Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

Fix the smp code in s390 by using this latter form of callback registration.

Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

Fix the cacheinfo code in s390 by using this latter form of callback
registration.

Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

On 03/15/2014 12:40 AM, Christoffer Dall wrote:
> On Fri, Mar 14, 2014 at 11:13:29AM +0530, Srivatsa S. Bhat wrote:
>> On 03/13/2014 04:51 AM, Christoffer Dall wrote:
>>> On Tue, Mar 11, 2014 at 02:05:38AM +0530, Srivatsa S. Bhat wrote:
>>>> Subsystems that want to register CPU hotplug callbacks, as well as perform
>>>> initialization for the CPUs that are already online, often do it as shown
>>>> below:
>>>>
[...]
>>> Just so we're clear, the existing code was simply racy as not prone to
>>> deadlocks, right?

> >>> This makes it clear that the test above for compatible CPUs can be quite
> >>> easily evaded by using CPU hotplug, but we don't really have a good
> >>> solution for handling that yet...  Hmmm, grumble grumble, I guess if you
> >>> hotplug unsupported CPUs on a KVM/ARM system for now, stuff will break.

>>
>> In this particular case, there was no deadlock possibility, rather the
>> existing code had insufficient synchronization against CPU hotplug.
>>
>> init_hyp_mode() would invoke cpu_init_hyp_mode() on currently online CPUs
>> using on_each_cpu(). If a CPU came online after this point and before calling
>> register_cpu_notifier(), that CPU would remain uninitialized because this
>> subsystem would miss the hot-online event. This patch fixes this bug and
>> also uses the new synchronization method (instead of get/put_online_cpus())
>> to ensure that we don't deadlock with CPU hotplug.
>>
>
> Yes, that was my conclusion as well.  Thanks for clarifying.  (It could
> be noted in the commit message as well if you should feel so inclined).
>

Please find the patch with updated changelog (and your Ack) below.
(No changes in code).

From: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Subject: [PATCH] arm, kvm: Fix CPU hotplug callback registration

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

In the existing arm kvm code, there is no synchronization with CPU hotplug
to avoid missing the hotplug events that might occur after invoking
init_hyp_mode() and before calling register_cpu_notifier(). Fix this bug
and also use the new synchronization method (instead of get/put_online_cpus())
to ensure that we don't deadlock with CPU hotplug.

Cc: Gleb Natapov <gleb@kernel.org>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Ingo Molnar <mingo@kernel.org>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Christoffer Dall <christoffer.dall@linaro.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

Fix the hw-breakpoint code in arm by using this latter form of callback
registration.

Cc: Russell King <linux@arm.linux.org.uk>
Cc: Ingo Molnar <mingo@kernel.org>
Acked-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

Fix the error injection code in ia64 by using this latter form of callback
registration.

Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

Fix the topology code in ia64 by using this latter form of callback
registration.

Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

Fix the palinfo code in ia64 by using this latter form of callback
registration.

Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

Fix the salinfo code in ia64 by using this latter form of callback
registration.

Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Subsystems that want to register CPU hotplug callbacks, as well as perform
initialization for the CPUs that are already online, often do it as shown
below:

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

This is wrong, since it is prone to ABBA deadlocks involving the
cpu_add_remove_lock and the cpu_hotplug.lock (when running concurrently
with CPU hotplug operations).

Instead, the correct and race-free way of performing the callback
registration is:

	cpu_notifier_register_begin();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* Note the use of the double underscored version of the API */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_notifier_register_done();

Fix the perf subsystem's hotplug notifier by using this latter form of
callback registration.

Also provide a bare-bones version of perf_cpu_notifier() that doesn't
invoke the notifiers for the already online CPUs. This would be useful
for subsystems that need to perform a different set of initialization
for the already online CPUs, or don't need the initialization altogether.

Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Recommend the usage of the new CPU hotplug callback registration APIs
(__register_cpu_notifier() etc), when subsystems need to also perform
initialization for already online CPUs. Provide examples of correct
and race-free ways of achieving this, and point out the kinds of code
that are error-prone.

Cc: Rob Landley <rob@landley.net>
Cc: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

The following method of CPU hotplug callback registration is not safe
due to the possibility of an ABBA deadlock involving the cpu_add_remove_lock
and the cpu_hotplug.lock.

	get_online_cpus();

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	register_cpu_notifier(&foobar_cpu_notifier);

	put_online_cpus();

The deadlock is shown below:

          CPU 0                                         CPU 1
          -----                                         -----

   Acquire cpu_hotplug.lock
   [via get_online_cpus()]

                                              CPU online/offline operation
                                              takes cpu_add_remove_lock
                                              [via cpu_maps_update_begin()]

   Try to acquire
   cpu_add_remove_lock
   [via register_cpu_notifier()]

                                              CPU online/offline operation
                                              tries to acquire cpu_hotplug.lock
                                              [via cpu_hotplug_begin()]

                            *** DEADLOCK! ***

The problem here is that callback registration takes the locks in one order
whereas the CPU hotplug operations take the same locks in the opposite order.
To avoid this issue and to provide a race-free method to register CPU hotplug
callbacks (along with initialization of already online CPUs), introduce new
variants of the callback registration APIs that simply register the callbacks
without holding the cpu_add_remove_lock during the registration. That way,
we can avoid the ABBA scenario. However, we will need to hold the
cpu_add_remove_lock throughout the entire critical section, to protect updates
to the callback/notifier chain.

This can be achieved by writing the callback registration code as follows:

	cpu_maps_update_begin(); [ or cpu_notifier_register_begin(); see below ]

	for_each_online_cpu(cpu)
		init_cpu(cpu);

	/* This doesn't take the cpu_add_remove_lock */
	__register_cpu_notifier(&foobar_cpu_notifier);

	cpu_maps_update_done();  [ or cpu_notifier_register_done(); see below ]

Note that we can't use get_online_cpus() here instead of cpu_maps_update_begin()
because the cpu_hotplug.lock is dropped during the invocation of CPU_POST_DEAD
notifiers, and hence get_online_cpus() cannot provide the necessary
synchronization to protect the callback/notifier chains against concurrent
reads and writes. On the other hand, since the cpu_add_remove_lock protects
the entire hotplug operation (including CPU_POST_DEAD), we can use
cpu_maps_update_begin/done() to guarantee proper synchronization.

Also, since cpu_maps_update_begin/done() is like a super-set of
get/put_online_cpus(), the former naturally protects the critical sections
from concurrent hotplug operations.

Since the names cpu_maps_update_begin/done() don't make much sense in CPU
hotplug callback registration scenarios, we'll introduce new APIs named
cpu_notifier_register_begin/done() and map them to cpu_maps_update_begin/done().

In summary, introduce the lockless variants of un/register_cpu_notifier() and
also export the cpu_notifier_register_begin/done() APIs for use by modules.
This way, we provide a race-free way to register hotplug callbacks as well as
perform initialization for the CPUs that are already online.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@kernel.org>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Toshi Kani <toshi.kani@hp.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Add lockdep annotations for get/put_online_cpus() and
cpu_hotplug_begin()/cpu_hotplug_end().

Cc: Ingo Molnar <mingo@kernel.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Pull scheduler fixes from Ingo Molnar:
 "Three small fixes"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/clock: Prevent tracing recursion in sched_clock_cpu()
  stop_machine: Fix^2 race between stop_two_cpus() and stop_cpus()
  sched/deadline: Deny unprivileged users to set/change SCHED_DEADLINE policy

Pull perf fixes from Ingo Molnar:
 "Misc smaller fixes"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86: Fix leak in uncore_type_init failure paths
  perf machine: Use map as success in ip__resolve_ams
  perf symbols: Fix crash in elf_section_by_name
  perf trace: Decode architecture-specific signal numbers

