mm, thp: remove __GFP_NORETRY from khugepaged and madvised allocations

After the previous patch, we can distinguish costly allocations that
should be really lightweight, such as THP page faults, with
__GFP_NORETRY.  This means we don't need to recognize khugepaged
allocations via PF_KTHREAD anymore.  We can also change THP page faults
in areas where madvise(MADV_HUGEPAGE) was used to try as hard as
khugepaged, as the process has indicated that it benefits from THP's and
is willing to pay some initial latency costs.

We can also make the flags handling less cryptic by distinguishing
GFP_TRANSHUGE_LIGHT (no reclaim at all, default mode in page fault) from
GFP_TRANSHUGE (only direct reclaim, khugepaged default).  Adding
__GFP_NORETRY or __GFP_KSWAPD_RECLAIM is done where needed.

The patch effectively changes the current GFP_TRANSHUGE users as
follows:

* get_huge_zero_page() - the zero page lifetime should be relatively
  long and it's shared by multiple users, so it's worth spending some
  effort on it.  We use GFP_TRANSHUGE, and __GFP_NORETRY is not added.
  This also restores direct reclaim to this allocation, which was
  unintentionally removed by commit e4a49efe4e7e ("mm: thp: set THP defrag
  by default to madvise and add a stall-free defrag option")

* alloc_hugepage_khugepaged_gfpmask() - this is khugepaged, so latency
  is not an issue.  So if khugepaged "defrag" is enabled (the default), do
  reclaim via GFP_TRANSHUGE without __GFP_NORETRY.  We can remove the
  PF_KTHREAD check from page alloc.

  As a side-effect, khugepaged will now no longer check if the initial
  compaction was deferred or contended.  This is OK, as khugepaged sleep
  times between collapsion attempts are long enough to prevent noticeable
  disruption, so we should allow it to spend some effort.

* migrate_misplaced_transhuge_page() - already was masking out
  __GFP_RECLAIM, so just convert to GFP_TRANSHUGE_LIGHT which is
  equivalent.

* alloc_hugepage_direct_gfpmask() - vma's with VM_HUGEPAGE (via madvise)
  are now allocating without __GFP_NORETRY.  Other vma's keep using
  __GFP_NORETRY if direct reclaim/compaction is at all allowed (by default
  it's allowed only for madvised vma's).  The rest is conversion to
  GFP_TRANSHUGE(_LIGHT).

[mhocko@suse.com: suggested GFP_TRANSHUGE_LIGHT]
Link: http://lkml.kernel.org/r/20160721073614.24395-7-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: make THP-specific decisions more generic

Since THP allocations during page faults can be costly, extra decisions
are employed for them to avoid excessive reclaim and compaction, if the
initial compaction doesn't look promising.  The detection has never been
perfect as there is no gfp flag specific to THP allocations.  At this
moment it checks the whole combination of flags that makes up
GFP_TRANSHUGE, and hopes that no other users of such combination exist,
or would mind being treated the same way.  Extra care is also taken to
separate allocations from khugepaged, where latency doesn't matter that
much.

It is however possible to distinguish these allocations in a simpler and
more reliable way.  The key observation is that after the initial
compaction followed by the first iteration of "standard"
reclaim/compaction, both __GFP_NORETRY allocations and costly
allocations without __GFP_REPEAT are declared as failures:

        /* Do not loop if specifically requested */
        if (gfp_mask & __GFP_NORETRY)
                goto nopage;

        /*
         * Do not retry costly high order allocations unless they are
         * __GFP_REPEAT
         */
        if (order > PAGE_ALLOC_COSTLY_ORDER && !(gfp_mask & __GFP_REPEAT))
                goto nopage;

This means we can further distinguish allocations that are costly order
*and* additionally include the __GFP_NORETRY flag.  As it happens,
GFP_TRANSHUGE allocations do already fall into this category.  This will
also allow other costly allocations with similar high-order benefit vs
latency considerations to use this semantic.  Furthermore, we can
distinguish THP allocations that should try a bit harder (such as from
khugepageed) by removing __GFP_NORETRY, as will be done in the next
patch.

Link: http://lkml.kernel.org/r/20160721073614.24395-6-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: restructure direct compaction handling in slowpath

The retry loop in __alloc_pages_slowpath is supposed to keep trying
reclaim and compaction (and OOM), until either the allocation succeeds,
or returns with failure.  Success here is more probable when reclaim
precedes compaction, as certain watermarks have to be met for compaction
to even try, and more free pages increase the probability of compaction
success.  On the other hand, starting with light async compaction (if
the watermarks allow it), can be more efficient, especially for smaller
orders, if there's enough free memory which is just fragmented.

Thus, the current code starts with compaction before reclaim, and to
make sure that the last reclaim is always followed by a final
compaction, there's another direct compaction call at the end of the
loop.  This makes the code hard to follow and adds some duplicated
handling of migration_mode decisions.  It's also somewhat inefficient
that even if reclaim or compaction decides not to retry, the final
compaction is still attempted.  Some gfp flags combination also shortcut
these retry decisions by "goto noretry;", making it even harder to
follow.

This patch attempts to restructure the code with only minimal functional
changes.  The call to the first compaction and THP-specific checks are
now placed above the retry loop, and the "noretry" direct compaction is
removed.

The initial compaction is additionally restricted only to costly orders,
as we can expect smaller orders to be held back by watermarks, and only
larger orders to suffer primarily from fragmentation.  This better
matches the checks in reclaim's shrink_zones().

There are two other smaller functional changes.  One is that the upgrade
from async migration to light sync migration will always occur after the
initial compaction.  This is how it has been until recent patch "mm,
oom: protect !costly allocations some more", which introduced upgrading
the mode based on COMPACT_COMPLETE result, but kept the final compaction
always upgraded, which made it even more special.  It's better to return
to the simpler handling for now, as migration modes will be further
modified later in the series.

The second change is that once both reclaim and compaction declare it's
not worth to retry the reclaim/compact loop, there is no final
compaction attempt.  As argued above, this is intentional.  If that
final compaction were to succeed, it would be due to a wrong retry
decision, or simply a race with somebody else freeing memory for us.

The main outcome of this patch should be simpler code.  Logically, the
initial compaction without reclaim is the exceptional case to the
reclaim/compaction scheme, but prior to the patch, it was the last loop
iteration that was exceptional.  Now the code matches the logic better.
The change also enable the following patches.

Link: http://lkml.kernel.org/r/20160721073614.24395-5-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: don't retry initial attempt in slowpath

After __alloc_pages_slowpath() sets up new alloc_flags and wakes up
kswapd, it first tries get_page_from_freelist() with the new
alloc_flags, as it may succeed e.g. due to using min watermark instead
of low watermark.  It makes sense to to do this attempt before adjusting
zonelist based on alloc_flags/gfp_mask, as it's still relatively a fast
path if we just wake up kswapd and successfully allocate.

This patch therefore moves the initial attempt above the retry label and
reorganizes a bit the part below the retry label.  We still have to
attempt get_page_from_freelist() on each retry, as some allocations
cannot do that as part of direct reclaim or compaction, and yet are not
allowed to fail (even though they do a WARN_ON_ONCE() and thus should
not exist).  We can reuse the call meant for ALLOC_NO_WATERMARKS attempt
and just set alloc_flags to ALLOC_NO_WATERMARKS if the context allows
it.  As a side-effect, the attempts from direct reclaim/compaction will
also no longer obey watermarks once this is set, but there's little harm
in that.

Kswapd wakeups are also done on each retry to be safe from potential
races resulting in kswapd going to sleep while a process (that may not
be able to reclaim by itself) is still looping.

Link: http://lkml.kernel.org/r/20160721073614.24395-4-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: set alloc_flags only once in slowpath

In __alloc_pages_slowpath(), alloc_flags doesn't change after it's
initialized, so move the initialization above the retry: label.  Also
make the comment above the initialization more descriptive.

The only exception in the alloc_flags being constant is
ALLOC_NO_WATERMARKS, which may change due to TIF_MEMDIE being set on the
allocating thread.  We can fix this, and make the code simpler and a bit
more effective at the same time, by moving the part that determines
ALLOC_NO_WATERMARKS from gfp_to_alloc_flags() to gfp_pfmemalloc_allowed().

This means we don't have to mask out ALLOC_NO_WATERMARKS in numerous
places in __alloc_pages_slowpath() anymore.  The only two tests for the
flag can instead call gfp_pfmemalloc_allowed().

Link: http://lkml.kernel.org/r/20160721073614.24395-3-vbabka@suse.cz
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lib/stackdepot.c: use __GFP_NOWARN for stack allocations

This (large, atomic) allocation attempt can fail.  We expect and handle
that, so avoid the scary warning.

Link: http://lkml.kernel.org/r/20160720151905.GB19146@node.shutemov.name
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Rik van Riel <riel@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, kasan: switch SLUB to stackdepot, enable memory quarantine for SLUB

For KASAN builds:
 - switch SLUB allocator to using stackdepot instead of storing the
   allocation/deallocation stacks in the objects;
 - change the freelist hook so that parts of the freelist can be put
   into the quarantine.

[aryabinin@virtuozzo.com: fixes]
Link: http://lkml.kernel.org/r/1468601423-28676-1-git-send-email-aryabinin@virtuozzo.com
Link: http://lkml.kernel.org/r/1468347165-41906-3-git-send-email-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Cc: Andrey Konovalov <adech.fo@gmail.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Steven Rostedt (Red Hat) <rostedt@goodmis.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Kuthonuzo Luruo <kuthonuzo.luruo@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, kasan: account for object redzone in SLUB's nearest_obj()

When looking up the nearest SLUB object for a given address, correctly
calculate its offset if SLAB_RED_ZONE is enabled for that cache.

Previously, when KASAN had detected an error on an object from a cache
with SLAB_RED_ZONE set, the actual start address of the object was
miscalculated, which led to random stacks having been reported.

When looking up the nearest SLUB object for a given address, correctly
calculate its offset if SLAB_RED_ZONE is enabled for that cache.

Fixes: 7ed2f9e663854db ("mm, kasan: SLAB support")
Link: http://lkml.kernel.org/r/1468347165-41906-2-git-send-email-glider@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Cc: Andrey Konovalov <adech.fo@gmail.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Steven Rostedt (Red Hat) <rostedt@goodmis.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kostya Serebryany <kcc@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Kuthonuzo Luruo <kuthonuzo.luruo@hpe.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: fix use-after-free if memory allocation failed in vma_adjust()

There's one case when vma_adjust() expands the vma, overlapping with
*two* next vma.  See case 6 of mprotect, described in the comment to
vma_merge().

To handle this (and only this) situation we iterate twice over main part
of the function.  See "goto again".

Vegard reported[1] that he sees out-of-bounds access complain from
KASAN, if anon_vma_clone() on the *second* iteration fails.

This happens because we free 'next' vma by the end of first iteration
and don't have a way to undo this if anon_vma_clone() fails on the
second iteration.

The solution is to do all required allocations upfront, before we touch
vmas.

The allocation on the second iteration is only required if first two
vmas don't have anon_vma, but third does.  So we need, in total, one
anon_vma_clone() call.

It's easy to adjust 'exporter' to the third vma for such case.

[1] http://lkml.kernel.org/r/1469514843-23778-1-git-send-email-vegard.nossum@oracle.com

Link: http://lkml.kernel.org/r/1469625255-126641-1-git-send-email-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Vegard Nossum <vegard.nossum@oracle.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zsmalloc: Delete an unnecessary check before the function call "iput"

iput() tests whether its argument is NULL and then returns immediately.
Thus the test around the call is not needed.

This issue was detected by using the Coccinelle software.

Link: http://lkml.kernel.org/r/559cf499-4a01-25f9-c87f-24d906626a57@users.sourceforge.net
Signed-off-by: Markus Elfring <elfring@users.sourceforge.net>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memblock.c: fix index adjustment error in __next_mem_range_rev()

Fix region index adjustment error when parameter type_b of
__next_mem_range_rev() == NULL.

Signed-off-by: zijun_hu <zijun_hu@htc.com>
Cc: Alexander Kuleshov <kuleshovmail@gmail.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Wei Yang <weiyang@linux.vnet.ibm.com>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Richard Leitner <dev@g0hl1n.net>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mem-hotplug: alloc new page from a nearest neighbor node when mem-offline

If we offline a node, alloc the new page from a nearest neighbor node
instead of the current node or other remote nodes, because re-migrate is
a waste of time and the distance of the remote nodes is often very
large.

Also use GFP_HIGHUSER_MOVABLE to alloc new page if the zone is movable
zone or highmem zone.

Link: http://lkml.kernel.org/r/5795E18B.5060302@huawei.com
Signed-off-by: Xishi Qiu <qiuxishi@huawei.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: optimize copy_page_to/from_iter_iovec

copy_page_to_iter_iovec() and copy_page_from_iter_iovec() copy some data
to userspace or from userspace.  These functions have a fast path where
they map a page using kmap_atomic and a slow path where they use kmap.

kmap is slower than kmap_atomic, so the fast path is preferred.

