memcg/sl[au]b: track all the memcg children of a kmem_cache

This enables us to remove all the children of a kmem_cache being
destroyed, if for example the kernel module it's being used in gets
unloaded.  Otherwise, the children will still point to the destroyed
parent.

Signed-off-by: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

move include of workqueue.h to top of slab.h file

Suggested by akpm. I originally decided to put it closer to the use of
the work struct, but let's move it to top.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: destroy memcg caches

Implement destruction of memcg caches.  Right now, only caches where our
reference counter is the last remaining are deleted.  If there are any
other reference counters around, we just leave the caches lying around
until they go away.

When that happens, a destruction function is called from the cache code.
Caches are only destroyed in process context, so we queue them up for
later processing in the general case.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

sl[au]b: allocate objects from memcg cache

We are able to match a cache allocation to a particular memcg.  If the
task doesn't change groups during the allocation itself - a rare event,
this will give us a good picture about who is the first group to touch a
cache page.

This patch uses the now available infrastructure by calling
memcg_kmem_get_cache() before all the cache allocations.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

sl[au]b: always get the cache from its page in kmem_cache_free()

struct page already has this information.  If we start chaining caches,
this information will always be more trustworthy than whatever is passed
into the function.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: remove test for current->mm in memcg_stop/resume_kmem_account

The original reason for the existence of this test, was that
memcg_kmem_cache_create could be called from either softirq context (where
memcg_stop/resume_account is not needed), or process context, (where
memcg_stop/resume_account is needed).  Just skipping it in-function was
the cleanest way to merge both behaviors.  The reason for that is that we
would try to create caches right away through memcg_kmem_cache_create if
the context would allow us to.

However, the final version of the code that merged did not have this
behavior and we always queue up new cache creation.  Thus, instead of a
comment explaining why current->mm test is needed, my proposal in this
patch is to remove memcg_stop/resume_account from the worker thread and
make sure all callers have a valid mm context.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Michal Hocko <mhocko@suse.cz>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: skip memcg kmem allocations in specified code regions

Create a mechanism that skip memcg allocations during certain pieces of
our core code.  It basically works in the same way as
preempt_disable()/preempt_enable(): By marking a region under which all
allocations will be accounted to the root memcg.

We need this to prevent races in early cache creation, when we
allocate data using caches that are not necessarily created already.

Signed-off-by: Glauber Costa <glommer@parallels.com>
yCc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: infrastructure to match an allocation to the right cache

The page allocator is able to bind a page to a memcg when it is allocated.
 But for the caches, we'd like to have as many objects as possible in a
page belonging to the same cache.

This is done in this patch by calling memcg_kmem_get_cache in the
beginning of every allocation function.  This function is patched out by
static branches when kernel memory controller is not being used.

It assumes that the task allocating, which determines the memcg in the
page allocator, belongs to the same cgroup throughout the whole process.
Misaccounting can happen if the task calls memcg_kmem_get_cache() while
belonging to a cgroup, and later on changes.  This is considered
acceptable, and should only happen upon task migration.

Before the cache is created by the memcg core, there is also a possible
imbalance: the task belongs to a memcg, but the cache being allocated from
is the global cache, since the child cache is not yet guaranteed to be
ready.  This case is also fine, since in this case the GFP_KMEMCG will not
be passed and the page allocator will not attempt any cgroup accounting.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: simplify ida initialization

As suggested by akpm, change the manual initialization of our kmem
index to DEFINE_IDA()

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: allocate memory for memcg caches whenever a new memcg appears

Every cache that is considered a root cache (basically the "original"
caches, tied to the root memcg/no-memcg) will have an array that should be
large enough to store a cache pointer per each memcg in the system.

Theoreticaly, this is as high as 1 << sizeof(css_id), which is currently
in the 64k pointers range.  Most of the time, we won't be using that much.

What goes in this patch, is a simple scheme to dynamically allocate such
an array, in order to minimize memory usage for memcg caches.  Because we
would also like to avoid allocations all the time, at least for now, the
array will only grow.  It will tend to be big enough to hold the maximum
number of kmem-limited memcgs ever achieved.

We'll allocate it to be a minimum of 64 kmem-limited memcgs.  When we have
more than that, we'll start doubling the size of this array every time the
limit is reached.

Because we are only considering kmem limited memcgs, a natural point for
this to happen is when we write to the limit.  At that point, we already
have set_limit_mutex held, so that will become our natural synchronization
mechanism.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

slab/slub: consider a memcg parameter in kmem_create_cache

Allow a memcg parameter to be passed during cache creation.  When the slub
allocator is being used, it will only merge caches that belong to the same
memcg.  We'll do this by scanning the global list, and then translating
the cache to a memcg-specific cache

Default function is created as a wrapper, passing NULL to the memcg
version.  We only merge caches that belong to the same memcg.

A helper is provided, memcg_css_id: because slub needs a unique cache name
for sysfs.  Since this is visible, but not the canonical location for slab
data, the cache name is not used, the css_id should suffice.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

slab: annotate on-slab caches nodelist locks

We currently provide lockdep annotation for kmalloc caches, and also
caches that have SLAB_DEBUG_OBJECTS enabled.  The reason for this is that
we can quite frequently nest in the l3->list_lock lock, which is not
something trivial to avoid.

My proposal with this patch, is to extend this to caches whose slab
management object lives within the slab as well ("on_slab").  The need for
this arose in the context of testing kmemcg-slab patches.  With such
patchset, we can have per-memcg kmalloc caches.  So the same path that led
to nesting between kmalloc caches will could then lead to in-memcg
nesting.  Because they are not annotated, lockdep will trigger.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

slab/slub: struct memcg_params

For the kmem slab controller, we need to record some extra information in
the kmem_cache structure.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Signed-off-by: Suleiman Souhlal <suleiman@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: add documentation about the kmem controller

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fork: protect architectures where THREAD_SIZE >= PAGE_SIZE against fork bombs

Because those architectures will draw their stacks directly from the page
allocator, rather than the slab cache, we can directly pass __GFP_KMEMCG
flag, and issue the corresponding free_pages.

This code path is taken when the architecture doesn't define
CONFIG_ARCH_THREAD_INFO_ALLOCATOR (only ia64 seems to), and has
THREAD_SIZE >= PAGE_SIZE.  Luckily, most - if not all - of the remaining
architectures fall in this category.

This will guarantee that every stack page is accounted to the memcg the
process currently lives on, and will have the allocations to fail if they
go over limit.

For the time being, I am defining a new variant of THREADINFO_GFP, not to
mess with the other path.  Once the slab is also tracked by memcg, we can
get rid of that flag.

Tested to successfully protect against :(){ :|:& };:

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Frederic Weisbecker <fweisbec@redhat.com>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: execute the whole memcg freeing in free_worker()

A lot of the initialization we do in mem_cgroup_create() is done with
softirqs enabled.  This include grabbing a css id, which holds
&ss->id_lock->rlock, and the per-zone trees, which holds
rtpz->lock->rlock.  All of those signal to the lockdep mechanism that
those locks can be used in SOFTIRQ-ON-W context.  This means that the
freeing of memcg structure must happen in a compatible context, otherwise
we'll get a deadlock, like the one below, caught by lockdep:

  [<ffffffff81103095>] free_accounted_pages+0x47/0x4c
  [<ffffffff81047f90>] free_task+0x31/0x5c
  [<ffffffff8104807d>] __put_task_struct+0xc2/0xdb
  [<ffffffff8104dfc7>] put_task_struct+0x1e/0x22
  [<ffffffff8104e144>] delayed_put_task_struct+0x7a/0x98
  [<ffffffff810cf0e5>] __rcu_process_callbacks+0x269/0x3df
  [<ffffffff810cf28c>] rcu_process_callbacks+0x31/0x5b
  [<ffffffff8105266d>] __do_softirq+0x122/0x277

This usage pattern could not be triggered before kmem came into play.
With the introduction of kmem stack handling, it is possible that we call
the last mem_cgroup_put() from the task destructor, which is run in an rcu
callback.  Such callbacks are run with softirqs disabled, leading to the
offensive usage pattern.