While testing and documenting the msgrcv() MSG_COPY flag that Stanislav
Kinsbursky added in commit 4a674f34 ("ipc: introduce message queue
copy feature" => kernel 3.8), I discovered a couple of bugs in the
implementation.  The two bugs concern MSG_COPY interactions with other
msgrcv() flags, namely:

 (A) MSG_COPY + MSG_EXCEPT
 (B) MSG_COPY + !IPC_NOWAIT

The bugs are distinct (and the fix for the first one is obvious),
however my fix for both is a single-line patch, which is why I'm
combining them in a single mail, rather than writing two mails+patches.

 ===== (A) MSG_COPY + MSG_EXCEPT =====

With the addition of the MSG_COPY flag, there are now two msgrcv()
flags--MSG_COPY and MSG_EXCEPT--that modify the meaning of the 'msgtyp'
argument in unrelated ways.  Specifying both in the same call is a
logical error that is currently permitted, with the effect that MSG_COPY
has priority and MSG_EXCEPT is ignored.  The call should give an error
if both flags are specified.  The patch below implements that behavior.

 ===== (B) (B) MSG_COPY + !IPC_NOWAIT =====

The test code that was submitted in commit 3a665531 ("selftests: IPC
message queue copy feature test") shows MSG_COPY being used in
conjunction with IPC_NOWAIT.  In other words, if there is no message at
the position 'msgtyp'.  return immediately with the error in ENOMSG.

What was not (fully) tested is the behavior if MSG_COPY is specified
*without* IPC_NOWAIT, and there is an odd behavior.  If the queue
contains less than 'msgtyp' messages, then the call blocks until the
next message is written to the queue.  At that point, the msgrcv() call
returns a copy of the newly added message, regardless of whether that
message is at the ordinal position 'msgtyp'.  This is clearly bogus, and
problematic for applications that might want to make use of the MSG_COPY
flag.

I considered the following possible solutions to this problem:

 (1) Force the call to block until a message *does* appear at the
     position 'msgtyp'.

 (2) If the MSG_COPY flag is specified, the kernel should implicitly add
     IPC_NOWAIT, so that the call fails with ENOMSG for this case.

 (3) If the MSG_COPY flag is specified, but IPC_NOWAIT is not, generate
     an error (probably, EINVAL is the right one).

I do not know if any application would really want to have the
functionality of solution (1), especially since an application can
determine in advance the number of messages in the queue using msgctl()
IPC_STAT.  Obviously, this solution would be the most work to implement.

Solution (2) would have the effect of silently fixing any applications
that tried to employ broken behavior.  However, it would mean that if we
later decided to implement solution (1), then user-space could not
easily detect what the kernel supports (but, since I'm somewhat doubtful
that solution (1) is needed, I'm not sure that this is much of a
problem).

Solution (3) would have the effect of informing broken applications that
they are doing something broken.  The downside is that this would cause
a ABI breakage for any applications that are currently employing the
broken behavior.  However:

a) Those applications are almost certainly not getting the results they
   expect.
b) Possibly, those applications don't even exist, because MSG_COPY is
   currently hidden behind CONFIG_CHECKPOINT_RESTORE.

The upside of solution (3) is that if we later decided to implement
solution (1), user-space could determine what the kernel supports, via
the error return.

In my view, solution (3) is mildly preferable to solution (2), and
solution (1) could still be done later if anyone really cares.  The
patch below implements solution (3).