However, on kernels without highmem support, kmap just calls
page_address, so there is no need to avoid kmap.  On kernels without
highmem support, the fast path just increases code size (and cache
footprint) and it doesn't improve copy performance in any way.

This patch enables the fast path only if CONFIG_HIGHMEM is defined.

Code size reduced by this patch:
  x86 (without highmem)   928
  x86-64   960
  sparc64   848
  alpha  1136
  pa-risc  1200

[akpm@linux-foundation.org: use IS_ENABLED(), per Andi]
Link: http://lkml.kernel.org/r/alpine.LRH.2.02.1607221711410.4818@file01.intranet.prod.int.rdu2.redhat.com
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Andi Kleen <andi@firstfloor.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add cond_resched() to generic_swapfile_activate()

generic_swapfile_activate() can take quite long time, it iterates over
all blocks of a file, so add cond_resched to it.  I observed about 1
second stalls when activating a swapfile that was almost unfragmented -
this patch fixes it.

Link: http://lkml.kernel.org/r/alpine.LRH.2.02.1607221710580.4818@file01.intranet.prod.int.rdu2.redhat.com
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Revert "mm, mempool: only set __GFP_NOMEMALLOC if there are free elements"

This reverts commit f9054c70d28b ("mm, mempool: only set __GFP_NOMEMALLOC
if there are free elements").

There has been a report about OOM killer invoked when swapping out to a
dm-crypt device.  The primary reason seems to be that the swapout out IO
managed to completely deplete memory reserves.  Ondrej was able to
bisect and explained the issue by pointing to f9054c70d28b ("mm,
mempool: only set __GFP_NOMEMALLOC if there are free elements").

The reason is that the swapout path is not throttled properly because
the md-raid layer needs to allocate from the generic_make_request path
which means it allocates from the PF_MEMALLOC context.  dm layer uses
mempool_alloc in order to guarantee a forward progress which used to
inhibit access to memory reserves when using page allocator.  This has
changed by f9054c70d28b ("mm, mempool: only set __GFP_NOMEMALLOC if
there are free elements") which has dropped the __GFP_NOMEMALLOC
protection when the memory pool is depleted.

If we are running out of memory and the only way forward to free memory
is to perform swapout we just keep consuming memory reserves rather than
throttling the mempool allocations and allowing the pending IO to
complete up to a moment when the memory is depleted completely and there
is no way forward but invoking the OOM killer.  This is less than
optimal.

The original intention of f9054c70d28b was to help with the OOM
situations where the oom victim depends on mempool allocation to make a
forward progress.  David has mentioned the following backtrace:

  schedule
  schedule_timeout
  io_schedule_timeout
  mempool_alloc
  __split_and_process_bio
  dm_request
  generic_make_request
  submit_bio
  mpage_readpages
  ext4_readpages
  __do_page_cache_readahead
  ra_submit
  filemap_fault
  handle_mm_fault
  __do_page_fault
  do_page_fault
  page_fault

We do not know more about why the mempool is depleted without being
replenished in time, though.  In any case the dm layer shouldn't depend
on any allocations outside of the dedicated pools so a forward progress
should be guaranteed.  If this is not the case then the dm should be
fixed rather than papering over the problem and postponing it to later
by accessing more memory reserves.

mempools are a mechanism to maintain dedicated memory reserves to
guaratee forward progress.  Allowing them an unbounded access to the
page allocator memory reserves is going against the whole purpose of
this mechanism.

Bisected by Ondrej Kozina.

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/20160721145309.GR26379@dhcp22.suse.cz
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Ondrej Kozina <okozina@redhat.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: NeilBrown <neilb@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Mikulas Patocka <mpatocka@redhat.com>
Cc: Ondrej Kozina <okozina@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: Mel Gorman <mgorman@suse.de>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, compaction: don't isolate PageWriteback pages in MIGRATE_SYNC_LIGHT mode

At present MIGRATE_SYNC_LIGHT is allowing __isolate_lru_page() to
isolate a PageWriteback page, which __unmap_and_move() then rejects with
-EBUSY: of course the writeback might complete in between, but that's
not what we usually expect, so probably better not to isolate it.

When tested by stress-highalloc from mmtests, this has reduced the
number of page migrate failures by 60-70%.

Link: http://lkml.kernel.org/r/20160721073614.24395-2-vbabka@suse.cz
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: hwpoison: remove incorrect comments

dequeue_hwpoisoned_huge_page() can be called without page lock hold, so
let's remove incorrect comment.

The reason why the page lock is not really needed is that
dequeue_hwpoisoned_huge_page() checks page_huge_active() inside
hugetlb_lock, which allows us to avoid trying to dequeue a hugepage that
are just allocated but not linked to active list yet, even without
taking page lock.

Link: http://lkml.kernel.org/r/20160720092901.GA15995@www9186uo.sakura.ne.jp
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Reported-by: Zhan Chen <zhanc1@andrew.cmu.edu>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

make __section_nr() more efficient

When CONFIG_SPARSEMEM_EXTREME is disabled, __section_nr can get the
section number with a subtraction directly.

Link: http://lkml.kernel.org/r/1468988310-11560-1-git-send-email-zhouchengming1@huawei.com
Signed-off-by: Zhou Chengming <zhouchengming1@huawei.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Hanjun Guo <guohanjun@huawei.com>
Cc: Li Bin <huawei.libin@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kmemleak: don't hang if user disables scanning early

If the user tries to disable automatic scanning early in the boot
process using e.g.:

  echo scan=off > /sys/kernel/debug/kmemleak

then this command will hang until SECS_FIRST_SCAN (= 60) seconds have
elapsed, even though the system is fully initialised.

We can fix this using interruptible sleep and checking if we're supposed
to stop whenever we wake up (like the rest of the code does).

Link: http://lkml.kernel.org/r/1468835005-2873-1-git-send-email-vegard.nossum@oracle.com
Signed-off-by: Vegard Nossum <vegard.nossum@oracle.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64:acpi: fix the acpi alignment exception when 'mem=' specified

When booting an ACPI enabled kernel with 'mem=x', there is the
possibility that ACPI data regions from the firmware will lie above the
memory limit.  Ordinarily these will be removed by
memblock_enforce_memory_limit(.).

Unfortunately, this means that these regions will then be mapped by
acpi_os_ioremap(.) as device memory (instead of normal) thus unaligned
accessess will then provoke alignment faults.

In this patch we adopt memblock_mem_limit_remove_map instead, and this
preserves these ACPI data regions (marked NOMAP) thus ensuring that
these regions are not mapped as device memory.

For example, below is an alignment exception observed on ARM platform
when booting the kernel with 'acpi=on mem=8G':

  ...
  Unable to handle kernel paging request at virtual address ffff0000080521e7
  pgd = ffff000008aa0000
  [ffff0000080521e7] *pgd=000000801fffe003, *pud=000000801fffd003, *pmd=000000801fffc003, *pte=00e80083ff1c1707
  Internal error: Oops: 96000021 [#1] PREEMPT SMP
  Modules linked in:
  CPU: 1 PID: 1 Comm: swapper/0 Not tainted 4.7.0-rc3-next-20160616+ #172
  Hardware name: AMD Overdrive/Supercharger/Default string, BIOS ROD1001A 02/09/2016
  task: ffff800001ef0000 ti: ffff800001ef8000 task.ti: ffff800001ef8000
  PC is at acpi_ns_lookup+0x520/0x734
  LR is at acpi_ns_lookup+0x4a4/0x734
  pc : [<ffff0000083b8b10>] lr : [<ffff0000083b8a94>] pstate: 60000045
  sp : ffff800001efb8b0
  x29: ffff800001efb8c0 x28: 000000000000001b
  x27: 0000000000000001 x26: 0000000000000000
  x25: ffff800001efb9e8 x24: ffff000008a10000
  x23: 0000000000000001 x22: 0000000000000001
  x21: ffff000008724000 x20: 000000000000001b
  x19: ffff0000080521e7 x18: 000000000000000d
  x17: 00000000000038ff x16: 0000000000000002
  x15: 0000000000000007 x14: 0000000000007fff
  x13: ffffff0000000000 x12: 0000000000000018
  x11: 000000001fffd200 x10: 00000000ffffff76
  x9 : 000000000000005f x8 : ffff000008725fa8
  x7 : ffff000008a8df70 x6 : ffff000008a8df70
  x5 : ffff000008a8d000 x4 : 0000000000000010
  x3 : 0000000000000010 x2 : 000000000000000c
  x1 : 0000000000000006 x0 : 0000000000000000
  ...
    acpi_ns_lookup+0x520/0x734
    acpi_ds_load1_begin_op+0x174/0x4fc
    acpi_ps_build_named_op+0xf8/0x220
    acpi_ps_create_op+0x208/0x33c
    acpi_ps_parse_loop+0x204/0x838
    acpi_ps_parse_aml+0x1bc/0x42c
    acpi_ns_one_complete_parse+0x1e8/0x22c
    acpi_ns_parse_table+0x8c/0x128
    acpi_ns_load_table+0xc0/0x1e8
    acpi_tb_load_namespace+0xf8/0x2e8
    acpi_load_tables+0x7c/0x110
    acpi_init+0x90/0x2c0
    do_one_initcall+0x38/0x12c
    kernel_init_freeable+0x148/0x1ec
    kernel_init+0x10/0xec
    ret_from_fork+0x10/0x40
  Code: b9009fbc 2a00037b 36380057 3219037b (b9400260)
  ---[ end trace 03381e5eb0a24de4 ]---
  Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b

With 'efi=debug', we can see those ACPI regions loaded by firmware on
that board as:

  efi:   0x0083ff185000-0x0083ff1b4fff [Reserved           |   |  |  |  |  |  |  |   |WB|WT|WC|UC]*
  efi:   0x0083ff1b5000-0x0083ff1c2fff [ACPI Reclaim Memory|   |  |  |  |  |  |  |   |WB|WT|WC|UC]*
  efi:   0x0083ff223000-0x0083ff224fff [ACPI Memory NVS    |   |  |  |  |  |  |  |   |WB|WT|WC|UC]*

Link: http://lkml.kernel.org/r/1468475036-5852-3-git-send-email-dennis.chen@arm.com
Acked-by: Steve Capper <steve.capper@arm.com>
Signed-off-by: Dennis Chen <dennis.chen@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Rafael J. Wysocki <rafael@kernel.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Kaly Xin <kaly.xin@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memblock.c: add new infrastructure to address the mem limit issue

In some cases, memblock is queried by kernel to determine whether a
specified address is RAM or not.  For example, the ACPI core needs this
information to determine which attributes to use when mapping ACPI
regions(acpi_os_ioremap).  Use of incorrect memory types can result in
faults, data corruption, or other issues.

Removing memory with memblock_enforce_memory_limit() throws away this
information, and so a kernel booted with 'mem=' may suffer from the
issues described above.  To avoid this, we need to keep those NOMAP
regions instead of removing all above the limit, which preserves the
information we need while preventing other use of those regions.

This patch adds new infrastructure to retain all NOMAP memblock regions
while removing others, to cater for this.

Link: http://lkml.kernel.org/r/1468475036-5852-2-git-send-email-dennis.chen@arm.com
Signed-off-by: Dennis Chen <dennis.chen@arm.com>
Acked-by: Steve Capper <steve.capper@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Rafael J. Wysocki <rafael@kernel.org>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Cc: Kaly Xin <kaly.xin@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

printk: when dumping regs, show the stack, not thread_info

We currently show:

  task: <current> ti: <current_thread_info()> task.ti: <task_thread_info(current)>"

"ti" and "task.ti" are redundant, and neither is actually what we want
to show, which the the base of the thread stack.  Change the display to
show the stack pointer explicitly.

Link: http://lkml.kernel.org/r/543ac5bd66ff94000a57a02e11af7239571a3055.1468523549.git.luto@kernel.org
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kdb: use task_cpu() instead of task_thread_info()->cpu

We'll need this cleanup to make the cpu field in thread_info be
optional.

Link: http://lkml.kernel.org/r/da298328dc77ea494576c2f20a934218e758a6fa.1468523549.git.luto@kernel.org
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Jason Wessel <jason.wessel@windriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: fix memcg stack accounting for sub-page stacks

We should account for stacks regardless of stack size, and we need to
account in sub-page units if THREAD_SIZE < PAGE_SIZE.  Change the units
to kilobytes and Move it into account_kernel_stack().

Fixes: 12580e4b54ba8 ("mm: memcontrol: report kernel stack usage in cgroup2 memory.stat")
Link: http://lkml.kernel.org/r/9b5314e3ee5eda61b0317ec1563768602c1ef438.1468523549.git.luto@kernel.org
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: track NR_KERNEL_STACK in KiB instead of number of stacks

Currently, NR_KERNEL_STACK tracks the number of kernel stacks in a zone.
This only makes sense if each kernel stack exists entirely in one zone,
and allowing vmapped stacks could break this assumption.