In general, we have little, if any, means to guarantee in which context
the last memcg_put will happen.  The best we can do is test it and try to
make sure no invalid context releases are happening.  But as we add more
code to memcg, the possible interactions grow in number and expose more
ways to get context conflicts.  One thing to keep in mind, is that part of
the freeing process is already deferred to a worker, such as vfree(), that
can only be called from process context.

For the moment, the only two functions we really need moved away are:

  * free_css_id(), and
  * mem_cgroup_remove_from_trees().

But because the later accesses per-zone info,
free_mem_cgroup_per_zone_info() needs to be moved as well.  With that, we
are left with the per_cpu stats only.  Better move it all.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Tested-by: Greg Thelen <gthelen@google.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: allow a memcg with kmem charges to be destructed

Because the ultimate goal of the kmem tracking in memcg is to track slab
pages as well, we can't guarantee that we'll always be able to point a
page to a particular process, and migrate the charges along with it -
since in the common case, a page will contain data belonging to multiple
processes.

Because of that, when we destroy a memcg, we only make sure the
destruction will succeed by discounting the kmem charges from the user
charges when we try to empty the cgroup.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: use static branches when code not in use

We can use static branches to patch the code in or out when not used.

Because the _ACTIVE bit on kmem_accounted is only set after the increment
is done, we guarantee that the root memcg will always be selected for kmem
charges until all call sites are patched (see memcg_kmem_enabled).  This
guarantees that no mischarges are applied.

Static branch decrement happens when the last reference count from the
kmem accounting in memcg dies.  This will only happen when the charges
drop down to 0.

When that happens, we need to disable the static branch only on those
memcgs that enabled it.  To achieve this, we would be forced to complicate
the code by keeping track of which memcgs were the ones that actually
enabled limits, and which ones got it from its parents.

It is a lot simpler just to do static_key_slow_inc() on every child
that is accounted.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: kmem accounting lifecycle management

Because kmem charges can outlive the cgroup, we need to make sure that we
won't free the memcg structure while charges are still in flight.  For
reviewing simplicity, the charge functions will issue mem_cgroup_get() at
every charge, and mem_cgroup_put() at every uncharge.

This can get expensive, however, and we can do better.  mem_cgroup_get()
only really needs to be issued once: when the first limit is set.  In the
same spirit, we only need to issue mem_cgroup_put() when the last charge
is gone.

We'll need an extra bit in kmem_account_flags for that:
KMEM_ACCOUNTED_DEAD.  it will be set when the cgroup dies, if there are
charges in the group.  If there aren't, we can proceed right away.

Our uncharge function will have to test that bit every time the charges
drop to 0.  Because that is not the likely output of res_counter_uncharge,
this should not impose a big hit on us: it is certainly much better than a
reference count decrease at every operation.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

res_counter: return amount of charges after res_counter_uncharge()

It is useful to know how many charges are still left after a call to
res_counter_uncharge.  While it is possible to issue a res_counter_read
after uncharge, this can be racy.

If we need, for instance, to take some action when the counters drop down
to 0, only one of the callers should see it.  This is the same semantics
as the atomic variables in the kernel.

Since the current return value is void, we don't need to worry about
anything breaking due to this change: nobody relied on that, and only
users appearing from now on will be checking this value.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: allocate kernel pages to the right memcg

When a process tries to allocate a page with the __GFP_KMEMCG flag, the
page allocator will call the corresponding memcg functions to validate the
allocation.  Tasks in the root memcg can always proceed.

To avoid adding markers to the page - and a kmem flag that would
necessarily follow, as much as doing page_cgroup lookups for no reason,
whoever is marking its allocations with __GFP_KMEMCG flag is responsible
for telling the page allocator that this is such an allocation at
free_pages() time.  This is done by the invocation of
__free_accounted_pages() and free_accounted_pages().

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: replace __always_inline with plain inline

Following the pattern found in the allocators, where we do our best to the
fast paths function-call free, all the externally visible functions for
kmemcg were marked __always_inline.

It is fair to say, however, that this should be up to the compiler.  We
will still keep as much of the flag testing as we can in memcontrol.h to
give the compiler the option to inline it, but won't force it.

I tested this with 4.7.2, it will inline all three functions anyway when
compiling with -O2, and will refrain from it when compiling with -Os.
This seems like a good behavior.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Cc: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: kmem controller infrastructure

Introduce infrastructure for tracking kernel memory pages to a given
memcg.  This will happen whenever the caller includes the flag
__GFP_KMEMCG flag, and the task belong to a memcg other than the root.

In memcontrol.h those functions are wrapped in inline acessors.  The idea
is to later on, patch those with static branches, so we don't incur any
overhead when no mem cgroups with limited kmem are being used.

Users of this functionality shall interact with the memcg core code
through the following functions:

memcg_kmem_newpage_charge: will return true if the group can handle the
                           allocation. At this point, struct page is not
                           yet allocated.

memcg_kmem_commit_charge: will either revert the charge, if struct page
                          allocation failed, or embed memcg information
                          into page_cgroup.

memcg_kmem_uncharge_page: called at free time, will revert the charge.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: add a __GFP_KMEMCG flag

This flag is used to indicate to the callees that this allocation is a
kernel allocation in process context, and should be accounted to current's
memcg.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Suleiman Souhlal <suleiman@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: kmem accounting basic infrastructure

Add the basic infrastructure for the accounting of kernel memory.  To
control that, the following files are created:

 * memory.kmem.usage_in_bytes
 * memory.kmem.limit_in_bytes
 * memory.kmem.failcnt
 * memory.kmem.max_usage_in_bytes

They have the same meaning of their user memory counterparts.  They
reflect the state of the "kmem" res_counter.

Per cgroup kmem memory accounting is not enabled until a limit is set for
the group.  Once the limit is set the accounting cannot be disabled for
that group.  This means that after the patch is applied, no behavioral
changes exists for whoever is still using memcg to control their memory
usage, until memory.kmem.limit_in_bytes is set for the first time.

We always account to both user and kernel resource_counters.  This
effectively means that an independent kernel limit is in place when the
limit is set to a lower value than the user memory.  A equal or higher
value means that the user limit will always hit first, meaning that kmem
is effectively unlimited.

People who want to track kernel memory but not limit it, can set this
limit to a very high number (like RESOURCE_MAX - 1page - that no one will
ever hit, or equal to the user memory)

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: change defines to an enum

This is just a cleanup patch for clarity of expression.  In earlier
submissions, people asked it to be in a separate patch, so here it is.

Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: reclaim when more than one page needed

mem_cgroup_do_charge() was written before kmem accounting, and expects
three cases: being called for 1 page, being called for a stock of 32
pages, or being called for a hugepage.  If we call for 2 or 3 pages (and
both the stack and several slabs used in process creation are such, at
least with the debug options I had), it assumed it's being called for
stock and just retried without reclaiming.

Fix that by passing down a minsize argument in addition to the csize.

And what to do about that (csize == PAGE_SIZE && ret) retry?  If it's
needed at all (and presumably is since it's there, perhaps to handle
races), then it should be extended to more than PAGE_SIZE, yet how far?
And should there be a retry count limit, of what?  For now retry up to
COSTLY_ORDER (as page_alloc.c does) and make sure not to do it if
__GFP_NORETRY.

v4: fixed nr pages calculation pointed out by Christoph Lameter.

Signed-off-by: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memcg: make it possible to use the stock for more than one page

We currently have a percpu stock cache scheme that charges one page at a
time from memcg->res, the user counter.  When the kernel memory controller
comes into play, we'll need to charge more than that.

This is because kernel memory allocations will also draw from the user
counter, and can be bigger than a single page, as it is the case with the
stack (usually 2 pages) or some higher order slabs.

