mm, hugetlb: unify core page allocation accounting and initialization

Patch series "mm, hugetlb: allocation API and migration improvements"

Motivation:

this is a follow up for [3] for the allocation API and [4] for the
hugetlb migration.  It wasn't really easy to split those into two
separate patch series as they share some code.

My primary motivation to touch this code is to make the gigantic pages
migration working.  The giga pages allocation code is just too fragile
and hacked into the hugetlb code now.  This series tries to move giga
pages closer to the first class citizen.  We are not there yet but
having 5 patches is quite a lot already and it will already make the
code much easier to follow.  I will come with other changes on top after
this sees some review.

The first two patches should be trivial to review.  The third patch
changes the way how we migrate huge pages.  Newly allocated pages are a
subject of the overcommit check and they participate surplus accounting
which is quite unfortunate as the changelog explains.  This patch
doesn't change anything wrt.  giga pages.

Patch #4 removes the surplus accounting hack from
__alloc_surplus_huge_page.  I hope I didn't miss anything there and a
deeper review is really due there.

Patch #5 finally unifies allocation paths and giga pages shouldn't be
any special anymore.  There is also some renaming going on as well.

This patch (of 6):

hugetlb allocator has two entry points to the page allocator
 - alloc_fresh_huge_page_node
 - __hugetlb_alloc_buddy_huge_page

The two differ very subtly in two aspects.  The first one doesn't care
about HTLB_BUDDY_* stats and it doesn't initialize the huge page.
prep_new_huge_page is not used because it not only initializes hugetlb
specific stuff but because it also put_page and releases the page to the
hugetlb pool which is not what is required in some contexts.  This makes
things more complicated than necessary.

Simplify things by a) removing the page allocator entry point duplicity
and only keep __hugetlb_alloc_buddy_huge_page and b) make
prep_new_huge_page more reusable by removing the put_page which moves
the page to the allocator pool.  All current callers are updated to call
put_page explicitly.  Later patches will add new callers which won't
need it.

This patch shouldn't introduce any functional change.

Link: http://lkml.kernel.org/r/20180103093213.26329-2-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Reale <ar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcontrol.c: try harder to decrease [memory,memsw].limit_in_bytes

mem_cgroup_resize_[memsw]_limit() tries to free only 32
(SWAP_CLUSTER_MAX) pages on each iteration.  This makes it practically
impossible to decrease limit of memory cgroup.  Tasks could easily
allocate back 32 pages, so we can't reduce memory usage, and once
retry_count reaches zero we return -EBUSY.

Easy to reproduce the problem by running the following commands:

  mkdir /sys/fs/cgroup/memory/test
  echo $$ >> /sys/fs/cgroup/memory/test/tasks
  cat big_file > /dev/null &
  sleep 1 && echo $((100*1024*1024)) > /sys/fs/cgroup/memory/test/memory.limit_in_bytes
  -bash: echo: write error: Device or resource busy

Instead of relying on retry_count, keep retrying the reclaim until the
desired limit is reached or fail if the reclaim doesn't make any
progress or a signal is pending.

Link: http://lkml.kernel.org/r/20180119132544.19569-1-aryabinin@virtuozzo.com
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcontrol.c: make local symbol static

Fix the following sparse warning:

  mm/memcontrol.c:1097:14: warning: symbol 'memcg1_stats' was not declared. Should it be static?

Link: http://lkml.kernel.org/r/20180118193327.14200-1-chrisadr@gentoo.org
Signed-off-by: Christopher Díaz Riveros <chrisadr@gentoo.org>
Cc: Michal Hocko <mhocko@kernel.org>
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/hmm: fix uninitialized use of 'entry' in hmm_vma_walk_pmd()

The variable 'entry' is used before being initialized in
hmm_vma_walk_pmd().

No bad effect (beside performance hit) so !non_swap_entry(0) evaluate to
true which trigger a fault as if CPU was trying to access migrated
memory and migrate memory back from device memory to regular memory.

This function (hmm_vma_walk_pmd()) is called when a device driver tries
to populate its own page table.  For migrated memory it should not
happen as the device driver should already have populated its page table
correctly during the migration.

Only case I can think of is multi-GPU where a second GPU triggers
migration back to regular memory.  Again this would just result in a
performance hit, nothing bad would happen.

Link: http://lkml.kernel.org/r/20180122185759.26286-1-jglisse@redhat.com
Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/mmzone.h: fix explanation of lower bits in the SPARSEMEM mem_map pointer

The comment is confusing.  On the one hand, it refers to 32-bit
alignment (struct page alignment on 32-bit platforms), but this would
only guarantee that the 2 lowest bits must be zero.  On the other hand,
it claims that at least 3 bits are available, and 3 bits are actually
used.

This is not broken, because there is a stronger alignment guarantee,
just less obvious.  Let's fix the comment to make it clear how many bits
are available and why.

