wbinvd_halt() which is broken anyway.
Originally-by: NVenkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>
Reviewed-by: NArjan van de Ven <arjan@linux.intel.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Venkatesh Pallipadi <venki@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.hl>
LKML-Reference: <20090522232230.162239000@intel.com>

Move enable_IR_x2apic() inside the default_setup_apic_routing(),
and for SMP platforms, move the default_setup_apic_routing() after
smp_sanity_check(). This cleans up the code that tries to avoid multiple
calls to default_setup_apic_routing() when smp_sanity_check() fails (which
goes through the APIC_init_uniprocessor() path).
Signed-off-by: NSuresh Siddha <suresh.b.siddha@intel.com>
LKML-Reference: <20100827181049.173087246@sbsiddha-MOBL3.sc.intel.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

With the recent innovations in CPU hardware acceleration technologies
from Intel and AMD, VMware ran a few experiments to compare these
techniques to guest paravirtualization technique on VMware's platform.
These hardware assisted virtualization techniques have outperformed the
performance benefits provided by VMI in most of the workloads. VMware
expects that these hardware features will be ubiquitous in a couple of
years, as a result, VMware has started a phased retirement of this
feature from the hypervisor.

Please note that VMI has always been an optimization and non-VMI kernels
still work fine on VMware's platform.
Latest versions of VMware's product which support VMI are,
Workstation 7.0 and VSphere 4.0 on ESX side, future maintainence
releases for these products will continue supporting VMI.

For more details about VMI retirement take a look at this,
http://blogs.vmware.com/guestosguide/2009/09/vmi-retirement.html

This feature removal was scheduled for 2.6.37 back in September 2009.
Signed-off-by: NAlok N Kataria <akataria@vmware.com>
LKML-Reference: <1282600151.19396.22.camel@ank32.eng.vmware.com>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

When testing cpu hotplug code on 32-bit we kept hitting the "CPU%d:
Stuck ??" message due to multiple cores concurrently accessing the
cpu_callin_mask, among others.

Since these codepaths are not protected from concurrent access due to
the fact that there's no sane reason for making an already complex
code unnecessarily more complex - we hit the issue only when insanely
switching cores off- and online - serialize hotplugging cores on the
sysfs level and be done with it.

[ v2.1: fix !HOTPLUG_CPU build ]

Cc: <stable@kernel.org>
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
LKML-Reference: <20100819181029.GC17171@aftab>
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

This patch fixes machine crashes which occur when heavily exercising the
CPU hotplug codepaths on a 32-bit kernel. These crashes are caused by
AMD Erratum 383 and result in a fatal machine check exception. Here's
the scenario:

1. On 32-bit, the swapper_pg_dir page table is used as the initial page
table for booting a secondary CPU.

2. To make this work, swapper_pg_dir needs a direct mapping of physical
memory in it (the low mappings). By adding those low, large page (2M)
mappings (PAE kernel), we create the necessary conditions for Erratum
383 to occur.

3. Other CPUs which do not participate in the off- and onlining game may
use swapper_pg_dir while the low mappings are present (when leave_mm is
called). For all steps below, the CPU referred to is a CPU that is using
swapper_pg_dir, and not the CPU which is being onlined.

4. The presence of the low mappings in swapper_pg_dir can result
in TLB entries for addresses below __PAGE_OFFSET to be established
speculatively. These TLB entries are marked global and large.

5. When the CPU with such TLB entry switches to another page table, this
TLB entry remains because it is global.

6. The process then generates an access to an address covered by the
above TLB entry but there is a permission mismatch - the TLB entry
covers a large global page not accessible to userspace.

7. Due to this permission mismatch a new 4kb, user TLB entry gets
established. Further, Erratum 383 provides for a small window of time
where both TLB entries are present. This results in an uncorrectable
machine check exception signalling a TLB multimatch which panics the
machine.

There are two ways to fix this issue:

        1. Always do a global TLB flush when a new cr3 is loaded and the
        old page table was swapper_pg_dir. I consider this a hack hard
        to understand and with performance implications

        2. Do not use swapper_pg_dir to boot secondary CPUs like 64-bit
        does.

This patch implements solution 2. It introduces a trampoline_pg_dir
which has the same layout as swapper_pg_dir with low_mappings. This page
table is used as the initial page table of the booting CPU. Later in the
bringup process, it switches to swapper_pg_dir and does a global TLB
flush. This fixes the crashes in our test cases.

-v2: switch to swapper_pg_dir right after entering start_secondary() so
that we are able to access percpu data which might not be mapped in the
trampoline page table.
Signed-off-by: NJoerg Roedel <joerg.roedel@amd.com>
LKML-Reference: <20100816123833.GB28147@aftab>
Signed-off-by: NBorislav Petkov <borislav.petkov@amd.com>
Signed-off-by: NH. Peter Anvin <hpa@zytor.com>

Workqueues are now initialized as part of the early_initcall().  So they
are available for use during cold boot process aswell.
Signed-off-by: NSuresh Siddha <suresh.b.siddha@intel.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Tejun Heo <tj@kernel.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Tony Luck <tony.luck@intel.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Use the stop machine context rather than IPI's to rendezvous all the cpus for
MTRR initialization that happens during cpu bringup or for MTRR modifications
during runtime.

This avoids deadlock scenario (reported by Prarit) like:

cpu A holds a read_lock (tasklist_lock for example) with irqs enabled
cpu B waits for the same lock with irqs disabled using write_lock_irq
cpu C doing set_mtrr() (during AP bringup for example), which will try to
rendezvous all the cpus using IPI's

This will result in C and A come to the rendezvous point and waiting
for B. B is stuck forever waiting for the lock and thus not
reaching the rendezvous point.

Using stop cpu (run in the process context of per cpu based keventd) to do
this rendezvous, avoids this deadlock scenario.

Also make sure all the cpu's are in the rendezvous handler before we proceed
with the local_irq_save() on each cpu. This lock step disabling irqs on all
the cpus will avoid other deadlock scenarios (for example involving
with the blocking smp_call_function's etc).

   [ This problem is very old. Marking -stable only for 2.6.35 as the
     stop_one_cpu_nowait() API is present only in 2.6.35. Any older
     kernel interested in this fix need to do some more work in backporting
     this patch. ]
Reported-by: NPrarit Bhargava <prarit@redhat.com>
Signed-off-by: NSuresh Siddha <suresh.b.siddha@intel.com>
LKML-Reference: <1280515602.2682.10.camel@sbsiddha-MOBL3.sc.intel.com>
Acked-by: NPrarit Bhargava <prarit@redhat.com>
Cc: stable@kernel.org	[2.6.35]
Signed-off-by: NH. Peter Anvin <hpa@linux.intel.com>

Define WQ_MAX_ACTIVE and create keventd with max_active set to half of
it which means that keventd now can process upto WQ_MAX_ACTIVE / 2 - 1
works concurrently.  Unless some combination can result in dependency
loop longer than max_active, deadlock won't happen and thus it's
unnecessary to check whether current_is_keventd() before trying to
schedule a work.  Kill current_is_keventd().

(Lockdep annotations are broken.  We need lock_map_acquire_read_norecurse())
Signed-off-by: NTejun Heo <tj@kernel.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Oleg Nesterov <oleg@redhat.com>