[glommer@parallels.com: added a changelog ]
Signed-off-by: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Frederic Weisbecker <fweisbec@redhat.com>
Cc: Greg Thelen <gthelen@google.com>
Cc: JoonSoo Kim <js1304@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Rik van Riel <riel@redhat.com>
Cc: Suleiman Souhlal <suleiman@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm, oom: allow exiting threads to have access to memory reserves

Exiting threads, those with PF_EXITING set, can pagefault and require
memory before they can make forward progress.  This happens, for instance,
when a process must fault task->robust_list, a userspace structure, before
detaching its memory.

These threads also aren't guaranteed to get access to memory reserves
unless oom killed or killed from userspace.  The oom killer won't grant
memory reserves if other threads are also exiting other than current and
stalling at the same point.  This prevents needlessly killing processes
when others are already exiting.

Instead of special casing all the possible situations between PF_EXITING
getting set and a thread detaching its mm where it may allocate memory,
which probably wouldn't get updated when a change is made to the exit
path, the solution is to give all exiting threads access to memory
reserves if they call the oom killer.  This allows them to quickly
allocate, detach its mm, and free the memory it represents.

Summary of Luigi's bug report:

: He had an oom condition where threads were faulting on task->robust_list
: and repeatedly called the oom killer but it would defer killing a thread
: because it saw other PF_EXITING threads.  This can happen anytime we need
: to allocate memory after setting PF_EXITING and before detaching our mm;
: if there are other threads in the same state then the oom killer won't do
: anything unless one of them happens to be killed from userspace.
:
: So instead of only deferring for PF_EXITING and !task->robust_list, it's
: better to just give them access to memory reserves to prevent a potential
: livelock so that any other faults that may be introduced in the future in
: the exit path don't cause the same problem (and hopefully we don't allow
: too many of those!).

Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Tested-by: Luigi Semenzato <semenzato@google.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

Documentation/cgroups/memory.txt: s/mem_cgroup_charge/mem_cgroup_change_common/

mem_cgroup_charge_common() is invoked as the entry point for cgroup limits
charge rather than mem_cgroup_charge(), as the later has been removed for
years.  Update the cgroup/memory.txt to reflect this change.

Signed-off-by: Jie Liu <jeff.liu@oracle.com>
Cc: Ying Han <yinghan@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: memmap_init_zone() performance improvement

We have what we call an "architectural simulator".  It is a computer
program that pretends that it is a computer system.  We use it to test the
firmware before real hardware is available.  We have booted Linux on our
simulator.  As you would expect it takes longer to boot on the simulator
than it does on real hardware.

With my patch - boot time 41 minutes
Without patch - boot time 94 minutes

These numbers do not scale linearly to real hardware.  But indicate to me
a place where Linux can be improved.

memmap_init_zone() loops through every Page Frame Number (pfn), including
pfn values that are within the gaps between existing memory sections.  The
unneeded looping will become a boot performance issue when machines
configure larger memory ranges that will contain larger and more numerous
gaps.

The code will skip across invalid pfn values to reduce the number of loops
executed.

Signed-off-by: Mike Yoknis <mike.yoknis@hp.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: thp: set the accessed flag for old pages on access fault

On x86 memory accesses to pages without the ACCESSED flag set result in
the ACCESSED flag being set automatically.  With the ARM architecture a
page access fault is raised instead (and it will continue to be raised
until the ACCESSED flag is set for the appropriate PTE/PMD).

For normal memory pages, handle_pte_fault will call pte_mkyoung
(effectively setting the ACCESSED flag).  For transparent huge pages,
pmd_mkyoung will only be called for a write fault.

This patch ensures that faults on transparent hugepages which do not
result in a CoW update the access flags for the faulting pmd.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Ni zhan Chen <nizhan.chen@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm, highmem: get virtual address of the page using PKMAP_ADDR()

In flush_all_zero_pkmaps(), we have an index of the pkmap associated with
the page.  Using this index, we can simply get virtual address of the
page.  So change it.

Signed-off-by: Joonsoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm-highmem-remove-page_address_pool-list-v2

Signed-off-by: Joonsoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm, highmem: remove page_address_pool list

We can find free page_address_map instance without the page_address_pool.
So remove it.

Signed-off-by: Joonsoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm, highmem: remove useless pool_lock

The pool_lock protects the page_address_pool from concurrent access.  But,
access to the page_address_pool is already protected by kmap_lock.  So
remove it.

Signed-off-by: Joonsoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Minchan Kin <minchan@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm, highmem: use PKMAP_NR() to calculate an index of pkmap

To calculate an index of pkmap, using PKMAP_NR() is more understandable
and maintainable, so change it.

Signed-off-by: Joonsoo Kim <js1304@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: do not call frontswap_init() during swapoff

The call to frontswap_init() was added within enable_swap_info(), which
was called not only during sys_swapon, but also to reinsert the swap_info
into the swap_list in case of failure of try_to_unuse() within
sys_swapoff.  This means that frontswap_init() might be called more than
once for the same swap area.

While as far as I could see no frontswap implementation has any problem
with it (and in fact, all the ones I found ignore the parameter passed to
frontswap_init), this could change in the future.

To prevent future problems, move the call to frontswap_init() to outside
the code shared between sys_swapon and sys_swapoff.

Signed-off-by: Cesar Eduardo Barros <cesarb@cesarb.net>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Acked-by: Dan Magenheimer <dan.magenheimer@oracle.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: refactor reinsert of swap_info in sys_swapoff()

The block within sys_swapoff() which re-inserts the swap_info into the
swap_list in case of failure of try_to_unuse() reads a few values outside
the swap_lock.  While this is safe at that point, it is subtle code.

Simplify the code by moving the reading of these values to a separate
function, refactoring it a bit so they are read from within the swap_lock.
 This is easier to understand, and matches better the way it worked before
I unified the insertion of the swap_info from both sys_swapon and
sys_swapoff.

This change should make no functional difference.  The only real change is
moving the read of two or three structure fields to within the lock
(frontswap_map_get() is nothing more than a read of p->frontswap_map).

Signed-off-by: Cesar Eduardo Barros <cesarb@cesarb.net>
Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: Dan Magenheimer <dan.magenheimer@oracle.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mmvmscan-only-evict-file-pages-when-we-have-plenty-fix

adjust comment layout

Cc: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm,vmscan: only evict file pages when we have plenty

If we have more inactive file pages than active file pages, we skip
scanning the active file pages altogether, with the idea that we do not
want to evict the working set when there is plenty of streaming IO in the
cache.

However, the code forgot to also skip scanning anonymous pages in that
situation.  That leads to the curious situation of keeping the active file
pages protected from being paged out when there are lots of inactive file
pages, while still scanning and evicting anonymous pages.

This patch fixes that situation, by only evicting file pages when we have
plenty of them and most are inactive.

Signed-off-by: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: add comment on storage key dirty bit semantics

Add comments that dirty bit in storage key gets set whenever page content
is changed.  Hopefully if someone will use this function, he'll have a
look at one of the two places where we comment on this.

Signed-off-by: Jan Kara <jack@suse.cz>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/memory_hotplug.c: update start_pfn in zone and pg_data when spanned_pages == 0.

If we hot-remove memory only and leave the cpus alive, the corresponding
node will not be removed.  But the node_start_pfn and node_spanned_pages
in pg_data will be reset to 0.  In this case, when we hot-add the memory
back next time, the node_start_pfn will always be 0 because no pfn is less
than 0.  After that, if we hot-remove the memory again, it will cause
kernel panic in function find_biggest_section_pfn() when it tries to scan
all the pfns.

The zone will also have the same problem.

This patch sets start_pfn to the start_pfn of the section being added when
spanned_pages of the zone or pg_data is 0.

---How to reproduce---

1. hot-add a container with some memory and cpus;
2. hot-remove the container's memory, and leave cpus there;
3. hot-add these memory again;
4. hot-remove them again;

then, the kernel will panic.