PS.  For anyone out there still listening, it's the usual story:
documenting an API (and the thinking about, and the testing of the API,
that documentation entails) is the one of the single best ways of
finding bugs in the API, as I've learned from a lot of experience.  Best
to do that documentation before releasing the API.
Signed-off-by: Michael Kerrisk <mtk.manpages@gmail.com>
Acked-by: Stanislav Kinsbursky <skinsbursky@parallels.com>
Cc: Stanislav Kinsbursky <skinsbursky@parallels.com>
Cc: stable@vger.kernel.org
Cc: Serge Hallyn <serge.hallyn@canonical.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Pavel Emelyanov <xemul@parallels.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Pull SCSI fixes from James Bottomley:
 "This is a set of six fixes.  Two are instant crash/null deref types
  (storvsc and isci).  The two qla2xxx are initialisation problems that
  cause MSI-X failures and card misdetection, the isci erroneous macro
  is actually illegal C that's causing a miscompile with certain gcc
  versions and the be2iscsi bad if expression is a static checker fix"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  [SCSI] storvsc: NULL pointer dereference fix
  [SCSI] qla2xxx: Poll during initialization for ISP25xx and ISP83xx
  [SCSI] isci: correct erroneous for_each_isci_host macro
  [SCSI] isci: fix reset timeout handling
  [SCSI] be2iscsi: fix bad if expression
  [SCSI] qla2xxx: Fix multiqueue MSI-X registration.

Pull x86 fixes from Peter Anvin:
 "Two x86 fixes: Suresh's eager FPU fix, and a fix to the NUMA quirk for
  AMD northbridges.

  This only includes Suresh's fix patch, not the "mostly a cleanup"
  patch which had __init issues"

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/amd/numa: Fix northbridge quirk to assign correct NUMA node
  x86, fpu: Check tsk_used_math() in kernel_fpu_end() for eager FPU

Pull ACPI and power management fixes from Rafael Wysocki:
 "Three of these are regression fixes, for two recent regressions and
  one introduced during the 3.13 cycle, and the fourth one is a working
  version of the fix that had to be reverted last time.

  Specifics:

   - A recent ACPI resources handling fix overlooked the fact that it
     had to update the ACPI PNP subsystem's resources parsing too and
     caused confusing warning messages to be printed during system
     intialization on some systems (with arguably buggy ACPI tables).
     Fix from Zhang Rui.

   - Moving the early ACPI initialization before timekeeping_init()
     earlier in this cycle broke fast TSC calibration on at least one
     system, so it needs to be done later, but still before
     efi_enter_virtual_mode() to allow the EFI initialization to refer
     to ACPI.

   - A change related to code duplication reduction in the cpufreq core
     inadvertently caused cpufreq intialization to fail for some CPUs
     handled by intel_pstate by adding checks that may fail for that
     driver, but aren't even necessary when it is used.  The issue is
     addressed by preventing those checks from run in the configurations
     in which they aren't needed.

   - If the Hardware Reduced ACPI flag is set in the ACPI tables, system
     suspend, hibernation and ACPI power off will only work when special
     sleep control and sleep status registeres are provided (their
     addresses in the ACPI tables are not zero).  If those registers are
     not available, the features in question have no chances to work, so
     they shouldn't even be regarded as supported.  That helps with
     power off in particular, because alternative power off methods may
     be used then and they may actually work"

* tag 'pm+acpi-3.14-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  ACPI / sleep: Add extra checks for HW Reduced ACPI mode sleep states
  ACPI / init: Invoke early ACPI initialization later
  cpufreq: Skip current frequency initialization for ->setpolicy drivers
  PNP / ACPI: proper handling of ACPI IO/Memory resource parsing failures

Pull device-mapper fixes form Mike Snitzer:
 "Two small fixes for the DM cache target:

   - fix corruption with >2TB fast device due to truncation bug
   - fix access beyond end of origin device due to a partial block"

* tag 'dm-3.14-fixes-4' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm cache: fix access beyond end of origin device
  dm cache: fix truncation bug when copying a block to/from >2TB fast device

For systems with multiple servers and routed fabric, all
northbridges get assigned to the first server. Fix this by also
using the node reported from the PCI bus. For single-fabric
systems, the northbriges are on PCI bus 0 by definition, which
are on NUMA node 0 by definition, so this is invarient on most
systems.