Since frv has THREAD_SIZE < PAGE_SIZE, we need to track kernel stack
allocations in a unit that divides both THREAD_SIZE and PAGE_SIZE on all
architectures.  Keep it simple and use KiB.

Link: http://lkml.kernel.org/r/083c71e642c5fa5f1b6898902e1b2db7b48940d4.1468523549.git.luto@kernel.org
Signed-off-by: Andy Lutomirski <luto@kernel.org>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: cleanup ifdef guards for vmem_altmap

Now that ZONE_DEVICE depends on SPARSEMEM_VMEMMAP we can simplify some
ifdef guards to just ZONE_DEVICE.

Link: http://lkml.kernel.org/r/146687646788.39261.8020536391978771940.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Eric Sandeen <sandeen@redhat.com>
Cc: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: CONFIG_ZONE_DEVICE stop depending on CONFIG_EXPERT

When it was first introduced CONFIG_ZONE_DEVICE depended on disabling
CONFIG_ZONE_DMA, a configuration choice reserved for "experts".
However, now that the ZONE_DMA conflict has been eliminated it no longer
makes sense to require CONFIG_EXPERT.

Link: http://lkml.kernel.org/r/146687646274.39261.14267596518720371009.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: Eric Sandeen <sandeen@redhat.com>
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Acked-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memblock: include <asm/sections.h> instead of <asm-generic/sections.h>

asm-generic headers are generic implementations for architecture
specific code and should not be included by common code.  Thus use the
asm/ version of sections.h to get at the linker sections.

Link: http://lkml.kernel.org/r/1468285103-7470-1-git-send-email-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, THP: clean up return value of madvise_free_huge_pmd

The definition of return value of madvise_free_huge_pmd is not clear
before.  According to the suggestion of Minchan Kim, change the type of
return value to bool and return true if we do MADV_FREE successfully on
entire pmd page, otherwise, return false.  Comments are added too.

Link: http://lkml.kernel.org/r/1467135452-16688-2-git-send-email-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Jerome Marchand <jmarchan@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Ebru Akagunduz <ebru.akagunduz@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc: use helper to clear page->flags bit

Use ClearPagePrivate/ClearPagePrivate2 helpers to clear
PG_private/PG_private_2 in page->flags

Link: http://lkml.kernel.org/r/1467882338-4300-7-git-send-email-opensource.ganesh@gmail.com
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc: add __init,__exit attribute

Add __init,__exit attribute for function that only called in module
init/exit to save memory.

Link: http://lkml.kernel.org/r/1467882338-4300-6-git-send-email-opensource.ganesh@gmail.com
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc: keep comments consistent with code

Some minor commebnt changes:

 1). update zs_malloc(),zs_create_pool() function header
 2). update "Usage of struct page fields"

Link: http://lkml.kernel.org/r/1467882338-4300-5-git-send-email-opensource.ganesh@gmail.com
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc: avoid calculate max objects of zspage twice

Currently, if a class can not be merged, the max objects of zspage in
that class may be calculated twice.

This patch calculate max objects of zspage at the begin, and pass the
value to can_merge() to decide whether the class can be merged.

Also this patch remove function get_maxobj_per_zspage(), as there is no
other place to call this function.

Link: http://lkml.kernel.org/r/1467882338-4300-4-git-send-email-opensource.ganesh@gmail.com
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc: use class->objs_per_zspage to get num of max objects

num of max objects in zspage is stored in each size_class now.  So there
is no need to re-calculate it.

Link: http://lkml.kernel.org/r/1467882338-4300-3-git-send-email-opensource.ganesh@gmail.com
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc: take obj index back from find_alloced_obj

the obj index value should be updated after return from
find_alloced_obj() to avoid CPU burning caused by unnecessary object
scanning.

Link: http://lkml.kernel.org/r/1467882338-4300-2-git-send-email-opensource.ganesh@gmail.com
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc: use obj_index to keep consistent with others

This is a cleanup patch.  Change "index" to "obj_index" to keep
consistent with others in zsmalloc.

Link: http://lkml.kernel.org/r/1467882338-4300-1-git-send-email-opensource.ganesh@gmail.com
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: bail out in shrink_inactive_list()

With node-lru, if there are enough reclaimable pages in highmem but
nothing in lowmem, VM can try to shrink inactive list although the
requested zone is lowmem.

The problem is that if the inactive list is full of highmem pages then a
direct reclaimer searching for a lowmem page waste CPU scanning
uselessly.  It just burns out CPU.  Even, many direct reclaimers are
stalled by too_many_isolated if lots of parallel reclaimer are going on
although there are no reclaimable memory in inactive list.

I tried the experiment 4 times in 32bit 2G 8 CPU KVM machine to get
elapsed time.

hackbench 500 process 2

 = Old =

  1st: 289s 2nd: 310s 3rd: 112s 4th: 272s

 = Now =

  1st: 31s  2nd: 132s 3rd: 162s 4th: 50s.

[akpm@linux-foundation.org: fixes per Mel]
Link: http://lkml.kernel.org/r/1469433119-1543-1-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: account for skipped pages as a partial scan

Page reclaim determines whether a pgdat is unreclaimable by examining
how many pages have been scanned since a page was freed and comparing
that to the LRU sizes.  Skipped pages are not reclaim candidates but
contribute to scanned.  This can prematurely mark a pgdat as
unreclaimable and trigger an OOM kill.

This patch accounts for skipped pages as a partial scan so that an
unreclaimable pgdat will still be marked as such but by scaling the cost
of a skip, it'll avoid the pgdat being marked prematurely.

Link: http://lkml.kernel.org/r/1469110261-7365-6-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: consider whether to decivate based on eligible zones inactive ratio

Minchan Kim reported that with per-zone lru state it was possible to
identify that a normal zone with 8^M anonymous pages could trigger OOM
with non-atomic order-0 allocations as all pages in the zone were in the
active list.

   gfp_mask=0x26004c0(GFP_KERNEL|__GFP_REPEAT|__GFP_NOTRACK), order=0
   Call Trace:
     __alloc_pages_nodemask+0xe52/0xe60
     ? new_slab+0x39c/0x3b0
     new_slab+0x39c/0x3b0
     ___slab_alloc.constprop.87+0x6da/0x840
     ? __alloc_skb+0x3c/0x260
     ? enqueue_task_fair+0x73/0xbf0
     ? poll_select_copy_remaining+0x140/0x140
     __slab_alloc.isra.81.constprop.86+0x40/0x6d
     ? __alloc_skb+0x3c/0x260
     kmem_cache_alloc+0x22c/0x260
     ? __alloc_skb+0x3c/0x260
     __alloc_skb+0x3c/0x260
     alloc_skb_with_frags+0x4e/0x1a0
     sock_alloc_send_pskb+0x16a/0x1b0
     ? wait_for_unix_gc+0x31/0x90
     unix_stream_sendmsg+0x28d/0x340
     sock_sendmsg+0x2d/0x40
     sock_write_iter+0x6c/0xc0
     __vfs_write+0xc0/0x120
     vfs_write+0x9b/0x1a0
     ? __might_fault+0x49/0xa0
     SyS_write+0x44/0x90
     do_fast_syscall_32+0xa6/0x1e0

   Mem-Info:
   active_anon:101103 inactive_anon:102219 isolated_anon:0
    active_file:503 inactive_file:544 isolated_file:0
    unevictable:0 dirty:0 writeback:34 unstable:0
    slab_reclaimable:6298 slab_unreclaimable:74669
    mapped:863 shmem:0 pagetables:100998 bounce:0
    free:23573 free_pcp:1861 free_cma:0
   Node 0 active_anon:404412kB inactive_anon:409040kB active_file:2012kB inactive_file:2176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:3452kB dirty:0kB writeback:136kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1320845 all_unreclaimable? yes
   DMA free:3296kB min:68kB low:84kB high:100kB active_anon:5540kB inactive_anon:0kB active_file:0kB inactive_file:0kB present:15992kB managed:15916kB mlocked:0kB slab_reclaimable:248kB slab_unreclaimable:2628kB kernel_stack:792kB pagetables:2316kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
   lowmem_reserve[]: 0 809 1965 1965
   Normal free:3600kB min:3604kB low:4504kB high:5404kB active_anon:86304kB inactive_anon:0kB active_file:160kB inactive_file:376kB present:897016kB managed:858524kB mlocked:0kB slab_reclaimable:24944kB slab_unreclaimable:296048kB kernel_stack:163832kB pagetables:35892kB bounce:0kB free_pcp:3076kB local_pcp:656kB free_cma:0kB
   lowmem_reserve[]: 0 0 9247 9247
   HighMem free:86156kB min:512kB low:1796kB high:3080kB active_anon:312852kB inactive_anon:410024kB active_file:1924kB inactive_file:2012kB present:1183736kB managed:1183736kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:365784kB bounce:0kB free_pcp:3868kB local_pcp:720kB free_cma:0kB
   lowmem_reserve[]: 0 0 0 0
   DMA: 8*4kB (UM) 8*8kB (UM) 4*16kB (M) 2*32kB (UM) 2*64kB (UM) 1*128kB (M) 3*256kB (UME) 2*512kB (UE) 1*1024kB (E) 0*2048kB 0*4096kB = 3296kB
   Normal: 240*4kB (UME) 160*8kB (UME) 23*16kB (ME) 3*32kB (UE) 3*64kB (UME) 2*128kB (ME) 1*256kB (U) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 3408kB
   HighMem: 10942*4kB (UM) 3102*8kB (UM) 866*16kB (UM) 76*32kB (UM) 11*64kB (UM) 4*128kB (UM) 1*256kB (M) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 86344kB
   Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=2048kB
   54409 total pagecache pages
   53215 pages in swap cache
   Swap cache stats: add 300982, delete 247765, find 157978/226539
   Free swap  = 3803244kB
   Total swap = 4192252kB
   524186 pages RAM
   295934 pages HighMem/MovableOnly
   9642 pages reserved
   0 pages cma reserved

The problem is due to the active deactivation logic in
inactive_list_is_low:

Node 0 active_anon:404412kB inactive_anon:409040kB

IOW, (inactive_anon of node * inactive_ratio > active_anon of node) due
to highmem anonymous stat so VM never deactivates normal zone's
anonymous pages.

This patch is a modified version of Minchan's original solution but
based upon it.  The problem with Minchan's patch is that any low zone
with an imbalanced list could force a rotation.

In this patch, a zone-constrained global reclaim will rotate the list if
the inactive/active ratio of all eligible zones needs to be corrected.
It is possible that higher zone pages will be initially rotated
prematurely but this is the safer choice to maintain overall LRU age.

Link: http://lkml.kernel.org/r/20160722090929.GJ10438@techsingularity.net
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove reclaim and compaction retry approximations

If per-zone LRU accounting is available then there is no point
approximating whether reclaim and compaction should retry based on pgdat
statistics.  This is effectively a revert of "mm, vmstat: remove zone
and node double accounting by approximating retries" with the difference
that inactive/active stats are still available.  This preserves the
history of why the approximation was retried and why it had to be
reverted to handle OOM kills on 32-bit systems.

Link: http://lkml.kernel.org/r/1469110261-7365-4-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: remove highmem_file_pages

With the reintroduction of per-zone LRU stats, highmem_file_pages is
redundant so remove it.