---Call trace---

[10530.646285] BUG: unable to handle kernel paging request at
00000fff82a8cc38
[10530.729670] IP: [<ffffffff811c0d55>] find_biggest_section_pfn+0xe5/0x180
......
[10533.064975] Call Trace:
[10533.094162]  [<ffffffff811c0fcf>] ? __remove_zone+0x2f/0x1b0
[10533.161757]  [<ffffffff811c1124>] __remove_zone+0x184/0x1b0
[10533.228318]  [<ffffffff811c11dc>] __remove_section+0x8c/0xb0
[10533.295916]  [<ffffffff811c12e7>] __remove_pages+0xe7/0x120
[10533.362476]  [<ffffffff81654f7c>] arch_remove_memory+0x2c/0x80
[10533.432151]  [<ffffffff81655bb6>] remove_memory+0x56/0x90
[10533.496633]  [<ffffffff813da0c8>]
acpi_memory_device_remove_memory+0x48/0x73
[10533.580846]  [<ffffffff813da55a>] acpi_memory_device_notify+0x153/0x274
[10533.659865]  [<ffffffff813a63cf>] ? acpi_bus_get_device+0x2f/0x77
[10533.732653]  [<ffffffff813a6589>] ? acpi_bus_notify+0xb5/0xec
[10533.801291]  [<ffffffff813b6786>] acpi_ev_notify_dispatch+0x41/0x5f
[10533.876156]  [<ffffffff813a3867>] acpi_os_execute_deferred+0x27/0x34
[10533.952062]  [<ffffffff81090589>] process_one_work+0x219/0x680
[10534.021736]  [<ffffffff81090528>] ? process_one_work+0x1b8/0x680
[10534.093488]  [<ffffffff813a3840>] ?
acpi_os_wait_events_complete+0x23/0x23
[10534.175622]  [<ffffffff810923be>] worker_thread+0x12e/0x320
[10534.242181]  [<ffffffff81092290>] ? manage_workers+0x110/0x110
[10534.311855]  [<ffffffff81098396>] kthread+0xc6/0xd0
[10534.370111]  [<ffffffff8167c7c4>] kernel_thread_helper+0x4/0x10
[10534.440824]  [<ffffffff81672230>] ? retint_restore_args+0x13/0x13
[10534.513612]  [<ffffffff810982d0>] ? __init_kthread_worker+0x70/0x70
[10534.588480]  [<ffffffff8167c7c0>] ? gs_change+0x13/0x13
......
[10535.045543] ---[ end trace 96d845dbf33fee11 ]---

Signed-off-by: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

slub, hotplug: ignore unrelated node's hot-adding and hot-removing

SLUB only focuses on the nodes which have normal memory and it ignores the
other node's hot-adding and hot-removing.

Aka: if some memory of a node which has no onlined memory is online, but
this new memory onlined is not normal memory (for example, highmem), we
should not allocate kmem_cache_node for SLUB.

And if the last normal memory is offlined, but the node still has memory,
we should remove kmem_cache_node for that node.  (The current code delays
it when all of the memory is offlined)

So we only do something when marg->status_change_nid_normal > 0.
marg->status_change_nid is not suitable here.

The same problem doesn't exist in SLAB, because SLAB allocates kmem_list3
for every node even the node don't have normal memory, SLAB tolerates
kmem_list3 on alien nodes.  SLUB only focuses on the nodes which have
normal memory, it don't tolerate alien kmem_cache_node.  The patch makes
SLUB become self-compatible and avoids WARNs and BUGs in rare conditions.

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Rob Landley <rob@landley.net>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Jiang Liu <jiang.liu@huawei.com>
Cc: Kay Sievers <kay.sievers@vrfy.org>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory_hotplug: fix possible incorrect node_states[N_NORMAL_MEMORY]

Currently memory_hotplug only manages the node_states[N_HIGH_MEMORY], it
forgets to manage node_states[N_NORMAL_MEMORY].  This may cause
node_states[N_NORMAL_MEMORY] to become incorrect.

Example, if a node is empty before online, and we online a memory which is
in ZONE_NORMAL.  And after online, node_states[N_HIGH_MEMORY] is correct,
but node_states[N_NORMAL_MEMORY] is incorrect, the online code doesn't set
the new online node to node_states[N_NORMAL_MEMORY].

The same thing will happen when offlining (the offline code doesn't clear
the node from node_states[N_NORMAL_MEMORY] when needed).  Some memory
managment code depends node_states[N_NORMAL_MEMORY], so we have to fix up
the node_states[N_NORMAL_MEMORY].

We add node_states_check_changes_online() and
node_states_check_changes_offline() to detect whether
node_states[N_HIGH_MEMORY] and node_states[N_NORMAL_MEMORY] are changed
while hotpluging.

Also add @status_change_nid_normal to struct memory_notify, thus the
memory hotplug callbacks know whether the node_states[N_NORMAL_MEMORY] are
changed.  (We can add a @flags and reuse @status_change_nid instead of
introducing @status_change_nid_normal, but it will add much more
complexity in memory hotplug callback in every subsystem.  So introducing
@status_change_nid_normal is better and it doesn't change the sematics of
@status_change_nid)

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Rob Landley <rob@landley.net>
Cc: Jiang Liu <jiang.liu@huawei.com>
Cc: Kay Sievers <kay.sievers@vrfy.org>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug: build zonelist if a zone is populated after onlining pages

After "memory-hotplug: allocate zone's pcp before onlining pages", we
build zone list before onlining pages to allocate zone's pcp.  But the
zone doesn't have pages before onlining pages, and the zone is not in
zonelist, so we still need to build zonelist after onlining pages.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: make zone_pcp_reset independent of MEMORY_HOTREMOVE

340175b7 (mm/hotplug: free zone->pageset when a zone becomes empty)
introduced zone_pcp_reset and hided it inside CONFIG_MEMORY_HOTREMOVE.
Since "memory-hotplug: allocate zone's pcp before onlining pages" the
function is also called from online_pages which is defined outside
CONFIG_MEMORY_HOTREMOVE which causes a linkage error.

The function, although not used outside of MEMORY_{HOTPLUT,HOTREMOVE},
seems like universal enough so let's keep it at its current location
and only remove the HOTREMOVE guard.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug: allocate zone's pcp before onlining pages

We use __free_page() to put a page to buddy system when onlining pages.
__free_page() will store NR_FREE_PAGES in zone's pcp.vm_stat_diff, so we
should allocate zone's pcp before onlining pages, otherwise we will lose
some free pages.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug, mm/sparse.c: clear the memory to store struct page

If sparse memory vmemmap is enabled, we can't free the memory to store
struct page when a memory device is hotremoved, because we may store
struct page in the memory to manage the memory which doesn't belong to
this memory device.  When we hotadded this memory device again, we will
reuse this memory to store struct page, and struct page may contain some
obsolete information, and we will get bad-page state:

[   59.611278] init_memory_mapping: [mem 0x80000000-0x9fffffff]
[   59.637836] Built 2 zonelists in Node order, mobility grouping on.  Total pages: 547617
[   59.638739] Policy zone: Normal
[   59.650840] BUG: Bad page state in process bash  pfn:9b6dc
[   59.651124] page:ffffea0002200020 count:0 mapcount:0 mapping:          (null) index:0xfdfdfdfdfdfdfdfd
[   59.651494] page flags: 0x2fdfdfdfd5df9fd(locked|referenced|uptodate|dirty|lru|active|slab|owner_priv_1|private|private_2|writeback|head|tail|swapcache|reclaim|swapbacked|unevictable|uncached|compound_lock)
[   59.653604] Modules linked in: netconsole acpiphp pci_hotplug acpi_memhotplug loop kvm_amd kvm microcode tpm_tis tpm tpm_bios evdev psmouse serio_raw i2c_piix4 i2c_core parport_pc parport processor button thermal_sys ext3 jbd mbcache sg sr_mod cdrom ata_generic virtio_net ata_piix virtio_blk libata virtio_pci virtio_ring virtio scsi_mod
[   59.656998] Pid: 988, comm: bash Not tainted 3.6.0-rc7-guest #12
[   59.657172] Call Trace:
[   59.657275]  [<ffffffff810e9b30>] ? bad_page+0xb0/0x100
[   59.657434]  [<ffffffff810ea4c3>] ? free_pages_prepare+0xb3/0x100
[   59.657610]  [<ffffffff810ea668>] ? free_hot_cold_page+0x48/0x1a0
[   59.657787]  [<ffffffff8112cc08>] ? online_pages_range+0x68/0xa0
[   59.657961]  [<ffffffff8112cba0>] ? __online_page_increment_counters+0x10/0x10
[   59.658162]  [<ffffffff81045561>] ? walk_system_ram_range+0x101/0x110
[   59.658346]  [<ffffffff814c4f95>] ? online_pages+0x1a5/0x2b0
[   59.658515]  [<ffffffff8135663d>] ? __memory_block_change_state+0x20d/0x270
[   59.658710]  [<ffffffff81356756>] ? store_mem_state+0xb6/0xf0
[   59.658878]  [<ffffffff8119e482>] ? sysfs_write_file+0xd2/0x160
[   59.659052]  [<ffffffff8113769a>] ? vfs_write+0xaa/0x160
[   59.659212]  [<ffffffff81137977>] ? sys_write+0x47/0x90
[   59.659371]  [<ffffffff814e2f25>] ? async_page_fault+0x25/0x30
[   59.659543]  [<ffffffff814ea239>] ? system_call_fastpath+0x16/0x1b
[   59.659720] Disabling lock debugging due to kernel taint