Tested on fam10h and fam15h single and multi-fabric systems and
candidate for stable.
Signed-off-by: Daniel J Blueman <daniel@numascale.com>
Acked-by: Steffen Persvold <sp@numascale.com>
Acked-by: Borislav Petkov <bp@suse.de>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/1394710981-3596-1-git-send-email-daniel@numascale.comSigned-off-by: Ingo Molnar <mingo@kernel.org>

Pull drm fixes from Dave Airlie:
 "Pretty minor set of fixes for radeon, ttm and vmwgfx.  The ttm ones
  are a regression and an oops seen on server chipsets"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/vmwgfx: Fix a surface reference corner-case in legacy emulation mode
  drm/radeon/cik: properly set compute ring status on disable
  drm/radeon/cik: stop the sdma engines in the enable() function
  drm/radeon/cik: properly set sdma ring status on disable
  drm/radeon: fix runpm disabling on non-PX harder
  drm/ttm: don't oops if no invalidate_caches()
  drm/ttm: Work around performance regression with VM_PFNMAP

Pull i2c Kconfig fix from Wolfram Sang.

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: Remove usage of orphaned symbol OF_I2C

Pull networking fixes from David Miller:
 "I know this is a bit more than you want to see, and I've told the
  wireless folks under no uncertain terms that they must severely scale
  back the extent of the fixes they are submitting this late in the
  game.

  Anyways:

   1) vmxnet3's netpoll doesn't perform the equivalent of an ISR, which
      is the correct implementation, like it should.  Instead it does
      something like a NAPI poll operation.  This leads to crashes.

      From Neil Horman and Arnd Bergmann.

   2) Segmentation of SKBs requires proper socket orphaning of the
      fragments, otherwise we might access stale state released by the
      release callbacks.

      This is a 5 patch fix, but the initial patches are giving
      variables and such significantly clearer names such that the
      actual fix itself at the end looks trivial.

      From Michael S.  Tsirkin.

   3) TCP control block release can deadlock if invoked from a timer on
      an already "owned" socket.  Fix from Eric Dumazet.

   4) In the bridge multicast code, we must validate that the
      destination address of general queries is the link local all-nodes
      multicast address.  From Linus Lüssing.

   5) The x86 BPF JIT support for negative offsets puts the parameter
      for the helper function call in the wrong register.  Fix from
      Alexei Starovoitov.

   6) The descriptor type used for RTL_GIGA_MAC_VER_17 chips in the
      r8169 driver is incorrect.  Fix from Hayes Wang.

   7) The xen-netback driver tests skb_shinfo(skb)->gso_type bits to see
      if a packet is a GSO frame, but that's not the correct test.  It
      should use skb_is_gso(skb) instead.  Fix from Wei Liu.

   8) Negative msg->msg_namelen values should generate an error, from
      Matthew Leach.

   9) at86rf230 can deadlock because it takes the same lock from it's
      ISR and it's hard_start_xmit method, without disabling interrupts
      in the latter.  Fix from Alexander Aring.

  10) The FEC driver's restart doesn't perform operations in the correct
      order, so promiscuous settings can get lost.  Fix from Stefan
      Wahren.

  11) Fix SKB leak in SCTP cookie handling, from Daniel Borkmann.

  12) Reference count and memory leak fixes in TIPC from Ying Xue and
      Erik Hugne.

  13) Forced eviction in inet_frag_evictor() must strictly make sure all
      frags are deleted, otherwise module unload (f.e.  6lowpan) can
      crash.  Fix from Florian Westphal.

  14) Remove assumptions in AF_UNIX's use of csum_partial() (which it
      uses as a hash function), which breaks on PowerPC.  From Anton
      Blanchard.

      The main gist of the issue is that csum_partial() is defined only
      as a value that, once folded (f.e.  via csum_fold()) produces a
      correct 16-bit checksum.  It is legitimate, therefore, for
      csum_partial() to produce two different 32-bit values over the
      same data if their respective alignments are different.