[mgorman@techsingularity.net: wrong stat is being accumulated in highmem_dirtyable_memory]
Link: http://lkml.kernel.org/r/20160725092324.GM10438@techsingularity.netLink:
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add per-zone lru list stat

When I did stress test with hackbench, I got OOM message frequently
which didn't ever happen in zone-lru.

  gfp_mask=0x26004c0(GFP_KERNEL|__GFP_REPEAT|__GFP_NOTRACK), order=0
  ..
  ..
   __alloc_pages_nodemask+0xe52/0xe60
   ? new_slab+0x39c/0x3b0
   new_slab+0x39c/0x3b0
   ___slab_alloc.constprop.87+0x6da/0x840
   ? __alloc_skb+0x3c/0x260
   ? _raw_spin_unlock_irq+0x27/0x60
   ? trace_hardirqs_on_caller+0xec/0x1b0
   ? finish_task_switch+0xa6/0x220
   ? poll_select_copy_remaining+0x140/0x140
   __slab_alloc.isra.81.constprop.86+0x40/0x6d
   ? __alloc_skb+0x3c/0x260
   kmem_cache_alloc+0x22c/0x260
   ? __alloc_skb+0x3c/0x260
   __alloc_skb+0x3c/0x260
   alloc_skb_with_frags+0x4e/0x1a0
   sock_alloc_send_pskb+0x16a/0x1b0
   ? wait_for_unix_gc+0x31/0x90
   ? alloc_set_pte+0x2ad/0x310
   unix_stream_sendmsg+0x28d/0x340
   sock_sendmsg+0x2d/0x40
   sock_write_iter+0x6c/0xc0
   __vfs_write+0xc0/0x120
   vfs_write+0x9b/0x1a0
   ? __might_fault+0x49/0xa0
   SyS_write+0x44/0x90
   do_fast_syscall_32+0xa6/0x1e0
   sysenter_past_esp+0x45/0x74

  Mem-Info:
  active_anon:104698 inactive_anon:105791 isolated_anon:192
   active_file:433 inactive_file:283 isolated_file:22
   unevictable:0 dirty:0 writeback:296 unstable:0
   slab_reclaimable:6389 slab_unreclaimable:78927
   mapped:474 shmem:0 pagetables:101426 bounce:0
   free:10518 free_pcp:334 free_cma:0
  Node 0 active_anon:418792kB inactive_anon:423164kB active_file:1732kB inactive_file:1132kB unevictable:0kB isolated(anon):768kB isolated(file):88kB mapped:1896kB dirty:0kB writeback:1184kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1478632 all_unreclaimable? yes
  DMA free:3304kB min:68kB low:84kB high:100kB present:15992kB managed:15916kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:4088kB kernel_stack:0kB pagetables:2480kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
  lowmem_reserve[]: 0 809 1965 1965
  Normal free:3436kB min:3604kB low:4504kB high:5404kB present:897016kB managed:858460kB mlocked:0kB slab_reclaimable:25556kB slab_unreclaimable:311712kB kernel_stack:164608kB pagetables:30844kB bounce:0kB free_pcp:620kB local_pcp:104kB free_cma:0kB
  lowmem_reserve[]: 0 0 9247 9247
  HighMem free:33808kB min:512kB low:1796kB high:3080kB present:1183736kB managed:1183736kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:372252kB bounce:0kB free_pcp:428kB local_pcp:72kB free_cma:0kB
  lowmem_reserve[]: 0 0 0 0
  DMA: 2*4kB (UM) 2*8kB (UM) 0*16kB 1*32kB (U) 1*64kB (U) 2*128kB (UM) 1*256kB (U) 1*512kB (M) 0*1024kB 1*2048kB (U) 0*4096kB = 3192kB
  Normal: 33*4kB (MH) 79*8kB (ME) 11*16kB (M) 4*32kB (M) 2*64kB (ME) 2*128kB (EH) 7*256kB (EH) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 3244kB
  HighMem: 2590*4kB (UM) 1568*8kB (UM) 491*16kB (UM) 60*32kB (UM) 6*64kB (M) 0*128kB 0*256kB 0*512kB 0*1024kB 0*2048kB 0*4096kB = 33064kB
  Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=2048kB
  25121 total pagecache pages
  24160 pages in swap cache
  Swap cache stats: add 86371, delete 62211, find 42865/60187
  Free swap  = 4015560kB
  Total swap = 4192252kB
  524186 pages RAM
  295934 pages HighMem/MovableOnly
  9658 pages reserved
  0 pages cma reserved

The order-0 allocation for normal zone failed while there are a lot of
reclaimable memory(i.e., anonymous memory with free swap).  I wanted to
analyze the problem but it was hard because we removed per-zone lru stat
so I couldn't know how many of anonymous memory there are in normal/dma
zone.

When we investigate OOM problem, reclaimable memory count is crucial
stat to find a problem.  Without it, it's hard to parse the OOM message
so I believe we should keep it.

With per-zone lru stat,

  gfp_mask=0x26004c0(GFP_KERNEL|__GFP_REPEAT|__GFP_NOTRACK), order=0
  Mem-Info:
  active_anon:101103 inactive_anon:102219 isolated_anon:0
   active_file:503 inactive_file:544 isolated_file:0
   unevictable:0 dirty:0 writeback:34 unstable:0
   slab_reclaimable:6298 slab_unreclaimable:74669
   mapped:863 shmem:0 pagetables:100998 bounce:0
   free:23573 free_pcp:1861 free_cma:0
  Node 0 active_anon:404412kB inactive_anon:409040kB active_file:2012kB inactive_file:2176kB unevictable:0kB isolated(anon):0kB isolated(file):0kB mapped:3452kB dirty:0kB writeback:136kB shmem:0kB writeback_tmp:0kB unstable:0kB pages_scanned:1320845 all_unreclaimable? yes
  DMA free:3296kB min:68kB low:84kB high:100kB active_anon:5540kB inactive_anon:0kB active_file:0kB inactive_file:0kB present:15992kB managed:15916kB mlocked:0kB slab_reclaimable:248kB slab_unreclaimable:2628kB kernel_stack:792kB pagetables:2316kB bounce:0kB free_pcp:0kB local_pcp:0kB free_cma:0kB
  lowmem_reserve[]: 0 809 1965 1965
  Normal free:3600kB min:3604kB low:4504kB high:5404kB active_anon:86304kB inactive_anon:0kB active_file:160kB inactive_file:376kB present:897016kB managed:858524kB mlocked:0kB slab_reclaimable:24944kB slab_unreclaimable:296048kB kernel_stack:163832kB pagetables:35892kB bounce:0kB free_pcp:3076kB local_pcp:656kB free_cma:0kB
  lowmem_reserve[]: 0 0 9247 9247
  HighMem free:86156kB min:512kB low:1796kB high:3080kB active_anon:312852kB inactive_anon:410024kB active_file:1924kB inactive_file:2012kB present:1183736kB managed:1183736kB mlocked:0kB slab_reclaimable:0kB slab_unreclaimable:0kB kernel_stack:0kB pagetables:365784kB bounce:0kB free_pcp:3868kB local_pcp:720kB free_cma:0kB
  lowmem_reserve[]: 0 0 0 0
  DMA: 8*4kB (UM) 8*8kB (UM) 4*16kB (M) 2*32kB (UM) 2*64kB (UM) 1*128kB (M) 3*256kB (UME) 2*512kB (UE) 1*1024kB (E) 0*2048kB 0*4096kB = 3296kB
  Normal: 240*4kB (UME) 160*8kB (UME) 23*16kB (ME) 3*32kB (UE) 3*64kB (UME) 2*128kB (ME) 1*256kB (U) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 3408kB
  HighMem: 10942*4kB (UM) 3102*8kB (UM) 866*16kB (UM) 76*32kB (UM) 11*64kB (UM) 4*128kB (UM) 1*256kB (M) 0*512kB 0*1024kB 0*2048kB 0*4096kB = 86344kB
  Node 0 hugepages_total=0 hugepages_free=0 hugepages_surp=0 hugepages_size=2048kB
  54409 total pagecache pages
  53215 pages in swap cache
  Swap cache stats: add 300982, delete 247765, find 157978/226539
  Free swap  = 3803244kB
  Total swap = 4192252kB
  524186 pages RAM
  295934 pages HighMem/MovableOnly
  9642 pages reserved
  0 pages cma reserved

With that, we can see normal zone has a 86M reclaimable memory so we can
know something goes wrong(I will fix the problem in next patch) in
reclaim.

[mgorman@techsingularity.net: rename zone LRU stats in /proc/vmstat]
Link: http://lkml.kernel.org/r/20160725072300.GK10438@techsingularity.net
Link: http://lkml.kernel.org/r/1469110261-7365-2-git-send-email-mgorman@techsingularity.net
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: release/reacquire lru_lock on pgdat change

With node-lru, the locking is based on the pgdat.  As Minchan pointed
out, there is an opportunity to reduce LRU lock release/acquire in
check_move_unevictable_pages by only changing lock on a pgdat change.

[mgorman@techsingularity.net: remove double initialisation]
Link: http://lkml.kernel.org/r/20160719074835.GC10438@techsingularity.net
Link: http://lkml.kernel.org/r/1468853426-12858-3-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: remove redundant check in shrink_zones()

As pointed out by Minchan Kim, shrink_zones() checks for populated zones
in a zonelist but a zonelist can never contain unpopulated zones.  While
it's not related to the node-lru series, it can be cleaned up now.

Link: http://lkml.kernel.org/r/1468853426-12858-2-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Suggested-by: Minchan Kim <minchan@kernel.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: Update all zone LRU sizes before updating memcg

Minchan Kim reported setting the following warning on a 32-bit system
although it can affect 64-bit systems.

  WARNING: CPU: 4 PID: 1322 at mm/memcontrol.c:998 mem_cgroup_update_lru_size+0x103/0x110
  mem_cgroup_update_lru_size(f44b4000, 1, -7): zid 1 lru_size 1 but empty
  Modules linked in:
  CPU: 4 PID: 1322 Comm: cp Not tainted 4.7.0-rc4-mm1+ #143
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
  Call Trace:
    dump_stack+0x76/0xaf
    __warn+0xea/0x110
    ? mem_cgroup_update_lru_size+0x103/0x110
    warn_slowpath_fmt+0x3b/0x40
    mem_cgroup_update_lru_size+0x103/0x110
    isolate_lru_pages.isra.61+0x2e2/0x360
    shrink_active_list+0xac/0x2a0
    ? __delay+0xe/0x10
    shrink_node_memcg+0x53c/0x7a0
    shrink_node+0xab/0x2a0
    do_try_to_free_pages+0xc6/0x390
    try_to_free_pages+0x245/0x590

LRU list contents and counts are updated separately.  Counts are updated
before pages are added to the LRU and updated after pages are removed.
The warning above is from a check in mem_cgroup_update_lru_size that
ensures that list sizes of zero are empty.

The problem is that node-lru needs to account for highmem pages if
CONFIG_HIGHMEM is set.  One impact of the implementation is that the
sizes are updated in multiple passes when pages from multiple zones were
isolated.  This happens whether HIGHMEM is set or not.  When multiple
zones are isolated, it's possible for a debugging check in memcg to be
tripped.

This patch forces all the zone counts to be updated before the memcg
function is called.

Link: http://lkml.kernel.org/r/1468588165-12461-6-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Minchan Kim <minchan@kernel.org>
Reported-by: Minchan Kim <minchan@kernel.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: show node_pages_scanned per node, not zone

The node_pages_scanned represents the number of scanned pages of node
for reclaim so it's pointless to show it as kilobytes.

As well, node_pages_scanned is per-node value, not per-zone.

This patch changes node_pages_scanned per-zone-killobytes with
per-node-count.

[minchan@kernel.org: fix node_pages_scanned]
Link: http://lkml.kernel.org/r/20160716101431.GA10305@bbox
Link: http://lkml.kernel.org/r/1468588165-12461-5-git-send-email-mgorman@techsingularity.net
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, pagevec: release/reacquire lru_lock on pgdat change

With node-lru, the locking is based on the pgdat.  Previously it was
required that a pagevec drain released one zone lru_lock and acquired
another zone lru_lock on every zone change.  Now, it's only necessary if
the node changes.  The end-result is fewer lock release/acquires if the
pages are all on the same node but in different zones.

Link: http://lkml.kernel.org/r/1468588165-12461-4-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Minchan Kim <minchan@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: fix dirtyable highmem calculation

When I tested vmscale in mmtest in 32bit, I found the benchmark was slow
down 0.5 times.

                base        node
                   1    global-1
User           12.98       16.04
System        147.61      166.42
Elapsed        26.48       38.08

With vmstat, I found IO wait avg is much increased compared to base.

The reason was highmem_dirtyable_memory accumulates free pages and
highmem_file_pages from HIGHMEM to MOVABLE zones which was wrong.  With
that, dirth_thresh in throtlle_vm_write is always 0 so that it calls
congestion_wait frequently if writeback starts.

With this patch, it is much recovered.

                base        node          fi
                   1    global-1         fix
User           12.98       16.04       13.78
System        147.61      166.42      143.92
Elapsed        26.48       38.08       29.64

Link: http://lkml.kernel.org/r/1468404004-5085-4-git-send-email-mgorman@techsingularity.net
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmstat: remove zone and node double accounting by approximating retries

The number of LRU pages, dirty pages and writeback pages must be
accounted for on both zones and nodes because of the reclaim retry
logic, compaction retry logic and highmem calculations all depending on
per-zone stats.

Many lowmem allocations are immune from OOM kill due to a check in
__alloc_pages_may_oom for (ac->high_zoneidx < ZONE_NORMAL) since commit
03668b3ceb0c ("oom: avoid oom killer for lowmem allocations").  The
exception is costly high-order allocations or allocations that cannot
fail.  If the __alloc_pages_may_oom avoids OOM-kill for low-order lowmem
allocations then it would fall through to __alloc_pages_direct_compact.

This patch will blindly retry reclaim for zone-constrained allocations
in should_reclaim_retry up to MAX_RECLAIM_RETRIES.  This is not ideal
but without per-zone stats there are not many alternatives.  The impact
it that zone-constrained allocations may delay before considering the
OOM killer.

As there is no guarantee enough memory can ever be freed to satisfy
compaction, this patch avoids retrying compaction for zone-contrained
allocations.

In combination, that means that the per-node stats can be used when
deciding whether to continue reclaim using a rough approximation.  While
it is possible this will make the wrong decision on occasion, it will
not infinite loop as the number of reclaim attempts is capped by
MAX_RECLAIM_RETRIES.

The final step is calculating the number of dirtyable highmem pages.  As
those calculations only care about the global count of file pages in
highmem.  This patch uses a global counter used instead of per-zone
stats as it is sufficient.