This patch clears the memory to store struct page to avoid unexpected error.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Reported-by: Vasilis Liaskovitis <vasilis.liaskovitis@profitbricks.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug: suppress "Device nodeX does not have a release() function" warning

When calling unregister_node(), the function shows following message at
device_release().

"Device 'node2' does not have a release() function, it is broken and must
be fixed."

The reason is node's device struct does not have a release() function.

So the patch registers node_device_release() to the device's release()
function for suppressing the warning message.  Additionally, the patch
adds memset() to initialize a node struct into register_node().  Because
the node struct is part of node_devices[] array and it cannot be freed by
node_device_release().  So if system reuses the node struct, it has a
garbage.

Signed-off-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

numa: convert static memory to dynamically allocated memory for per node device

We use a static array to store struct node.  In many cases, we don't have
too many nodes, and some memory will be unused.  Convert it to per-device
dynamically allocated memory.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug: fix NR_FREE_PAGES mismatch's fix

When a page is freed and put into pcp list, get_freepage_migratetype()
doesn't return MIGRATE_ISOLATE even if this pageblock is isolated.
So we should use get_freepage_migratetype() instead of mt to check
whether it is isolated.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Reported-by: Jianguo Wu <wujianguo106@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug: fix NR_FREE_PAGES mismatch

NR_FREE_PAGES will be wrong after offlining pages.  We add/dec
NR_FREE_PAGES like this now:

1. move all pages in buddy system to MIGRATE_ISOLATE, and dec NR_FREE_PAGES

2. don't add NR_FREE_PAGES when it is freed and the migratetype is
   MIGRATE_ISOLATE

3. dec NR_FREE_PAGES when offlining isolated pages.

4. add NR_FREE_PAGES when undoing isolate pages.

When we come to step 3, all pages are in MIGRATE_ISOLATE list, and
NR_FREE_PAGES are right.  When we come to step4, all pages are not in
buddy system, so we don't change NR_FREE_PAGES in this step, but we change
NR_FREE_PAGES in step3.  So NR_FREE_PAGES is wrong after offlining pages.
So there is no need to change NR_FREE_PAGES in step3.

This patch also fixs a problem in step2: if the migratetype is
MIGRATE_ISOLATE, we should not add NR_FRR_PAGES when we remove pages from
pcppages.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug: auto offline page_cgroup when onlining memory block failed

When a memory block is onlined, we will try allocate memory on that node
to store page_cgroup.  If onlining the memory block failed, we don't
offline the page cgroup, and we have no chance to offline this page cgroup
unless the memory block is onlined successfully again.  It will cause that
we can't hot-remove the memory device on that node, because some memory is
used to store page cgroup.  If onlining the memory block is failed, there
is no need to stort page cgroup for this memory.  So auto offline
page_cgroup when onlining memory block failed.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug-update-mce_bad_pages-when-removing-the-memory-fix

cleanup ifdefs

Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug: update mce_bad_pages when removing the memory

When we hotremove a memory device, we will free the memory to store struct
page.  If the page is hwpoisoned page, we should decrease mce_bad_pages.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory-hotplug: skip HWPoisoned page when offlining pages

hwpoisoned may be set when we offline a page by the sysfs interface
/sys/devices/system/memory/soft_offline_page or
/sys/devices/system/memory/hard_offline_page. If we don't clear
this flag when onlining pages, this page can't be freed, and will
not in free list. So we can't offline these pages again. So we
should skip such page when offlining pages.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory hotplug: suppress "Device memoryX does not have a release() function" warning

When calling remove_memory_block(), the function shows following message
at device_release().

"Device 'memory528' does not have a release() function, it is broken and
must be fixed."

The reason is memory_block's device struct does not have a release()
function.

So the patch registers memory_block_release() to the device's release()
function for suppressing the warning message.  Additionally, the patch
moves kfree(mem) into the release function since the release function is
prepared as a means to free a memory_block struct.

Signed-off-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Wen Congyang <wency@cn.fujitsu.com>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

thp: cleanup: introduce mk_huge_pmd()

Introduce mk_huge_pmd() to simplify the code

Signed-off-by: Bob Liu <lliubbo@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Ni zhan Chen <nizhan.chen@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

thp: introduce hugepage_vma_check()

Multiple places do the same check.

Signed-off-by: Bob Liu <lliubbo@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Ni zhan Chen <nizhan.chen@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm-introduce-mm_find_pmd-fix

mm/rmap.c: In function 'try_to_unmap_cluster':
mm/rmap.c:1364:9: warning: unused variable 'pud' [-Wunused-variable]
mm/rmap.c:1363:9: warning: unused variable 'pgd' [-Wunused-variable]

Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Bob Liu <lliubbo@gmail.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Ni zhan Chen <nizhan.chen@gmail.com>
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: introduce mm_find_pmd()

Several place need to find the pmd by(mm_struct, address), so introduce a
function to simplify it.

Signed-off-by: Bob Liu <lliubbo@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Ni zhan Chen <nizhan.chen@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

thp-clean-up-__collapse_huge_page_isolate v2

mv label out of condition.

Signed-off-by: Bob Liu <lliubbo@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Ni zhan Chen <nizhan.chen@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

thp: clean up __collapse_huge_page_isolate

There are duplicated places using release_pte_pages().
And release_all_pte_pages() can be removed.

Signed-off-by: Bob Liu <lliubbo@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Ni zhan Chen <nizhan.chen@gmail.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm-use-is_enabledconfig_compaction-instead-of-compaction_build-fix

Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: use IS_ENABLED(CONFIG_COMPACTION) instead of COMPACTION_BUILD

We don't need custom COMPACTION_BUILD anymore, since we have handy
IS_ENABLED().

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: use IS_ENABLED(CONFIG_NUMA) instead of NUMA_BUILD

We don't need custom NUMA_BUILD anymore, since we have handy
IS_ENABLED().

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm, memcg: make mem_cgroup_out_of_memory() static

mem_cgroup_out_of_memory() is only referenced from within file scope, so
it can be marked static.

Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: show migration types in show_mem

This is useful to diagnose the reason for page allocation failure for
cases where there appear to be several free pages.