  15) Fix endiannes bug in MAC address handling of ibmveth driver, also
      from Anton Blanchard.

  16) Error checks for ipv6 exthdrs offload registration are reversed,
      from Anton Nayshtut.

  17) Externally triggered ipv6 addrconf routes should count against the
      garbage collection threshold.  Fix from Sabrina Dubroca.

  18) The PCI shutdown handler added to the bnx2 driver can wedge the
      chip if it was not brought up earlier already, which in particular
      causes the firmware to shut down the PHY.  Fix from Michael Chan.

  19) Adjust the sanity WARN_ON_ONCE() in qdisc_list_add() because as
      currently coded it can and does trigger in legitimate situations.
      From Eric Dumazet.

  20) BNA driver fails to build on ARM because of a too large udelay()
      call, fix from Ben Hutchings.

  21) Fair-Queue qdisc holds locks during GFP_KERNEL allocations, fix
      from Eric Dumazet.

  22) The vlan passthrough ops added in the previous release causes a
      regression in source MAC address setting of outgoing headers in
      some circumstances.  Fix from Peter Boström"

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (70 commits)
  ipv6: Avoid unnecessary temporary addresses being generated
  eth: fec: Fix lost promiscuous mode after reconnecting cable
  bonding: set correct vlan id for alb xmit path
  at86rf230: fix lockdep splats
  net/mlx4_en: Deregister multicast vxlan steering rules when going down
  vmxnet3: fix building without CONFIG_PCI_MSI
  MAINTAINERS: add networking selftests to NETWORKING
  net: socket: error on a negative msg_namelen
  MAINTAINERS: Add tools/net to NETWORKING [GENERAL]
  packet: doc: Spelling s/than/that/
  net/mlx4_core: Load the IB driver when the device supports IBoE
  net/mlx4_en: Handle vxlan steering rules for mac address changes
  net/mlx4_core: Fix wrong dump of the vxlan offloads device capability
  xen-netback: use skb_is_gso in xenvif_start_xmit
  r8169: fix the incorrect tx descriptor version
  tools/net/Makefile: Define PACKAGE to fix build problems
  x86: bpf_jit: support negative offsets
  bridge: multicast: enable snooping on general queries only
  bridge: multicast: add sanity check for general query destination
  tcp: tcp_release_cb() should release socket ownership
  ...

The symbol is an orphan, don't depend on it anymore.
Signed-off-by: Richard Weinberger <richard@nod.at>
[wsa: enhanced commit message]
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Fixes: 687b81d0 (i2c: move OF helpers into the core)
Cc: stable@kernel.org

* pnp:
  PNP / ACPI: proper handling of ACPI IO/Memory resource parsing failures

* acpi-init:
  ACPI / init: Invoke early ACPI initialization later

* acpi-sleep:
  ACPI / sleep: Add extra checks for HW Reduced ACPI mode sleep states

* pm-cpufreq:
  cpufreq: Skip current frequency initialization for ->setpolicy drivers

If the HW Reduced ACPI mode bit is set in the FADT, ACPICA uses
the optional sleep control and sleep status registers for making
the system enter sleep states (including S5), so it is not possible
to use system sleep states or power it off using ACPI if the HW
Reduced ACPI mode bit is set and those registers are not available.

For this reason, add a new function, acpi_sleep_state_supported(),
checking if the HW Reduced ACPI mode bit is set and whether or not
system sleep states are usable in that case in addition to checking
the return value of acpi_get_sleep_type_data() and make the ACPI
sleep setup routines use that function to check the availability of
system sleep states.

Among other things, this prevents the kernel from attempting to
use ACPI for powering off HW Reduced ACPI systems without the sleep
control and sleep status registers, because ACPI power off doesn't
have a chance to work on them.  That allows alternative power off
mechanisms that may actually work to be used on those systems.  The
affected machines include Dell Venue 8 Pro, Asus T100TA, Haswell
Desktop SDP and Ivy Bridge EP Demo depot.