In combination, this allows the per-zone LRU and dirty state counters to
be removed.

[mgorman@techsingularity.net: fix acct_highmem_file_pages()]
Link: http://lkml.kernel.org/r/1468853426-12858-4-git-send-email-mgorman@techsingularity.netLink:
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Suggested by: Michal Hocko <mhocko@kernel.org>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmstat: print node-based stats in zoneinfo file

There are a number of stats that were previously accessible via zoneinfo
that are now invisible.  While it is possible to create a new file for
the node stats, this may be missed by users.  Instead this patch prints
the stats under the first populated zone in /proc/zoneinfo.

Link: http://lkml.kernel.org/r/1467970510-21195-34-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: vmstat: account per-zone stalls and pages skipped during reclaim

The vmstat allocstall was fairly useful in the general sense but
node-based LRUs change that.  It's important to know if a stall was for
an address-limited allocation request as this will require skipping
pages from other zones.  This patch adds pgstall_* counters to replace
allocstall.  The sum of the counters will equal the old allocstall so it
can be trivially recalculated.  A high number of address-limited
allocation requests may result in a lot of useless LRU scanning for
suitable pages.

As address-limited allocations require pages to be skipped, it's
important to know how much useless LRU scanning took place so this patch
adds pgskip* counters.  This yields the following model

1. The number of address-space limited stalls can be accounted for (pgstall)
2. The amount of useless work required to reclaim the data is accounted (pgskip)
3. The total number of scans is available from pgscan_kswapd and pgscan_direct
   so from that the ratio of useful to useless scans can be calculated.

[mgorman@techsingularity.net: s/pgstall/allocstall/]
Link: http://lkml.kernel.org/r/1468404004-5085-3-git-send-email-mgorman@techsingularity.netLink:
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: vmstat: replace __count_zone_vm_events with a zone id equivalent

This is partially a preparation patch for more vmstat work but it also
has the slight advantage that __count_zid_vm_events is cheaper to
calculate than __count_zone_vm_events().

Link: http://lkml.kernel.org/r/1467970510-21195-32-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: page_alloc: cache the last node whose dirty limit is reached

If a page is about to be dirtied then the page allocator attempts to
limit the total number of dirty pages that exists in any given zone.
The call to node_dirty_ok is expensive so this patch records if the last
pgdat examined hit the dirty limits.  In some cases, this reduces the
number of calls to node_dirty_ok().

Link: http://lkml.kernel.org/r/1467970510-21195-31-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: remove fair zone allocation policy

The fair zone allocation policy interleaves allocation requests between
zones to avoid an age inversion problem whereby new pages are reclaimed
to balance a zone.  Reclaim is now node-based so this should no longer
be an issue and the fair zone allocation policy is not free.  This patch
removes it.

Link: http://lkml.kernel.org/r/1467970510-21195-30-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: add classzone information to tracepoints

This is convenient when tracking down why the skip count is high because
it'll show what classzone kswapd woke up at and what zones are being
isolated.

Link: http://lkml.kernel.org/r/1467970510-21195-29-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: Have kswapd reclaim from all zones if reclaiming and buffer_heads_over_limit

The buffer_heads_over_limit limit in kswapd is inconsistent with direct
reclaim behaviour.  It may force an an attempt to reclaim from all zones
and then not reclaim at all because higher zones were balanced than
required by the original request.

This patch will causes kswapd to consider reclaiming from all zones if
buffer_heads_over_limit.  However, if there are eligible zones for the
allocation request that woke kswapd then no reclaim will occur even if
buffer_heads_over_limit.  This avoids kswapd over-reclaiming just
because buffer_heads_over_limit.

[mgorman@techsingularity.net: fix comment about buffer_heads_over_limit]
Link: http://lkml.kernel.org/r/1468404004-5085-2-git-send-email-mgorman@techsingularity.net
Link: http://lkml.kernel.org/r/1467970510-21195-28-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: avoid passing in `remaining' unnecessarily to prepare_kswapd_sleep()

As pointed out by Minchan Kim, the first call to prepare_kswapd_sleep()
always passes in 0 for `remaining' and the second call can trivially
check the parameter in advance.

Suggested-by: Minchan Kim <minchan@kernel.org>
Link: http://lkml.kernel.org/r/1467970510-21195-27-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: avoid passing in classzone_idx unnecessarily to compaction_ready

The scan_control structure has enough information available for
compaction_ready() to make a decision.  The classzone_idx manipulations
in shrink_zones() are no longer necessary as the highest populated zone
is no longer used to determine if shrink_slab should be called or not.

[mgorman@techsingularity.net remove redundant check in shrink_zones()]
Link: http://lkml.kernel.org/r/1468588165-12461-3-git-send-email-mgorman@techsingularity.net
Link: http://lkml.kernel.org/r/1467970510-21195-26-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: avoid passing in classzone_idx unnecessarily to shrink_node

shrink_node receives all information it needs about classzone_idx from
sc->reclaim_idx so remove the aliases.

Link: http://lkml.kernel.org/r/1467970510-21195-25-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: convert zone_reclaim to node_reclaim

As reclaim is now per-node based, convert zone_reclaim to be
node_reclaim.  It is possible that a node will be reclaimed multiple
times if it has multiple zones but this is unavoidable without caching
all nodes traversed so far.  The documentation and interface to
userspace is the same from a configuration perspective and will will be
similar in behaviour unless the node-local allocation requests were also
limited to lower zones.

Link: http://lkml.kernel.org/r/1467970510-21195-24-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: wake kswapd based on the highest eligible zone

The ac_classzone_idx is used as the basis for waking kswapd and that is
based on the preferred zoneref.  If the preferred zoneref's first zone
is lower than what is available on other nodes, it's possible that
kswapd is woken on a zone with only higher, but still eligible, zones.
As classzone_idx is strictly adhered to now, it causes a problem because
eligible pages are skipped.

For example, node 0 has only DMA32 and node 1 has only NORMAL.  An
allocating context running on node 0 may wake kswapd on node 1 telling
it to skip all NORMAL pages.

Link: http://lkml.kernel.org/r/1467970510-21195-23-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: only wakeup kswapd once per node for the requested classzone

kswapd is woken when zones are below the low watermark but the wakeup
decision is not taking the classzone into account.  Now that reclaim is
node-based, it is only required to wake kswapd once per node and only if
all zones are unbalanced for the requested classzone.

Note that one node might be checked multiple times if the zonelist is
ordered by node because there is no cheap way of tracking what nodes
have already been visited.  For zone-ordering, each node should be
checked only once.

Link: http://lkml.kernel.org/r/1467970510-21195-22-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move vmscan writes and file write accounting to the node

As reclaim is now node-based, it follows that page write activity due to
page reclaim should also be accounted for on the node.  For consistency,
also account page writes and page dirtying on a per-node basis.

After this patch, there are a few remaining zone counters that may appear
strange but are fine.  NUMA stats are still per-zone as this is a
user-space interface that tools consume.  NR_MLOCK, NR_SLAB_*,
NR_PAGETABLE, NR_KERNEL_STACK and NR_BOUNCE are all allocations that
potentially pin low memory and cannot trivially be reclaimed on demand.
This information is still useful for debugging a page allocation failure
warning.

Link: http://lkml.kernel.org/r/1467970510-21195-21-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move most file-based accounting to the node

There are now a number of accounting oddities such as mapped file pages
being accounted for on the node while the total number of file pages are
accounted on the zone.  This can be coped with to some extent but it's
confusing so this patch moves the relevant file-based accounted.  Due to
throttling logic in the page allocator for reliable OOM detection, it is
still necessary to track dirty and writeback pages on a per-zone basis.

[mgorman@techsingularity.net: fix NR_ZONE_WRITE_PENDING accounting]
Link: http://lkml.kernel.org/r/1468404004-5085-5-git-send-email-mgorman@techsingularity.net
Link: http://lkml.kernel.org/r/1467970510-21195-20-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: rename NR_ANON_PAGES to NR_ANON_MAPPED

NR_FILE_PAGES  is the number of        file pages.
NR_FILE_MAPPED is the number of mapped file pages.
NR_ANON_PAGES  is the number of mapped anon pages.

This is unhelpful naming as it's easy to confuse NR_FILE_MAPPED and
NR_ANON_PAGES for mapped pages.  This patch renames NR_ANON_PAGES so we
have

NR_FILE_PAGES  is the number of        file pages.
NR_FILE_MAPPED is the number of mapped file pages.
NR_ANON_MAPPED is the number of mapped anon pages.

Link: http://lkml.kernel.org/r/1467970510-21195-19-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move page mapped accounting to the node

Reclaim makes decisions based on the number of pages that are mapped but
it's mixing node and zone information.  Account NR_FILE_MAPPED and
NR_ANON_PAGES pages on the node.

Link: http://lkml.kernel.org/r/1467970510-21195-18-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, page_alloc: consider dirtyable memory in terms of nodes

Historically dirty pages were spread among zones but now that LRUs are
per-node it is more appropriate to consider dirty pages in a node.

Link: http://lkml.kernel.org/r/1467970510-21195-17-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, workingset: make working set detection node-aware

Working set and refault detection is still zone-based, fix it.

Link: http://lkml.kernel.org/r/1467970510-21195-16-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, memcg: move memcg limit enforcement from zones to nodes

Memcg needs adjustment after moving LRUs to the node.  Limits are
tracked per memcg but the soft-limit excess is tracked per zone.  As
global page reclaim is based on the node, it is easy to imagine a
situation where a zone soft limit is exceeded even though the memcg
limit is fine.

This patch moves the soft limit tree the node.  Technically, all the
variable names should also change but people are already familiar by the
meaning of "mz" even if "mn" would be a more appropriate name now.

Link: http://lkml.kernel.org/r/1467970510-21195-15-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: make shrink_node decisions more node-centric

Earlier patches focused on having direct reclaim and kswapd use data
that is node-centric for reclaiming but shrink_node() itself still uses
too much zone information.  This patch removes unnecessary zone-based
information with the most important decision being whether to continue
reclaim or not.  Some memcg APIs are adjusted as a result even though
memcg itself still uses some zone information.

[mgorman@techsingularity.net: optimization]
Link: http://lkml.kernel.org/r/1468588165-12461-2-git-send-email-mgorman@techsingularity.net
Link: http://lkml.kernel.org/r/1467970510-21195-14-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: vmscan: do not reclaim from kswapd if there is any eligible zone

kswapd scans from highest to lowest for a zone that requires balancing.
This was necessary when reclaim was per-zone to fairly age pages on
lower zones.  Now that we are reclaiming on a per-node basis, any
eligible zone can be used and pages will still be aged fairly.  This
patch avoids reclaiming excessively unless buffer_heads are over the
limit and it's necessary to reclaim from a higher zone than requested by
the waker of kswapd to relieve low memory pressure.

[hillf.zj@alibaba-inc.com: Force kswapd reclaim no more than needed]
Link: http://lkml.kernel.org/r/1466518566-30034-12-git-send-email-mgorman@techsingularity.net
Link: http://lkml.kernel.org/r/1467970510-21195-13-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: remove duplicate logic clearing node congestion and dirty state

Reclaim may stall if there is too much dirty or congested data on a
node.  This was previously based on zone flags and the logic for
clearing the flags is in two places.  As congestion/dirty tracking is
now tracked on a per-node basis, we can remove some duplicate logic.

Link: http://lkml.kernel.org/r/1467970510-21195-12-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: by default have direct reclaim only shrink once per node

Direct reclaim iterates over all zones in the zonelist and shrinking
them but this is in conflict with node-based reclaim.  In the default
case, only shrink once per node.

Link: http://lkml.kernel.org/r/1467970510-21195-11-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: simplify the logic deciding whether kswapd sleeps

kswapd goes through some complex steps trying to figure out if it should
stay awake based on the classzone_idx and the requested order.  It is
unnecessarily complex and passes in an invalid classzone_idx to
balance_pgdat().  What matters most of all is whether a larger order has
been requsted and whether kswapd successfully reclaimed at the previous
order.  This patch irons out the logic to check just that and the end
result is less headache inducing.

Link: http://lkml.kernel.org/r/1467970510-21195-10-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: remove balance gap

The balance gap was introduced to apply equal pressure to all zones when
reclaiming for a higher zone.  With node-based LRU, the need for the
balance gap is removed and the code is dead so remove it.

[vbabka@suse.cz: Also remove KSWAPD_ZONE_BALANCE_GAP_RATIO]
Link: http://lkml.kernel.org/r/1467970510-21195-9-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: make kswapd reclaim in terms of nodes

Patch "mm: vmscan: Begin reclaiming pages on a per-node basis" started
thinking of reclaim in terms of nodes but kswapd is still zone-centric.
This patch gets rid of many of the node-based versus zone-based
decisions.

o A node is considered balanced when any eligible lower zone is balanced.
  This eliminates one class of age-inversion problem because we avoid
  reclaiming a newer page just because it's in the wrong zone
o pgdat_balanced disappears because we now only care about one zone being
  balanced.
o Some anomalies related to writeback and congestion tracking being based on
  zones disappear.
o kswapd no longer has to take care to reclaim zones in the reverse order
  that the page allocator uses.
o Most importantly of all, reclaim from node 0 with multiple zones will
  have similar aging and reclaiming characteristics as every
  other node.