Example, with this alloc_pages(GFP_ATOMIC) failure:

 swapper/0: page allocation failure: order:0, mode:0x0
 ...
 Mem-info:
 Normal per-cpu:
 CPU    0: hi:   90, btch:  15 usd:  48
 CPU    1: hi:   90, btch:  15 usd:  21
 active_anon:0 inactive_anon:0 isolated_anon:0
  active_file:0 inactive_file:84 isolated_file:0
  unevictable:0 dirty:0 writeback:0 unstable:0
  free:4026 slab_reclaimable:75 slab_unreclaimable:484
  mapped:0 shmem:0 pagetables:0 bounce:0
 Normal free:16104kB min:2296kB low:2868kB high:3444kB active_anon:0kB
 inactive_anon:0kB active_file:0kB inactive_file:336kB unevictable:0kB
 isolated(anon):0kB isolated(file):0kB present:331776kB mlocked:0kB
 dirty:0kB writeback:0kB mapped:0kB shmem:0kB slab_reclaimable:300kB
 slab_unreclaimable:1936kB kernel_stack:328kB pagetables:0kB unstable:0kB
 bounce:0kB writeback_tmp:0kB pages_scanned:0 all_unreclaimable? no
 lowmem_reserve[]: 0 0

Before the patch, it's hard (for me, at least) to say why all these free
chunks weren't considered for allocation:

 Normal: 0*4kB 0*8kB 0*16kB 0*32kB 0*64kB 0*128kB 1*256kB 1*512kB
 1*1024kB 1*2048kB 3*4096kB = 16128kB

After the patch, it's obvious that the reason is that all of these are
in the MIGRATE_CMA (C) freelist:

 Normal: 0*4kB 0*8kB 0*16kB 0*32kB 0*64kB 0*128kB 1*256kB (C) 1*512kB
 (C) 1*1024kB (C) 1*2048kB (C) 3*4096kB (C) = 16128kB

Signed-off-by: Rabin Vincent <rabin.vincent@stericsson.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

writeback: remove nr_pages_dirtied arg from balance_dirty_pages_ratelimited_nr()

There is no reason to pass the nr_pages_dirtied argument, because
nr_pages_dirtied value from the caller is unused in
balance_dirty_pages_ratelimited_nr().

Signed-off-by: Namjae Jeon <linkinjeon@gmail.com>
Signed-off-by: Vivek Trivedi <vtrivedi018@gmail.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc.c: remove duplicate check

While allocating pages using buddy allocator, the compound page is
probably split up to free pages.  Under these circumstances, the compound
page should be destroyed by destroy_compound_page().  However, there is a
duplicate check to judge if the page is compound.

Remove the duplicate check since the compound_order() returns 0 when the
page doesn't have PG_head set in destroy_compound_page().  That is to say,
destroy_compound_page() needn't check PageHead().

Signed-off-by: Gavin Shan <shangw@linux.vnet.ibm.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

watchdog: trigger all-cpu backtrace when locked up and going to panic

Send an NMI to all CPUs when a lockup is detected and the lockup watchdog
code is configured to panic.  This gives us a fairly uptodate snapshot of
all CPUs in the system.

This lets us get stack trace of all CPUs which makes life easier trying to
debug a deadlock, and the NMI doesn't change anything since the next step
is a kernel panic.

Signed-off-by: Sasha Levin <sasha.levin@oracle.com>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fs/block_dev.c: no need to check inode->i_bdev in bd_forget()

Its only caller evict() has promised a non-NULL inode->i_bdev.

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fs: change return values from -EACCES to -EPERM

According to SUSv3:

[EACCES] Permission denied. An attempt was made to access a file in a way
forbidden by its file access permissions.

[EPERM] Operation not permitted. An attempt was made to perform an operation
limited to processes with appropriate privileges or to the owner of a file
or other resource.

So -EPERM should be returned if capability checks fails.

Strictly speaking this is an API change since the error code user sees is

Signed-off-by: Zhao Hongjiang <zhaohongjiang@huawei.com>
Acked-by: Jan Kara <jack@suse.cz>
Acked-by: Steven Whitehouse <swhiteho@redhat.com>
Acked-by: Ian Kent <raven@themaw.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

vfs: increment iversion when a file is truncated

When a file is truncated with truncate()/ftruncate() and then closed,
iversion is not updated.  This patch uses ATTR_SIZE flag as an indication
to increment iversion.

Mimi said:

On fput(), i_version is used to detect and flag files that have changed
and need to be re-measured in the IMA measurement policy.  When a file
is truncated with truncate()/ftruncate() and then closed, i_version is
not updated.  As a result, although the file has changed, it will not be
re-measured and added to the IMA measurement list on subsequent access.

Signed-off-by: Dmitry Kasatkin <dmitry.kasatkin@intel.com>
Acked-by: Mimi Zohar <zohar@us.ibm.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

drivers/message/fusion/mptscsih.c: missing break

This happens to do the right thing in all cases on fibre channel but not on
other media types

Signed-off-by: Alan Cox <alan@linux.intel.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Nagalakshmi Nandigama <nagalakshmi.nandigama@lsi.com>
Cc: Kashyap Desai <kashyap.desai@lsi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

h8300: select generic atomic64_t support

Rationales from Eric:

So I just looked a little deeper and it appears architectures that do
not support atomic64_t are broken.

The generic atomic64 support came in 2009 to support the perf subsystem
with the expectation that all architectures would implement atomic64
support.

Furthermore upon inspection of the kernel atomic64_t is used in a fair
number of places beyond the performance counters:

block/blk-cgroup.c
drivers/acpi/apei/
drivers/block/rbd.c
drivers/crypto/nx/nx.h
drivers/gpu/drm/radeon/radeon.h
drivers/infiniband/hw/ipath/
drivers/infiniband/hw/qib/
drivers/staging/octeon/
fs/xfs/
include/linux/perf_event.h
include/net/netfilter/nf_conntrack_acct.h
kernel/events/
kernel/trace/
net/mac80211/key.h
net/rds/

The block control group, infiniband, xfs, crypto, 802.11, netfilter.
Nothing quite so fundamental as fs/namespace.c but definitely in
multiplatform-code that should work, and is already broken on those
architecutres.

Looking at the implementation of atomic64_add_return in lib/atomic64.c the
code looks as efficient as these kinds of things get.

Which leads me to the conclusion that we need atomic64 support on all
architectures.

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

Coccinelle: Add api/d_find_alias.cocci.

Ensure that calls to d_find_alias() have a corresponding dput().

Signed-off-by: Cyril Roelandt <tipecaml@gmail.com>
Cc: Julia Lawall <Julia.Lawall@lip6.fr>
Cc: Gilles Muller <Gilles.Muller@lip6.fr>
Cc: Nicolas Palix <nicolas.palix@imag.fr>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

time: don't inline EXPORT_SYMBOL functions

How is the compiler even handling exported functions that are marked
inline?  Anyway, these shouldn't be inline because of that, so remove that
marking.

Based on a larger patch by Mark Charlebois to get LLVM to build the
kernel.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Mark Charlebois <mcharleb@qualcomm.com>
Cc: Paul Gortmaker <paul.gortmaker@windriver.com>
Cc: hank <pyu@redhat.com>
Cc: John Stultz <john.stultz@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

timeconst.pl: remove deprecated defined(@array)

The use of defined() on arrays and hashes has been deprecated since perl
5.6, but until 5.17.6 it only warned on lexicals, not package globals.

Signed-off-by: Dagfinn Ilmari Mannsåker <ilmari@ilmari.org>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

irq: tsk->comm is an array

The array check is useless so remove it.