References: https://bugzilla.kernel.org/show_bug.cgi?id=70931Reported-by: Adam Williamson <awilliam@redhat.com>
Tested-by: Aubrey Li <aubrey.li@linux.intel.com>
Cc: 3.4+ <stable@vger.kernel.org> # 3.4+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

tmp_prefered_lft is an offset to ifp->tstamp, not now. Therefore
age needs to be added to the condition.

Age calculation in ipv6_create_tempaddr is different from the one
in addrconf_verify and doesn't consider ADDRCONF_TIMER_FUZZ_MINUS.
This can cause age in ipv6_create_tempaddr to be less than the one
in addrconf_verify and therefore unnecessary temporary address to
be generated.
Use age calculation as in addrconf_modify to avoid this.
Signed-off-by: Heiner Kallweit <heiner.kallweit@web.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

If the Freescale fec is in promiscuous mode and network cable is
reconnected then the promiscuous mode get lost. The problem is caused
by a too soon call of set_multicast_list to re-enable promisc mode.
The FEC_R_CNTRL register changes are overwritten by fec_restart.

This patch fixes this by moving the call behind the init of FEC_R_CNTRL
register in fec_restart.

Successful tested on a i.MX28 board.
Signed-off-by: Stefan Wahren <stefan.wahren@i2se.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

The commit d3ab3ffd
(bonding: use rlb_client_info->vlan_id instead of ->tag)
remove the rlb_client_info->tag, but occur some issues,
The vlan_get_tag() will return 0 for success and -EINVAL for
error, so the client_info->vlan_id always be set to 0 if the
vlan_get_tag return 0 for success, so the client_info would
never get a correct vlan id.

We should only set the vlan id to 0 when the vlan_get_tag return error.

Fixes: d3ab3ffd (bonding: use rlb_client_info->vlan_id instead of ->tag)

CC: Ding Tianhong <dingtianhong@huawei.com>
CC: Jay Vosburgh <fubar@us.ibm.com>
CC: Andy Gospodarek <andy@greyhouse.net>
Signed-off-by: Ding Tianhong <dingtianhong@huawei.com>
Acked-by: Veaceslav Falico <vfalico@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

This patch fix a lockdep in the at86rf230 driver, otherwise we get:

[   30.206517] =================================
[   30.211078] [ INFO: inconsistent lock state ]
[   30.215647] 3.14.0-20140108-1-00994-g32e9426 #163 Not tainted
[   30.221660] ---------------------------------
[   30.226222] inconsistent {HARDIRQ-ON-W} -> {IN-HARDIRQ-W} usage.
[   30.232514] systemd-udevd/157 [HC1[1]:SC0[0]:HE0:SE1] takes:
[   30.238439]  (&(&lp->lock)->rlock){?.+...}, at: [<c03600f8>] at86rf230_isr+0x18/0x44
[   30.246621] {HARDIRQ-ON-W} state was registered at:
[   30.251728]   [<c0061ce4>] __lock_acquire+0x7a4/0x18d8
[   30.257135]   [<c0063500>] lock_acquire+0x68/0x7c
[   30.262071]   [<c0588820>] _raw_spin_lock+0x28/0x38
[   30.267203]   [<c0361240>] at86rf230_xmit+0x1c/0x144
[   30.272412]   [<c057ba6c>] mac802154_xmit_worker+0x88/0x148
[   30.278271]   [<c0047844>] process_one_work+0x274/0x404
[   30.283761]   [<c00484c0>] worker_thread+0x228/0x374
[   30.288971]   [<c004cfb8>] kthread+0xd0/0xe4
[   30.293455]   [<c000dac8>] ret_from_fork+0x14/0x2c
[   30.298493] irq event stamp: 8948
[   30.301963] hardirqs last  enabled at (8947): [<c00cb290>] __kmalloc+0xb4/0x110
[   30.309636] hardirqs last disabled at (8948): [<c00115d4>] __irq_svc+0x34/0x5c
[   30.317215] softirqs last  enabled at (8452): [<c0037324>] __do_softirq+0x1dc/0x264
[   30.325243] softirqs last disabled at (8439): [<c0037638>] irq_exit+0x80/0xf4