Although memmap arrays are allocated in various places, the resulting
pointer is encoded eventually, so I am adding a BUG_ON() here to enforce
at runtime that all expected bits are indeed available.

I have also added a BUILD_BUG_ON to check that PFN_SECTION_SHIFT is
sufficient, because this part of the calculation can be easily checked
at build time.

[ptesarik@suse.com: v2]
Link: http://lkml.kernel.org/r/20180125100516.589ea6af@ezekiel.suse.cz
Link: http://lkml.kernel.org/r/20180119080908.3a662e6f@ezekiel.suse.cz
Signed-off-by: Petr Tesarik <ptesarik@suse.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: YASUAKI ISHIMATSU <yasu.isimatu@gmail.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/compaction.c: fix comment for try_to_compact_pages()

"mode" argument is not used by try_to_compact_pages() and sub functions
anymore, it has been replaced by "prio".  Fix the comment to explain the
use of "prio" argument.

Link: http://lkml.kernel.org/r/1515801336-20611-1-git-send-email-yang.shi@linux.alibaba.com
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/page_ext.c: make page_ext_init a noop when CONFIG_PAGE_EXTENSION but nothing uses it

static struct page_ext_operations *page_ext_ops[] always contains debug_guardpage_ops,

static struct page_ext_operations *page_ext_ops[] = {
        &debug_guardpage_ops,
 #ifdef CONFIG_PAGE_OWNER
        &page_owner_ops,
 #endif
...
}

but for it to work, CONFIG_DEBUG_PAGEALLOC must be enabled first.  If
someone has CONFIG_PAGE_EXTENSION, but has none of its users, eg:
(CONFIG_PAGE_OWNER, CONFIG_DEBUG_PAGEALLOC, CONFIG_IDLE_PAGE_TRACKING),
we can shrink page_ext_init() to a simple retq.

  $ size vmlinux  (before patch)
        text      data       bss       dec       hex  filename
    14356698   5681582   1687748  21726028   14b834c  vmlinux

  $ size vmlinux  (after patch)
        text      data       bss       dec       hex  filename
    14356008   5681538   1687748  21725294   14b806e  vmlinux

On the other hand, it might does not even make sense, since if someone
enables CONFIG_PAGE_EXTENSION, I would expect him to enable also at
least one of its users.

Link: http://lkml.kernel.org/r/20180105130235.GA21241@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@techadventures.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zsmalloc: use U suffix for negative literals being shifted

Fix warning about shifting unsigned literals being undefined behavior.

Link: http://lkml.kernel.org/r/1515642078-4259-1-git-send-email-nick.desaulniers@gmail.com
Signed-off-by: Nick Desaulniers <nick.desaulniers@gmail.com>
Suggested-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Andy Shevchenko <andy.shevchenko@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Nick Desaulniers <nick.desaulniers@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/page_owner.c: clean up init_pages_in_zone()

Remove two redundant assignments in init_pages_in_zone().

[osalvador@techadventures.net: v3]
Link: http://lkml.kernel.org/r/20180117124513.GA876@techadventures.net
[akpm@linux-foundation.org: coding style tweaks]
Link: http://lkml.kernel.org/r/20180110084355.GA22822@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@techadventures.net>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/page_alloc.c: fix typos in comments

Link: http://lkml.kernel.org/r/1515485774-4768-1-git-send-email-zhangshile@gmail.com
Signed-off-by: Shile Zhang <zhangshile@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: refactor mem_cgroup_resize_limit()

mem_cgroup_resize_limit() and mem_cgroup_resize_memsw_limit() have
identical logics.  Refactor code so we don't need to keep two pieces of
code that does same thing.

Link: http://lkml.kernel.org/r/20180108224238.14583-1-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zswap: only save zswap header when necessary

We waste sizeof(swp_entry_t) for zswap header when using zsmalloc as
zpool driver because zsmalloc doesn't support eviction.

Add zpool_evictable() to detect if zpool is potentially evictable, and
use it in zswap to avoid waste memory for zswap header.

[yuzhao@google.com: The zpool->" prefix is a result of copy & paste]
Link: http://lkml.kernel.org/r/20180110225626.110330-1-yuzhao@google.com
Link: http://lkml.kernel.org/r/20180110224741.83751-1-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Dan Streetman <ddstreet@ieee.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Seth Jennings <sjenning@redhat.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/fadvise: discard partial page if endbyte is also EOF

During our recent testing with fadvise(FADV_DONTNEED), we find that if
given offset/length is not page-aligned, the last page will not be
discarded.  The tool we use is vmtouch (https://hoytech.com/vmtouch/),
we map a 10KB-sized file into memory and then try to run this tool to
evict the whole file mapping, but the last single page always remains
staying in the memory:

$./vmtouch -e test_10K
           Files: 1
     Directories: 0
   Evicted Pages: 3 (12K)
         Elapsed: 2.1e-05 seconds