Link: http://lkml.kernel.org/r/1467970510-21195-8-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: have kswapd only scan based on the highest requested zone

kswapd checks all eligible zones to see if they need balancing even if
it was woken for a lower zone.  This made sense when we reclaimed on a
per-zone basis because we wanted to shrink zones fairly so avoid
age-inversion problems.  Ideally this is completely unnecessary when
reclaiming on a per-node basis.  In theory, there may still be anomalies
when all requests are for lower zones and very old pages are preserved
in higher zones but this should be the exceptional case.

Link: http://lkml.kernel.org/r/1467970510-21195-7-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: begin reclaiming pages on a per-node basis

This patch makes reclaim decisions on a per-node basis.  A reclaimer
knows what zone is required by the allocation request and skips pages
from higher zones.  In many cases this will be ok because it's a
GFP_HIGHMEM request of some description.  On 64-bit, ZONE_DMA32 requests
will cause some problems but 32-bit devices on 64-bit platforms are
increasingly rare.  Historically it would have been a major problem on
32-bit with big Highmem:Lowmem ratios but such configurations are also
now rare and even where they exist, they are not encouraged.  If it
really becomes a problem, it'll manifest as very low reclaim
efficiencies.

Link: http://lkml.kernel.org/r/1467970510-21195-6-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, mmzone: clarify the usage of zone padding

Zone padding separates write-intensive fields used by page allocation,
compaction and vmstats but the comments are a little misleading and need
clarification.

Link: http://lkml.kernel.org/r/1467970510-21195-5-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: move LRU lists to node

This moves the LRU lists from the zone to the node and related data such
as counters, tracing, congestion tracking and writeback tracking.

Unfortunately, due to reclaim and compaction retry logic, it is
necessary to account for the number of LRU pages on both zone and node
logic.  Most reclaim logic is based on the node counters but the retry
logic uses the zone counters which do not distinguish inactive and
active sizes.  It would be possible to leave the LRU counters on a
per-zone basis but it's a heavier calculation across multiple cache
lines that is much more frequent than the retry checks.

Other than the LRU counters, this is mostly a mechanical patch but note
that it introduces a number of anomalies.  For example, the scans are
per-zone but using per-node counters.  We also mark a node as congested
when a zone is congested.  This causes weird problems that are fixed
later but is easier to review.

In the event that there is excessive overhead on 32-bit systems due to
the nodes being on LRU then there are two potential solutions

1. Long-term isolation of highmem pages when reclaim is lowmem

   When pages are skipped, they are immediately added back onto the LRU
   list. If lowmem reclaim persisted for long periods of time, the same
   highmem pages get continually scanned. The idea would be that lowmem
   keeps those pages on a separate list until a reclaim for highmem pages
   arrives that splices the highmem pages back onto the LRU. It potentially
   could be implemented similar to the UNEVICTABLE list.

   That would reduce the skip rate with the potential corner case is that
   highmem pages have to be scanned and reclaimed to free lowmem slab pages.

2. Linear scan lowmem pages if the initial LRU shrink fails

   This will break LRU ordering but may be preferable and faster during
   memory pressure than skipping LRU pages.

Link: http://lkml.kernel.org/r/1467970510-21195-4-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmscan: move lru_lock to the node

Node-based reclaim requires node-based LRUs and locking.  This is a
preparation patch that just moves the lru_lock to the node so later
patches are easier to review.  It is a mechanical change but note this
patch makes contention worse because the LRU lock is hotter and direct
reclaim and kswapd can contend on the same lock even when reclaiming
from different zones.

Link: http://lkml.kernel.org/r/1467970510-21195-3-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Minchan Kim <minchan@kernel.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, vmstat: add infrastructure for per-node vmstats

Patchset: "Move LRU page reclaim from zones to nodes v9"

This series moves LRUs from the zones to the node.  While this is a
current rebase, the test results were based on mmotm as of June 23rd.
Conceptually, this series is simple but there are a lot of details.
Some of the broad motivations for this are;

1. The residency of a page partially depends on what zone the page was
   allocated from.  This is partially combatted by the fair zone allocation
   policy but that is a partial solution that introduces overhead in the
   page allocator paths.

2. Currently, reclaim on node 0 behaves slightly different to node 1. For
   example, direct reclaim scans in zonelist order and reclaims even if
   the zone is over the high watermark regardless of the age of pages
   in that LRU. Kswapd on the other hand starts reclaim on the highest
   unbalanced zone. A difference in distribution of file/anon pages due
   to when they were allocated results can result in a difference in
   again. While the fair zone allocation policy mitigates some of the
   problems here, the page reclaim results on a multi-zone node will
   always be different to a single-zone node.
   it was scheduled on as a result.

3. kswapd and the page allocator scan zones in the opposite order to
   avoid interfering with each other but it's sensitive to timing.  This
   mitigates the page allocator using pages that were allocated very recently
   in the ideal case but it's sensitive to timing. When kswapd is allocating
   from lower zones then it's great but during the rebalancing of the highest
   zone, the page allocator and kswapd interfere with each other. It's worse
   if the highest zone is small and difficult to balance.

4. slab shrinkers are node-based which makes it harder to identify the exact
   relationship between slab reclaim and LRU reclaim.

The reason we have zone-based reclaim is that we used to have
large highmem zones in common configurations and it was necessary
to quickly find ZONE_NORMAL pages for reclaim. Today, this is much
less of a concern as machines with lots of memory will (or should) use
64-bit kernels. Combinations of 32-bit hardware and 64-bit hardware are
rare. Machines that do use highmem should have relatively low highmem:lowmem
ratios than we worried about in the past.

Conceptually, moving to node LRUs should be easier to understand. The
page allocator plays fewer tricks to game reclaim and reclaim behaves
similarly on all nodes.

The series has been tested on a 16 core UMA machine and a 2-socket 48
core NUMA machine. The UMA results are presented in most cases as the NUMA
machine behaved similarly.

pagealloc
---------

This is a microbenchmark that shows the benefit of removing the fair zone
allocation policy. It was tested uip to order-4 but only orders 0 and 1 are
shown as the other orders were comparable.

                                           4.7.0-rc4                  4.7.0-rc4
                                      mmotm-20160623                 nodelru-v9
Min      total-odr0-1               490.00 (  0.00%)           457.00 (  6.73%)
Min      total-odr0-2               347.00 (  0.00%)           329.00 (  5.19%)
Min      total-odr0-4               288.00 (  0.00%)           273.00 (  5.21%)
Min      total-odr0-8               251.00 (  0.00%)           239.00 (  4.78%)
Min      total-odr0-16              234.00 (  0.00%)           222.00 (  5.13%)
Min      total-odr0-32              223.00 (  0.00%)           211.00 (  5.38%)
Min      total-odr0-64              217.00 (  0.00%)           208.00 (  4.15%)
Min      total-odr0-128             214.00 (  0.00%)           204.00 (  4.67%)
Min      total-odr0-256             250.00 (  0.00%)           230.00 (  8.00%)
Min      total-odr0-512             271.00 (  0.00%)           269.00 (  0.74%)
Min      total-odr0-1024            291.00 (  0.00%)           282.00 (  3.09%)
Min      total-odr0-2048            303.00 (  0.00%)           296.00 (  2.31%)
Min      total-odr0-4096            311.00 (  0.00%)           309.00 (  0.64%)
Min      total-odr0-8192            316.00 (  0.00%)           314.00 (  0.63%)
Min      total-odr0-16384           317.00 (  0.00%)           315.00 (  0.63%)
Min      total-odr1-1               742.00 (  0.00%)           712.00 (  4.04%)
Min      total-odr1-2               562.00 (  0.00%)           530.00 (  5.69%)
Min      total-odr1-4               457.00 (  0.00%)           433.00 (  5.25%)
Min      total-odr1-8               411.00 (  0.00%)           381.00 (  7.30%)
Min      total-odr1-16              381.00 (  0.00%)           356.00 (  6.56%)
Min      total-odr1-32              372.00 (  0.00%)           346.00 (  6.99%)
Min      total-odr1-64              372.00 (  0.00%)           343.00 (  7.80%)
Min      total-odr1-128             375.00 (  0.00%)           351.00 (  6.40%)
Min      total-odr1-256             379.00 (  0.00%)           351.00 (  7.39%)
Min      total-odr1-512             385.00 (  0.00%)           355.00 (  7.79%)
Min      total-odr1-1024            386.00 (  0.00%)           358.00 (  7.25%)
Min      total-odr1-2048            390.00 (  0.00%)           362.00 (  7.18%)
Min      total-odr1-4096            390.00 (  0.00%)           362.00 (  7.18%)
Min      total-odr1-8192            388.00 (  0.00%)           363.00 (  6.44%)

This shows a steady improvement throughout. The primary benefit is from
reduced system CPU usage which is obvious from the overall times;

           4.7.0-rc4   4.7.0-rc4
        mmotm-20160623nodelru-v8
User          189.19      191.80
System       2604.45     2533.56
Elapsed      2855.30     2786.39

The vmstats also showed that the fair zone allocation policy was definitely
removed as can be seen here;

                             4.7.0-rc3   4.7.0-rc3
                         mmotm-20160623 nodelru-v8
DMA32 allocs               28794729769           0
Normal allocs              48432501431 77227309877
Movable allocs                       0           0

tiobench on ext4
----------------

tiobench is a benchmark that artifically benefits if old pages remain resident
while new pages get reclaimed. The fair zone allocation policy mitigates this
problem so pages age fairly. While the benchmark has problems, it is important
that tiobench performance remains constant as it implies that page aging
problems that the fair zone allocation policy fixes are not re-introduced.

                                         4.7.0-rc4             4.7.0-rc4
                                    mmotm-20160623            nodelru-v9
Min      PotentialReadSpeed        89.65 (  0.00%)       90.21 (  0.62%)
Min      SeqRead-MB/sec-1          82.68 (  0.00%)       82.01 ( -0.81%)
Min      SeqRead-MB/sec-2          72.76 (  0.00%)       72.07 ( -0.95%)
Min      SeqRead-MB/sec-4          75.13 (  0.00%)       74.92 ( -0.28%)
Min      SeqRead-MB/sec-8          64.91 (  0.00%)       65.19 (  0.43%)
Min      SeqRead-MB/sec-16         62.24 (  0.00%)       62.22 ( -0.03%)
Min      RandRead-MB/sec-1          0.88 (  0.00%)        0.88 (  0.00%)
Min      RandRead-MB/sec-2          0.95 (  0.00%)        0.92 ( -3.16%)
Min      RandRead-MB/sec-4          1.43 (  0.00%)        1.34 ( -6.29%)
Min      RandRead-MB/sec-8          1.61 (  0.00%)        1.60 ( -0.62%)
Min      RandRead-MB/sec-16         1.80 (  0.00%)        1.90 (  5.56%)
Min      SeqWrite-MB/sec-1         76.41 (  0.00%)       76.85 (  0.58%)
Min      SeqWrite-MB/sec-2         74.11 (  0.00%)       73.54 ( -0.77%)
Min      SeqWrite-MB/sec-4         80.05 (  0.00%)       80.13 (  0.10%)
Min      SeqWrite-MB/sec-8         72.88 (  0.00%)       73.20 (  0.44%)
Min      SeqWrite-MB/sec-16        75.91 (  0.00%)       76.44 (  0.70%)
Min      RandWrite-MB/sec-1         1.18 (  0.00%)        1.14 ( -3.39%)
Min      RandWrite-MB/sec-2         1.02 (  0.00%)        1.03 (  0.98%)
Min      RandWrite-MB/sec-4         1.05 (  0.00%)        0.98 ( -6.67%)
Min      RandWrite-MB/sec-8         0.89 (  0.00%)        0.92 (  3.37%)
Min      RandWrite-MB/sec-16        0.92 (  0.00%)        0.93 (  1.09%)

           4.7.0-rc4   4.7.0-rc4
        mmotm-20160623 approx-v9
User          645.72      525.90
System        403.85      331.75
Elapsed      6795.36     6783.67

This shows that the series has little or not impact on tiobench which is
desirable and a reduction in system CPU usage. It indicates that the fair
zone allocation policy was removed in a manner that didn't reintroduce
one class of page aging bug. There were only minor differences in overall
reclaim activity

                             4.7.0-rc4   4.7.0-rc4
                          mmotm-20160623nodelru-v8
Minor Faults                    645838      647465
Major Faults                       573         640
Swap Ins                             0           0
Swap Outs                            0           0
DMA allocs                           0           0
DMA32 allocs                  46041453    44190646
Normal allocs                 78053072    79887245
Movable allocs                       0           0
Allocation stalls                   24          67
Stall zone DMA                       0           0
Stall zone DMA32                     0           0
Stall zone Normal                    0           2
Stall zone HighMem                   0           0
Stall zone Movable                   0          65
Direct pages scanned             10969       30609
Kswapd pages scanned          93375144    93492094
Kswapd pages reclaimed        93372243    93489370
Direct pages reclaimed           10969       30609
Kswapd efficiency                  99%         99%
Kswapd velocity              13741.015   13781.934
Direct efficiency                 100%        100%
Direct velocity                  1.614       4.512
Percentage direct scans             0%          0%

kswapd activity was roughly comparable. There were differences in direct
reclaim activity but negligible in the context of the overall workload
(velocity of 4 pages per second with the patches applied, 1.6 pages per
second in the baseline kernel).

pgbench read-only large configuration on ext4
---------------------------------------------

pgbench is a database benchmark that can be sensitive to page reclaim
decisions. This also checks if removing the fair zone allocation policy
is safe

pgbench Transactions
                        4.7.0-rc4             4.7.0-rc4
                   mmotm-20160623            nodelru-v8
Hmean    1       188.26 (  0.00%)      189.78 (  0.81%)
Hmean    5       330.66 (  0.00%)      328.69 ( -0.59%)
Hmean    12      370.32 (  0.00%)      380.72 (  2.81%)
Hmean    21      368.89 (  0.00%)      369.00 (  0.03%)
Hmean    30      382.14 (  0.00%)      360.89 ( -5.56%)
Hmean    32      428.87 (  0.00%)      432.96 (  0.95%)

Negligible differences again. As with tiobench, overall reclaim activity
was comparable.

bonnie++ on ext4
----------------

No interesting performance difference, negligible differences on reclaim
stats.

paralleldd on ext4
------------------

This workload uses varying numbers of dd instances to read large amounts of
data from disk.