We use the lp->lock inside the isr of at86rf230, that's why we need the
irqsave spinlock calls.
Signed-off-by: Alexander Aring <alex.aring@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

When mlx4_en_stop_port() is called, we need to deregister also the
tunnel steering rules that relate to multicast.
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Since commit d25f06ea "vmxnet3: fix netpoll race condition",
the vmxnet3 driver fails to build when CONFIG_PCI_MSI is disabled,
because it unconditionally references the vmxnet3_msix_rx()
function.

To fix this, use the same #ifdef in the caller that exists around
the function definition.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Neil Horman <nhorman@tuxdriver.com>
Cc: Shreyas Bhatewara <sbhatewara@vmware.com>
Cc: "VMware, Inc." <pv-drivers@vmware.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: stable@vger.kernel.org
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Add it to NETWORKING [GENERAL] to make sure patches for selftests
go to the netdev list as well.
Signed-off-by: Daniel Borkmann <dborkman@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Second pull request of 2014-03-12. The first one was requested to be canceled.

Rob's fix for oops on invalidate_caches() and a fix for a
performance regression.

* tag 'ttm-fixes-3.14-2014-03-12' of git://people.freedesktop.org/~thomash/linux:
  drm/ttm: don't oops if no invalidate_caches()
  drm/ttm: Work around performance regression with VM_PFNMAP

A few more radeon fixes.

* 'drm-fixes-3.14' of git://people.freedesktop.org/~agd5f/linux:
  drm/radeon/cik: properly set compute ring status on disable
  drm/radeon/cik: stop the sdma engines in the enable() function
  drm/radeon/cik: properly set sdma ring status on disable
  drm/radeon: fix runpm disabling on non-PX harder

Merge tag 'vmwgfx-fixes-3.14-2014-03-13' of git://people.freedesktop.org/~thomash/linux into drm-fixes

Pull request of 2014-03-13

one minor fix for new hw

* tag 'vmwgfx-fixes-3.14-2014-03-13' of git://people.freedesktop.org/~thomash/linux:
  drm/vmwgfx: Fix a surface reference corner-case in legacy emulation mode

If running on a gb-object capable device with a non-gb capable surface
exporter (X server) and a gb capable surface referencing client (GL driver),
the referencing client expects to find a shareable backing buffer attached to
the surface at reference time. This may not be the case if the surface has
not yet been validated. This would cause the surface reference IOCTL to
return an error.
Signed-off-by: Thomas Hellstrom <thellstrom@vmware.com>
Reviewed-by: Jakob Bornecrantz <jakob@vmware.com>

Pull PCI fixes from Bjorn Helgaas:
 "These are two important regression fixes for bugs we've introduced so
  far in v3.14.

  One of the resource allocation changes from the merge window is broken
  for 32-bit kernels where we don't use _CRS for PCI host bridges
  (mostly pre-2008 machines), so there's a fix for that.

  The INTx enable change we put in after the merge window turned out to
  break pciehp because we re-enable INTx on the hotplug bridge, which
  apparently breaks MSI for future hotplug events"

* tag 'pci-v3.14-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  PCI: Don't check resource_size() in pci_bus_alloc_resource()
  PCI: Enable INTx in pci_reenable_device() only when MSI/MSI-X not enabled

Pull KVM fixes from Paolo Bonzini:
 "The ARM patch fixes a build breakage with randconfig.  The x86 one
  fixes Windows guests on AMD processors"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: SVM: fix cr8 intercept window
  ARM: KVM: fix non-VGIC compilation