[akpm@linux-foundation.org: remove comment, per David]
Signed-off-by: Alan Cox <alan@linux.intel.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

drivers/video/console/softcursor.c: remove redundant NULL check before kfree()

Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Cc: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

drivers-video-add-support-for-the-solomon-ssd1307-oled-controller-checkpatch-fixes

WARNING: line over 80 characters
#150: FILE: drivers/video/ssd1307fb.c:58:
+static int ssd1307fb_write_array(struct i2c_client *client, u8 type, u8* cmd, u32 len)

ERROR: "foo* bar" should be "foo *bar"
#150: FILE: drivers/video/ssd1307fb.c:58:
+static int ssd1307fb_write_array(struct i2c_client *client, u8 type, u8* cmd, u32 len)

WARNING: line over 80 characters
#175: FILE: drivers/video/ssd1307fb.c:83:
+static inline int ssd1307fb_write_cmd_array(struct i2c_client *client, u8* cmd, u32 len)

ERROR: "foo* bar" should be "foo *bar"
#175: FILE: drivers/video/ssd1307fb.c:83:
+static inline int ssd1307fb_write_cmd_array(struct i2c_client *client, u8* cmd, u32 len)

WARNING: line over 80 characters
#185: FILE: drivers/video/ssd1307fb.c:93:
+static inline int ssd1307fb_write_data_array(struct i2c_client *client, u8* cmd, u32 len)

ERROR: "foo* bar" should be "foo *bar"
#185: FILE: drivers/video/ssd1307fb.c:93:
+static inline int ssd1307fb_write_data_array(struct i2c_client *client, u8* cmd, u32 len)

WARNING: line over 80 characters
#230: FILE: drivers/video/ssd1307fb.c:138:
+ ssd1307fb_write_cmd(par->client, SSD1307FB_START_PAGE_ADDRESS + (i + 1));

WARNING: line over 80 characters
#238: FILE: drivers/video/ssd1307fb.c:146:
+ u32 index = page_length + (SSD1307FB_WIDTH * k + j) / 8;

WARNING: line over 80 characters
#281: FILE: drivers/video/ssd1307fb.c:189:
+static void ssd1307fb_fillrect(struct fb_info *info, const struct fb_fillrect *rect)

WARNING: line over 80 characters
#288: FILE: drivers/video/ssd1307fb.c:196:
+static void ssd1307fb_copyarea(struct fb_info *info, const struct fb_copyarea *area)

WARNING: line over 80 characters
#295: FILE: drivers/video/ssd1307fb.c:203:
+static void ssd1307fb_imageblit(struct fb_info *info, const struct fb_image *image)

WARNING: line over 80 characters
#322: FILE: drivers/video/ssd1307fb.c:230:
+static int __devinit ssd1307fb_probe(struct i2c_client *client, const struct i2c_device_id *id)

WARNING: line over 80 characters
#396: FILE: drivers/video/ssd1307fb.c:304:
+ dev_dbg(&client->dev, "Using PWM%d with a %dns period.\n", par->pwm->pwm, par->pwm_period);

WARNING: line over 80 characters
#430: FILE: drivers/video/ssd1307fb.c:338:
+ dev_info(&client->dev, "fb%d: %s framebuffer device registered, using %d bytes of video memory\n", info->node, info->fix.id, vmem_size);

total: 3 errors, 11 warnings, 446 lines checked

./patches/drivers-video-add-support-for-the-solomon-ssd1307-oled-controller.patch has style problems, please review.

If any of these errors are false positives, please report
them to the maintainer, see CHECKPATCH in MAINTAINERS.

Please run checkpatch prior to sending patches

Cc: Brian Lilly <brian@crystalfontz.com>
Cc: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

drivers/video: add support for the Solomon SSD1307 OLED Controller

Add support for the Solomon SSD1307 OLED controller found on the
Crystalfontz CFA10036 board.

This controller can drive a display with a resolution up to 128x39 and can
operate over I2C or SPI.

The current driver has only been tested on the CFA-10036, that is using
this controller over I2C to driver a 96x16 OLED screen.

Signed-off-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Cc: Brian Lilly <brian@crystalfontz.com>
Cc: Greg KH <gregkh@linux-foundation.org>
Cc: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Thomas Petazzoni <thomas@free-electrons.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

pcmcia: move unbind/rebind into dev_pm_ops.complete

Move the device rebind procedures for cardbus devices from the pm.resume
into the pm.complete callback.

The reason for moving the code is: "[...] The PM code needs to send
suspend and resume messages to every device in the right order, and it
can't do that if new devices are being added at the same time.  [...]"

However the situation really isn't quite that rigid.  In particular,
adding new children during a resume callback shouldn't cause much of
problem because the children don't need to be resumed anyway (since they
were never suspended).  On the other hand, if you do it you will get a
dev_warn() from the PM core, something like 'parent should not be
sleeping'.

Still, it is considered bad form and should be avoided if possible."

(Alan Stern's full comment about the topic can
be found here: <https://lkml.org/lkml/2012/7/10/254>)

Signed-off-by: Christian Lamparter <chunkeey@googlemail.com>
Cc: Dominik Brodowski <linux@dominikbrodowski.net>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: Greg KH <greg@kroah.com>
Acked-by: "Rafael J. Wysocki" <rjw@sisk.pl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

ceph: fix dentry reference leak in ceph_encode_fh()

dput() was not called in the error path.

Signed-off-by: Cyril Roelandt <tipecaml@gmail.com>
Cc: Sage Weil <sage@inktank.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

audit: catch possible NULL audit buffers

It's possible for audit_log_start() to return NULL.  Handle it in the
various callers.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Eric Paris <eparis@redhat.com>
Cc: Jeff Layton <jlayton@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Julien Tinnes <jln@google.com>
Cc: Will Drewry <wad@google.com>
Cc: Steve Grubb <sgrubb@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

audit: create explicit AUDIT_SECCOMP event type

The seccomp path was using AUDIT_ANOM_ABEND from when seccomp mode 1 could
only kill a process.  While we still want to make sure an audit record is
forced on a kill, this should use a separate record type since seccomp
mode 2 introduces other behaviors.  In the case of "handled" behaviors
(process wasn't killed), only emit a record if the process is under
inspection.

Signed-off-by: Kees Cook <keescook@chromium.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Eric Paris <eparis@redhat.com>
Cc: Jeff Layton <jlayton@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Julien Tinnes <jln@google.com>
Cc: Will Drewry <wad@google.com>
Cc: Steve Grubb <sgrubb@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

x86: make 'mem=' option to work for efi platform

Current mem boot option only can work for non efi environment.  If the
user specifies add_efi_memmap, it cannot work for efi environment.  In the
efi environment, we call e820_add_region() to add the memory map.  So we
can modify __e820_add_region() and the mem boot option can work for efi
environment.

Note: Only E820_RAM is limited, and BOOT_SERVICES_{CODE,DATA} are always
mapped(If its address >= mem_limit, the memory won't be freed in
efi_free_boot_services()).

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: Matt Fleming <matt.fleming@intel.com>
Cc: Rob Landley <rob@landley.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Yasuaki ISIMATU <isimatu.yasuaki@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

x86: fix the argument passed to sync_global_pgds()

The address range of sync_global_pgds() should be [start, end], but we
pass [start, end) to this function.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Jiang Liu <liuj97@gmail.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

x86: convert update_mmu_cache() and update_mmu_cache_pmd() to functions

Converting macros to functions unhide type problems before changes will
be integrated and trigger problems on other architectures.

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

olpc: fix olpc-xo1-sci.c build errors

Fix build errors when CONFIG_INPUT=m.  This is not pretty, but all of the
OLPC kconfig options are bool instead of tristate.

arch/x86/built-in.o: In function `send_lid_state':
olpc-xo1-sci.c:(.text+0x1d323): undefined reference to `input_event'
olpc-xo1-sci.c:(.text+0x1d338): undefined reference to `input_event'
arch/x86/built-in.o: In function `free_ebook_switch':
olpc-xo1-sci.c:(.text+0x1d529): undefined reference to `input_unregister_device'
olpc-xo1-sci.c:(.text+0x1d533): undefined reference to `input_free_device'
arch/x86/built-in.o: In function `free_power_button':
olpc-xo1-sci.c:(.text+0x1d549): undefined reference to `input_unregister_device'
olpc-xo1-sci.c:(.text+0x1d553): undefined reference to `input_free_device'
arch/x86/built-in.o: In function `send_ebook_state':
olpc-xo1-sci.c:(.text+0x1d632): undefined reference to `input_event'
olpc-xo1-sci.c:(.text+0x1d647): undefined reference to `input_event'
arch/x86/built-in.o: In function `xo1_sci_intr':
olpc-xo1-sci.c:(.text+0x1d78e): undefined reference to `input_event'
olpc-xo1-sci.c:(.text+0x1d7a3): undefined reference to `input_event'
olpc-xo1-sci.c:(.text+0x1d7be): undefined reference to `input_event'
arch/x86/built-in.o:olpc-xo1-sci.c:(.text+0x1d7d3): more undefined references to `input_event' follow
arch/x86/built-in.o: In function `free_lid_switch':
olpc-xo1-sci.c:(.text+0x1d7fd): undefined reference to `input_unregister_device'
olpc-xo1-sci.c:(.text+0x1d807): undefined reference to `input_free_device'
arch/x86/built-in.o: In function `setup_lid_switch':
olpc-xo1-sci.c:(.devinit.text+0x155): undefined reference to `input_allocate_device'
olpc-xo1-sci.c:(.devinit.text+0x1a4): undefined reference to `input_register_device'
olpc-xo1-sci.c:(.devinit.text+0x1ce): undefined reference to `input_unregister_device'
olpc-xo1-sci.c:(.devinit.text+0x1d8): undefined reference to `input_free_device'
arch/x86/built-in.o: In function `xo1_sci_probe':
olpc-xo1-sci.c:(.devinit.text+0x235): undefined reference to `input_allocate_device'
olpc-xo1-sci.c:(.devinit.text+0x285): undefined reference to `input_register_device'
olpc-xo1-sci.c:(.devinit.text+0x299): undefined reference to `input_free_device'
olpc-xo1-sci.c:(.devinit.text+0x2e1): undefined reference to `input_register_device'
olpc-xo1-sci.c:(.devinit.text+0x2f5): undefined reference to `input_free_device'
olpc-xo1-sci.c:(.devinit.text+0x54c): undefined reference to `input_allocate_device'