                               4.7.0-rc3             4.7.0-rc3
                          mmotm-20160623            nodelru-v9
Amean    Elapsd-1       186.04 (  0.00%)      189.41 ( -1.82%)
Amean    Elapsd-3       192.27 (  0.00%)      191.38 (  0.46%)
Amean    Elapsd-5       185.21 (  0.00%)      182.75 (  1.33%)
Amean    Elapsd-7       183.71 (  0.00%)      182.11 (  0.87%)
Amean    Elapsd-12      180.96 (  0.00%)      181.58 ( -0.35%)
Amean    Elapsd-16      181.36 (  0.00%)      183.72 ( -1.30%)

           4.7.0-rc4   4.7.0-rc4
        mmotm-20160623 nodelru-v9
User         1548.01     1552.44
System       8609.71     8515.08
Elapsed      3587.10     3594.54

There is little or no change in performance but some drop in system CPU usage.

                             4.7.0-rc3   4.7.0-rc3
                        mmotm-20160623  nodelru-v9
Minor Faults                    362662      367360
Major Faults                      1204        1143
Swap Ins                            22           0
Swap Outs                         2855        1029
DMA allocs                           0           0
DMA32 allocs                  31409797    28837521
Normal allocs                 46611853    49231282
Movable allocs                       0           0
Direct pages scanned                 0           0
Kswapd pages scanned          40845270    40869088
Kswapd pages reclaimed        40830976    40855294
Direct pages reclaimed               0           0
Kswapd efficiency                  99%         99%
Kswapd velocity              11386.711   11369.769
Direct efficiency                 100%        100%
Direct velocity                  0.000       0.000
Percentage direct scans             0%          0%
Page writes by reclaim            2855        1029
Page writes file                     0           0
Page writes anon                  2855        1029
Page reclaim immediate             771        1628
Sector Reads                 293312636   293536360
Sector Writes                 18213568    18186480
Page rescued immediate               0           0
Slabs scanned                   128257      132747
Direct inode steals                181          56
Kswapd inode steals                 59        1131

It basically shows that kswapd was active at roughly the same rate in
both kernels. There was also comparable slab scanning activity and direct
reclaim was avoided in both cases. There appears to be a large difference
in numbers of inodes reclaimed but the workload has few active inodes and
is likely a timing artifact.

stutter
-------

stutter simulates a simple workload. One part uses a lot of anonymous
memory, a second measures mmap latency and a third copies a large file.
The primary metric is checking for mmap latency.

stutter
                             4.7.0-rc4             4.7.0-rc4
                        mmotm-20160623            nodelru-v8
Min         mmap     16.6283 (  0.00%)     13.4258 ( 19.26%)
1st-qrtle   mmap     54.7570 (  0.00%)     34.9121 ( 36.24%)
2nd-qrtle   mmap     57.3163 (  0.00%)     46.1147 ( 19.54%)
3rd-qrtle   mmap     58.9976 (  0.00%)     47.1882 ( 20.02%)
Max-90%     mmap     59.7433 (  0.00%)     47.4453 ( 20.58%)
Max-93%     mmap     60.1298 (  0.00%)     47.6037 ( 20.83%)
Max-95%     mmap     73.4112 (  0.00%)     82.8719 (-12.89%)
Max-99%     mmap     92.8542 (  0.00%)     88.8870 (  4.27%)
Max         mmap   1440.6569 (  0.00%)    121.4201 ( 91.57%)
Mean        mmap     59.3493 (  0.00%)     42.2991 ( 28.73%)
Best99%Mean mmap     57.2121 (  0.00%)     41.8207 ( 26.90%)
Best95%Mean mmap     55.9113 (  0.00%)     39.9620 ( 28.53%)
Best90%Mean mmap     55.6199 (  0.00%)     39.3124 ( 29.32%)
Best50%Mean mmap     53.2183 (  0.00%)     33.1307 ( 37.75%)
Best10%Mean mmap     45.9842 (  0.00%)     20.4040 ( 55.63%)
Best5%Mean  mmap     43.2256 (  0.00%)     17.9654 ( 58.44%)
Best1%Mean  mmap     32.9388 (  0.00%)     16.6875 ( 49.34%)

This shows a number of improvements with the worst-case outlier greatly
improved.

Some of the vmstats are interesting

                             4.7.0-rc4   4.7.0-rc4
                          mmotm-20160623nodelru-v8
Swap Ins                           163         502
Swap Outs                            0           0
DMA allocs                           0           0
DMA32 allocs                 618719206  1381662383
Normal allocs                891235743   564138421
Movable allocs                       0           0
Allocation stalls                 2603           1
Direct pages scanned            216787           2
Kswapd pages scanned          50719775    41778378
Kswapd pages reclaimed        41541765    41777639
Direct pages reclaimed          209159           0
Kswapd efficiency                  81%         99%
Kswapd velocity              16859.554   14329.059
Direct efficiency                  96%          0%
Direct velocity                 72.061       0.001
Percentage direct scans             0%          0%
Page writes by reclaim         6215049           0
Page writes file               6215049           0
Page writes anon                     0           0
Page reclaim immediate           70673          90
Sector Reads                  81940800    81680456
Sector Writes                100158984    98816036
Page rescued immediate               0           0
Slabs scanned                  1366954       22683

While this is not guaranteed in all cases, this particular test showed
a large reduction in direct reclaim activity. It's also worth noting
that no page writes were issued from reclaim context.

This series is not without its hazards. There are at least three areas
that I'm concerned with even though I could not reproduce any problems in
that area.

1. Reclaim/compaction is going to be affected because the amount of reclaim is
   no longer targetted at a specific zone. Compaction works on a per-zone basis
   so there is no guarantee that reclaiming a few THP's worth page pages will
   have a positive impact on compaction success rates.

2. The Slab/LRU reclaim ratio is affected because the frequency the shrinkers
   are called is now different. This may or may not be a problem but if it
   is, it'll be because shrinkers are not called enough and some balancing
   is required.

3. The anon/file reclaim ratio may be affected. Pages about to be dirtied are
   distributed between zones and the fair zone allocation policy used to do
   something very similar for anon. The distribution is now different but not
   necessarily in any way that matters but it's still worth bearing in mind.

VM statistic counters for reclaim decisions are zone-based.  If the kernel
is to reclaim on a per-node basis then we need to track per-node
statistics but there is no infrastructure for that.  The most notable
change is that the old node_page_state is renamed to
sum_zone_node_page_state.  The new node_page_state takes a pglist_data and
uses per-node stats but none exist yet.  There is some renaming such as
vm_stat to vm_zone_stat and the addition of vm_node_stat and the renaming
of mod_state to mod_zone_state.  Otherwise, this is mostly a mechanical
patch with no functional change.  There is a lot of similarity between the
node and zone helpers which is unfortunate but there was no obvious way of
reusing the code and maintaining type safety.

Link: http://lkml.kernel.org/r/1467970510-21195-2-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Rik van Riel <riel@surriel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Hillf Danton <hillf.zj@alibaba-inc.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, meminit: remove early_page_nid_uninitialised

The helper early_page_nid_uninitialised() has been dead since commit
974a786e63c9 ("mm, page_alloc: remove MIGRATE_RESERVE") so remove the
dead code.

Link: http://lkml.kernel.org/r/1468008031-3848-2-git-send-email-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cpuset, mm: fix TIF_MEMDIE check in cpuset_change_task_nodemask

Commit c0ff7453bb5c ("cpuset,mm: fix no node to alloc memory when
changing cpuset's mems") has added TIF_MEMDIE and PF_EXITING check but
it is checking the flag on the current task rather than the given one.

This doesn't make much sense and it is actually wrong.  If the current
task which updates the nodemask of a cpuset got killed by the OOM killer
then a part of the cpuset cgroup processes would have incompatible
nodemask which is surprising to say the least.

The comment suggests the intention was to skip oom victim or an exiting
task so we should be checking the given task.  But even then it would be
layering violation because it is the memory allocator to interpret the
TIF_MEMDIE meaning.  Simply drop both checks.  All tasks in the cpuset
should simply follow the same mask.

Link: http://lkml.kernel.org/r/1467029719-17602-3-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: Miao Xie <miaoxie@huawei.com>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

freezer, oom: check TIF_MEMDIE on the correct task

freezing_slow_path() is checking TIF_MEMDIE to skip OOM killed tasks.
It is, however, checking the flag on the current task rather than the
given one.  This is really confusing because freezing() can be called
also on !current tasks.  It would end up working correctly for its main
purpose because __refrigerator will be always called on the current task
so the oom victim will never get frozen.  But it could lead to
surprising results when a task which is freezing a cgroup got oom killed
because only part of the cgroup would get frozen.  This is highly
unlikely but worth fixing as the resulting code would be more clear
anyway.

Link: http://lkml.kernel.org/r/1467029719-17602-2-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: Miao Xie <miaoxie@huawei.com>
Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/compaction: remove unnecessary order check in try_to_compact_pages()

The caller __alloc_pages_direct_compact() already checked (order == 0)
so there's no need to check again.

Link: http://lkml.kernel.org/r/1465973568-3496-1-git-send-email-opensource.ganesh@gmail.com
Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Michal Nazarewicz <mina86@mina86.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: fix vm-scalability regression in cgroup-aware workingset code

Commit 23047a96d7cf ("mm: workingset: per-cgroup cache thrash
detection") added a page->mem_cgroup lookup to the cache eviction,
refault, and activation paths, as well as locking to the activation
path, and the vm-scalability tests showed a regression of -23%.

While the test in question is an artificial worst-case scenario that
doesn't occur in real workloads - reading two sparse files in parallel
at full CPU speed just to hammer the LRU paths - there is still some
optimizations that can be done in those paths.

Inline the lookup functions to eliminate calls.  Also, page->mem_cgroup
doesn't need to be stabilized when counting an activation; we merely
need to hold the RCU lock to prevent the memcg from being freed.

This cuts down on overhead quite a bit:

23047a96d7cfcfca 063f6715e77a7be5770d6081fe
---------------- --------------------------
         %stddev     %change         %stddev
             \          |                \
  21621405 +- 0%     +11.3%   24069657 +- 2%  vm-scalability.throughput

[linux@roeck-us.net: drop unnecessary include file]
[hannes@cmpxchg.org: add WARN_ON_ONCE()s]
Link: http://lkml.kernel.org/r/20160707194024.GA26580@cmpxchg.org
Link: http://lkml.kernel.org/r/20160624175101.GA3024@cmpxchg.org
Reported-by: Ye Xiaolong <xiaolong.ye@intel.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: update the comment in __isolate_free_page

We need to assure the comment is consistent with the code.

[akpm@linux-foundation.org: coding-style fixes]
Link: http://lkml.kernel.org/r/1466171914-21027-1-git-send-email-zhongjiang@huawei.com
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom: tighten task_will_free_mem() locking

"mm, oom: fortify task_will_free_mem" has dropped task_lock around
task_will_free_mem in oom_kill_process bacause it assumed that a
potential race when the selected task exits will not be a problem as the
oom_reaper will call exit_oom_victim.

Tetsuo was objecting that nommu doesn't have oom_reaper so the race
would be still possible.  The code would be racy and lockup prone
theoretically in other aspects without the oom reaper anyway so I didn't
considered this a big deal.  But it seems that further changes I am
planning in this area will benefit from stable task->mm in this path as
well.  So let's drop find_lock_task_mm from task_will_free_mem and call
it from under task_lock as we did previously.  Just pull the task->mm !=
NULL check inside the function.

Link: http://lkml.kernel.org/r/1467201562-6709-1-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@parallels.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom: hide mm which is shared with kthread or global init

The only case where the oom_reaper is not triggered for the oom victim
is when it shares the memory with a kernel thread (aka use_mm) or with
the global init.  After "mm, oom: skip vforked tasks from being
selected" the victim cannot be a vforked task of the global init so we
are left with clone(CLONE_VM) (without CLONE_SIGHAND).  use_mm() users
are quite rare as well.

In order to help forward progress for the OOM killer, make sure that
this really rare case will not get in the way - we do this by hiding the
mm from the oom killer by setting MMF_OOM_REAPED flag for it.
oom_scan_process_thread will ignore any TIF_MEMDIE task if it has
MMF_OOM_REAPED flag set to catch these oom victims.

After this patch we should guarantee forward progress for the OOM killer
even when the selected victim is sharing memory with a kernel thread or
global init as long as the victims mm is still alive.

Link: http://lkml.kernel.org/r/1466426628-15074-11-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom_reaper: do not attempt to reap a task more than twice

oom_reaper relies on the mmap_sem for read to do its job.  Many places
which might block readers have been converted to use down_write_killable
and that has reduced chances of the contention a lot.  Some paths where
the mmap_sem is held for write can take other locks and they might
either be not prepared to fail due to fatal signal pending or too
impractical to be changed.

This patch introduces MMF_OOM_NOT_REAPABLE flag which gets set after the
first attempt to reap a task's mm fails.  If the flag is present after
the failure then we set MMF_OOM_REAPED to hide this mm from the oom
killer completely so it can go and chose another victim.

As a result a risk of OOM deadlock when the oom victim would be blocked
indefinetly and so the oom killer cannot make any progress should be
mitigated considerably while we still try really hard to perform all
reclaim attempts and stay predictable in the behavior.

Link: http://lkml.kernel.org/r/1466426628-15074-10-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom: task_will_free_mem should skip oom_reaped tasks

The 0-day robot has encountered the following:

   Out of memory: Kill process 3914 (trinity-c0) score 167 or sacrifice child
   Killed process 3914 (trinity-c0) total-vm:55864kB, anon-rss:1512kB, file-rss:1088kB, shmem-rss:25616kB
   oom_reaper: reaped process 3914 (trinity-c0), now anon-rss:0kB, file-rss:0kB, shmem-rss:26488kB
   oom_reaper: reaped process 3914 (trinity-c0), now anon-rss:0kB, file-rss:0kB, shmem-rss:26900kB
   oom_reaper: reaped process 3914 (trinity-c0), now anon-rss:0kB, file-rss:0kB, shmem-rss:26900kB
   oom_reaper: reaped process 3914 (trinity-c0), now anon-rss:0kB, file-rss:0kB, shmem-rss:27296kB
   oom_reaper: reaped process 3914 (trinity-c0), now anon-rss:0kB, file-rss:0kB, shmem-rss:28148kB

oom_reaper is trying to reap the same task again and again.

This is possible only when the oom killer is bypassed because of
task_will_free_mem because we skip over tasks with MMF_OOM_REAPED
already set during select_bad_process.  Teach task_will_free_mem to skip
over MMF_OOM_REAPED tasks as well because they will be unlikely to free
anything more.

Analyzed by Tetsuo Handa.

Link: http://lkml.kernel.org/r/1466426628-15074-9-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom: fortify task_will_free_mem()

task_will_free_mem is rather weak.  It doesn't really tell whether the
task has chance to drop its mm.  98748bd72200 ("oom: consider
multi-threaded tasks in task_will_free_mem") made a first step into making
it more robust for multi-threaded applications so now we know that the
whole process is going down and probably drop the mm.

This patch builds on top for more complex scenarios where mm is shared
between different processes - CLONE_VM without CLONE_SIGHAND, or in kernel
use_mm().

Make sure that all processes sharing the mm are killed or exiting.  This
will allow us to replace try_oom_reaper by wake_oom_reaper because
task_will_free_mem implies the task is reapable now.  Therefore all paths
which bypass the oom killer are now reapable and so they shouldn't lock up
the oom killer.

Link: http://lkml.kernel.org/r/1466426628-15074-8-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom: kill all tasks sharing the mm

Currently oom_kill_process skips both the oom reaper and SIG_KILL if a
process sharing the same mm is unkillable via OOM_ADJUST_MIN.  After "mm,
oom_adj: make sure processes sharing mm have same view of oom_score_adj"
all such processes are sharing the same value so we shouldn't see such a
task at all (oom_badness would rule them out).

We can still encounter oom disabled vforked task which has to be killed as
well if we want to have other tasks sharing the mm reapable because it can
access the memory before doing exec.  Killing such a task should be
acceptable because it is highly unlikely it has done anything useful
because it cannot modify any memory before it calls exec.  An alternative
would be to keep the task alive and skip the oom reaper and risk all the
weird corner cases where the OOM killer cannot make forward progress
because the oom victim hung somewhere on the way to exit.

[rientjes@google.com - drop printk when OOM_SCORE_ADJ_MIN killed task
 the setting is inherently racy and we cannot do much about it without
 introducing locks in hot paths]
Link: http://lkml.kernel.org/r/1466426628-15074-7-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom: skip vforked tasks from being selected

vforked tasks are not really sitting on any memory.  They are sharing the
mm with parent until they exec into a new code.  Until then it is just
pinning the address space.  OOM killer will kill the vforked task along
with its parent but we still can end up selecting vforked task when the
parent wouldn't be selected.  E.g.  init doing vfork to launch a task or
vforked being a child of oom unkillable task with an updated oom_score_adj
to be killable.

Add a new helper to check whether a task is in the vfork sharing memory
with its parent and use it in oom_badness to skip over these tasks.

Link: http://lkml.kernel.org/r/1466426628-15074-6-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom_adj: make sure processes sharing mm have same view of oom_score_adj

oom_score_adj is shared for the thread groups (via struct signal) but this
is not sufficient to cover processes sharing mm (CLONE_VM without
CLONE_SIGHAND) and so we can easily end up in a situation when some
processes update their oom_score_adj and confuse the oom killer.  In the
worst case some of those processes might hide from the oom killer
altogether via OOM_SCORE_ADJ_MIN while others are eligible.  OOM killer
would then pick up those eligible but won't be allowed to kill others
sharing the same mm so the mm wouldn't release the mm and so the memory.

It would be ideal to have the oom_score_adj per mm_struct because that is
the natural entity OOM killer considers.  But this will not work because
some programs are doing

vfork()
set_oom_adj()
exec()

We can achieve the same though.  oom_score_adj write handler can set the
oom_score_adj for all processes sharing the same mm if the task is not in
the middle of vfork.  As a result all the processes will share the same
oom_score_adj.  The current implementation is rather pessimistic and
checks all the existing processes by default if there is more than 1
holder of the mm but we do not have any reliable way to check for external
users yet.

Link: http://lkml.kernel.org/r/1466426628-15074-5-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

proc, oom_adj: extract oom_score_adj setting into a helper

Currently we have two proc interfaces to set oom_score_adj.  The legacy
/proc/<pid>/oom_adj and /proc/<pid>/oom_score_adj which both have their
specific handlers.  Big part of the logic is duplicated so extract the
common code into __set_oom_adj helper.  Legacy knob still expects some
details slightly different so make sure those are handled same way - e.g.
the legacy mode ignores oom_score_adj_min and it warns about the usage.

This patch shouldn't introduce any functional changes.

Link: http://lkml.kernel.org/r/1466426628-15074-4-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

proc, oom: drop bogus sighand lock

Oleg has pointed out that can simplify both oom_adj_{read,write} and
oom_score_adj_{read,write} even further and drop the sighand lock.  The
main purpose of the lock was to protect p->signal from going away but this
will not happen since ea6d290ca34c ("signals: make task_struct->signal
immutable/refcountable").

The other role of the lock was to synchronize different writers,
especially those with CAP_SYS_RESOURCE.  Introduce a mutex for this
purpose.  Later patches will need this lock anyway.

Suggested-by: Oleg Nesterov <oleg@redhat.com>
Link: http://lkml.kernel.org/r/1466426628-15074-3-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Vladimir Davydov <vdavydov@virtuozzo.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

proc, oom: drop bogus task_lock and mm check

Series "Handle oom bypass more gracefully", V5

The following 10 patches should put some order to very rare cases of mm
shared between processes and make the paths which bypass the oom killer
oom reapable and therefore much more reliable finally.  Even though mm
shared outside of thread group is rare (either vforked tasks for a short
period, use_mm by kernel threads or exotic thread model of
clone(CLONE_VM) without CLONE_SIGHAND) it is better to cover them.  Not
only it makes the current oom killer logic quite hard to follow and
reason about it can lead to weird corner cases.  E.g.  it is possible to
select an oom victim which shares the mm with unkillable process or
bypass the oom killer even when other processes sharing the mm are still
alive and other weird cases.

Patch 1 drops bogus task_lock and mm check from oom_{score_}adj_write.
This can be considered a bug fix with a low impact as nobody has noticed
for years.

Patch 2 drops sighand lock because it is not needed anymore as pointed
by Oleg.

Patch 3 is a clean up of oom_score_adj handling and a preparatory work
for later patches.

Patch 4 enforces oom_adj_score to be consistent between processes
sharing the mm to behave consistently with the regular thread groups.
This can be considered a user visible behavior change because one thread
group updating oom_score_adj will affect others which share the same mm
via clone(CLONE_VM).  I argue that this should be acceptable because we
already have the same behavior for threads in the same thread group and
sharing the mm without signal struct is just a different model of
threading.  This is probably the most controversial part of the series,
I would like to find some consensus here.  There were some suggestions
to hook some counter/oom_score_adj into the mm_struct but I feel that
this is not necessary right now and we can rely on proc handler +
oom_kill_process to DTRT.  I can be convinced otherwise but I strongly
think that whatever we do the userspace has to have a way to see the
current oom priority as consistently as possible.

Patch 5 makes sure that no vforked task is selected if it is sharing the
mm with oom unkillable task.

Patch 6 ensures that all user tasks sharing the mm are killed which in
turn makes sure that all oom victims are oom reapable.

Patch 7 guarantees that task_will_free_mem will always imply reapable
bypass of the oom killer.

Patch 8 is new in this version and it addresses an issue pointed out by
0-day OOM report where an oom victim was reaped several times.

Patch 9 puts an upper bound on how many times oom_reaper tries to reap a
task and hides it from the oom killer to move on when no progress can be
made.  This will give an upper bound to how long an oom_reapable task
can block the oom killer from selecting another victim if the oom_reaper
is not able to reap the victim.

Patch 10 tries to plug the (hopefully) last hole when we can still lock
up when the oom victim is shared with oom unkillable tasks (kthreads and
global init).  We just try to be best effort in that case and rather
fallback to kill something else than risk a lockup.

This patch (of 10):

Both oom_adj_write and oom_score_adj_write are using task_lock, check for
task->mm and fail if it is NULL.  This is not needed because the
oom_score_adj is per signal struct so we do not need mm at all.  The code
has been introduced by 3d5992d2ac7d ("oom: add per-mm oom disable count")
but we do not do per-mm oom disable since c9f01245b6a7 ("oom: remove
oom_disable_count").

The task->mm check is even not correct because the current thread might
have exited but the thread group might be still alive - e.g.  thread group
leader would lead that echo $VAL > /proc/pid/oom_score_adj would always
fail with EINVAL while /proc/pid/task/$other_tid/oom_score_adj would
succeed.  This is unexpected at best.

Remove the lock along with the check to fix the unexpected behavior and
also because there is not real need for the lock in the first place.

Link: http://lkml.kernel.org/r/1466426628-15074-2-git-send-email-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Vladimir Davydov <vdavydov@virtuozzo.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add braces to avoid "ambiguous ‘else’" compiler warnings

Some of our "for_each_xyz()" macro constructs make gcc unhappy about
lack of braces around if-statements inside or outside the loop, because
the loop construct itself has a "if-then-else" statement inside of it.

The resulting warnings look something like this:

  drivers/gpu/drm/i915/i915_debugfs.c: In function ‘i915_dump_lrc’:
  drivers/gpu/drm/i915/i915_debugfs.c:2103:6: warning: suggest explicit braces to avoid ambiguous ‘else’ [-Wparentheses]
     if (ctx != dev_priv->kernel_context)
        ^

even if the code itself is fine.

Since the warning is fairly easy to avoid by adding a braces around the
if-statement near the for_each_xyz() construct, do so, rather than
disabling the otherwise potentially useful warning.

(The if-then-else statements used in the "for_each_xyz()" constructs are
designed to be inherently safe even with no braces, but in this case
it's quite understandable that gcc isn't really able to tell that).

This finally leaves the standard "allmodconfig" build with just a
handful of remaining warnings, so new and valid warnings hopefully will
stand out.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>