In the long run, fixing this driver kconfig to be tristate instead of bool
would be a very good change.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Cc: Andres Salomon <dilinger@queued.net>
Cc: Chris Ball <cjb@laptop.org>
Cc: Jon Nettleton <jon.nettleton@gmail.com>
Cc: Daniel Drake <dsd@laptop.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

arch/x86/platform/uv: fix incorrect tlb flush all issue

The flush tlb optimization code has logical issue on UV platform.  It
doesn't flush the full range at all, since it simply ignores its 'end'
parameter (and hence also the "all" indicator) in uv_flush_tlb_others()
function.

Cliff's notes:

: I tested the patch on a UV.  It has the effect of either clearing 1 or all
: TLBs in a cpu.  I added some debugging to test for the cases when clearing
: all TLBs is overkill, and in practice it happens very seldom.

Reported-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Alex Shi <alex.shi@intel.com>
Signed-off-by: Cliff Wickman <cpw@sgi.com>
Tested-by: Cliff Wickman <cpw@sgi.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

arch/x86/tools/insn_sanity.c: identify source of messages

The kernel build prints:

  Building modules, stage 2.
  TEST    posttest
  MODPOST 3821 modules
  TEST    posttest
Success: decoded and checked 1000000 random instructions with 0 errors (seed:0xaac4bc47)
  CC      arch/x86/boot/a20.o
  CC      arch/x86/boot/cmdline.o
  AS      arch/x86/boot/copy.o
  HOSTCC  arch/x86/boot/mkcpustr
  CC      arch/x86/boot/cpucheck.o
  CC      arch/x86/boot/early_serial_console.o

which is irritating because you don't know what program is proudly
pronouncing its success.

So, as described in "console mode programming user interface guidelines
version 101" which doesn't exist, change this program to identify the
source of its messages.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

x86 numa: don't check if node is NUMA_NO_NODE

If we aren't debugging per_cpu maps, the cpu's node is stored in per_cpu
variable numa_node.  If `node' is NUMA_NO_NODE, it means the caller wants
to clear the cpu's node.  So we should also call set_cpu_numa_node() in
this case.

Signed-off-by: Wen Congyang <wency@cn.fujitsu.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Pavel Machek <pavel@ucw.cz>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

arch/x86/platform/iris/iris.c: register a platform device and a platform driver

This makes the iris driver use the platform API, so it is properly exposed
in /sys.

[akpm@linux-foundation.org: remove commented-out code, add missing space to printk, clean up code layout]
Signed-off-by: Shérab <Sebastien.Hinderer@ens-lyon.org>
Cc: Len Brown <lenb@kernel.org>
Cc: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fs/block_dev.c: page cache wrongly left invalidated after revalidate_disk()

We found that bdev->bd_invalidated was left set once revalidate_disk() is
called, which results in page cache flush every time that device is open.

Specifically, we found this problem in MD block device.  Once we resize a
MD device, mdadm --monitor periodically flush all page cache for that
device every 60 or 1000 seconds when it opens the device.

This bug lies since at least 3.2.0 till the latest kernel(3.6.2).
Patch is attached.

The following steps will reproduce the problem.

1. prepair a block device(ex. /dev/sdb).
2. create two partitions.

sudo parted /dev/sdb
mklabel gpt
mkpart primary 0% 50%
mkpart primary 50% 100%

3. create a md device.

sudo mdadm -C /dev/md/hoge -l 1 -n 2 -e 1.2 --assume-clean --auto=md \
 --symlink=no /dev/sdb1 /dev/sdb2

4. create file system and mount it

sudo mkfs.ext3 /dev/md/hoge
sudo mkdir /mnt/test
sudo mount /dev/md/hoge /mnt/test

5. try to resize the device

sudo mdadm -G /dev/md/hoge --size=max

6. create a file to fill file cache.

sudo dd if=/dev/urandom of=/mnt/test/data bs=1M count=10
and verity the current status of file by free command.

7. mdadm monitor will open the md device every 1000 seconds
and you will find all file cache on the device are cleared.

The timing can be reduced by the following steps.

a) kill mdadm and restart it with --delay option
/sbin/mdadm --monitor --delay=30 --pid-file /var/run/mdadm/monitor.pid \
 --daemonise --scan --syslog

or open the md device directly.

sudo dd if=/dev/md/hoge of=/dev/null bs=4096 count=1

Signed-off-by: MITSUNARI Shigeo <herumi@nifty.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Jeff Moyer <jmoyer@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

vfs: d_obtain_alias() needs to use "/" as default name.

NFS appears to use d_obtain_alias() to create the root dentry rather than
d_make_root.  This can cause 'prepend_path()' to complain that the root
has a weird name if an NFS filesystem is lazily unmounted.  e.g.  if
"/mnt" is an NFS mount then

 { cd /mnt; umount -l /mnt ; ls -l /proc/self/cwd; }

will cause a WARN message like
   WARNING: at /home/git/linux/fs/dcache.c:2624 prepend_path+0x1d7/0x1e0()
   ...
   Root dentry has weird name <>

to appear in kernel logs.

So change d_obtain_alias() to use "/" rather than "" as the anonymous
name.

Signed-off-by: NeilBrown <neilb@suse.de>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

CRIS: Fix I/O macros

The inb/outb macros for CRIS are broken from a number of points of view,
missing () around parameters and they have an unprotected if statement in
them.  This was breaking the compile of IPMI on CRIS and thus I was being
annoyed by build regressions, so I fixed them.

Plus I don't think they would have worked at all, since the data values
were missing "&" and the outsl had a "3" instead of a "4" for the size.
From what I can tell, this stuff is not used at all, so this can't be any
more broken than it was before, anyway.

Signed-off-by: Corey Minyard <cminyard@mvista.com>
Cc: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <starvik@axis.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

thp: fix update_mmu_cache_pmd() calls

update_mmu_cache_pmd() takes pointer to pmd_t as third, not pmd_t.

mm/huge_memory.c: In function 'do_huge_pmd_numa_page':
mm/huge_memory.c:825:2: error: incompatible type for argument 3 of 'update_mmu_cache_pmd'
In file included from include/linux/mm.h:44:0,
                 from mm/huge_memory.c:8:
arch/mips/include/asm/pgtable.h:385:20: note: expected 'struct pmd_t *' but argument is of type 'pmd_t'
mm/huge_memory.c:895:2: error: incompatible type for argument 3 of 'update_mmu_cache_pmd'
In file included from include/linux/mm.h:44:0,
                 from mm/huge_memory.c:8:
arch/mips/include/asm/pgtable.h:385:20: note: expected 'struct pmd_t *' but argument is of type 'pmd_t'

Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>