$./vmtouch test_10K
           Files: 1
     Directories: 0
  Resident Pages: 1/3  4K/12K  33.3%
         Elapsed: 5.5e-05 seconds

However when we test with an older kernel, say 3.10, this problem is
gone.  So we wonder if this is a regression:

$./vmtouch -e test_10K
           Files: 1
     Directories: 0
   Evicted Pages: 3 (12K)
         Elapsed: 8.2e-05 seconds

$./vmtouch test_10K
           Files: 1
     Directories: 0
  Resident Pages: 0/3  0/12K  0%  <-- partial page also discarded
         Elapsed: 5e-05 seconds

After digging a little bit into this problem, we find it seems not a
regression.  Not discarding partial page is likely to be on purpose
according to commit 441c228f817f ("mm: fadvise: document the
fadvise(FADV_DONTNEED) behaviour for partial pages") written by Mel
Gorman.  He explained why partial pages should be preserved instead of
being discarded when using fadvise(FADV_DONTNEED).

However, the interesting part is that the actual code did NOT work as
the same as it was described, the partial page was still discarded
anyway, due to a calculation mistake of `end_index' passed to
invalidate_mapping_pages().  This mistake has not been fixed until
recently, that's why we fail to reproduce our problem in old kernels.
The fix is done in commit 18aba41cbf ("mm/fadvise.c: do not discard
partial pages with POSIX_FADV_DONTNEED") by Oleg Drokin.

Back to the original testing, our problem becomes that there is a
special case that, if the page-unaligned `endbyte' is also the end of
file, it is not necessary at all to preserve the last partial page, as
we all know no one else will use the rest of it.  It should be safe
enough if we just discard the whole page.  So we add an EOF check in
this patch.

We also find a poosbile real world issue in mainline kernel.  Assume
such scenario: A userspace backup application want to backup a huge
amount of small files (<4k) at once, the developer might (I guess) want
to use fadvise(FADV_DONTNEED) to save memory.  However, FADV_DONTNEED
won't really happen since the only page mapped is a partial page, and
kernel will preserve it.  Our patch also fixes this problem, since we
know the endbyte is EOF, so we discard it.

Here is a simple reproducer to reproduce and verify each scenario we
described above:

  test_fadvise.c
  ==============================
  #include <sys/mman.h>
  #include <sys/stat.h>
  #include <fcntl.h>
  #include <stdlib.h>
  #include <string.h>
  #include <stdio.h>
  #include <unistd.h>

  int main(int argc, char **argv)
  {
   int i, fd, ret, len;
   struct stat buf;
   void *addr;
   unsigned char *vec;
   char *strbuf;
   ssize_t pagesize = getpagesize();
   ssize_t filesize;

   fd = open(argv[1], O_RDWR|O_CREAT, S_IRUSR|S_IWUSR);
   if (fd < 0)
   return -1;
   filesize = strtoul(argv[2], NULL, 10);

   strbuf = malloc(filesize);
   memset(strbuf, 42, filesize);
   write(fd, strbuf, filesize);
   free(strbuf);
   fsync(fd);

   len = (filesize + pagesize - 1) / pagesize;
   printf("length of pages: %d\n", len);

   addr = mmap(NULL, filesize, PROT_READ, MAP_SHARED, fd, 0);
   if (addr == MAP_FAILED)
   return -1;

   ret = posix_fadvise(fd, 0, filesize, POSIX_FADV_DONTNEED);
   if (ret < 0)
   return -1;

   vec = malloc(len);
   ret = mincore(addr, filesize, (void *)vec);
   if (ret < 0)
   return -1;

   for (i = 0; i < len; i++)
   printf("pages[%d]: %x\n", i, vec[i] & 0x1);

   free(vec);
   close(fd);

   return 0;
  }
  ==============================

Test 1: running on kernel with commit 18aba41cbf reverted:

  [root@caspar ~]# uname -r
  4.15.0-rc6.revert+
  [root@caspar ~]# ./test_fadvise file1 1024
  length of pages: 1
  pages[0]: 0    # <-- partial page discarded
  [root@caspar ~]# ./test_fadvise file2 8192
  length of pages: 2
  pages[0]: 0
  pages[1]: 0
  [root@caspar ~]# ./test_fadvise file3 10240
  length of pages: 3
  pages[0]: 0
  pages[1]: 0
  pages[2]: 0    # <-- partial page discarded

Test 2: running on mainline kernel:

  [root@caspar ~]# uname -r
  4.15.0-rc6+
  [root@caspar ~]# ./test_fadvise test1 1024
  length of pages: 1
  pages[0]: 1    # <-- partial and the only page not discarded
  [root@caspar ~]# ./test_fadvise test2 8192
  length of pages: 2
  pages[0]: 0
  pages[1]: 0
  [root@caspar ~]# ./test_fadvise test3 10240
  length of pages: 3
  pages[0]: 0
  pages[1]: 0
  pages[2]: 1    # <-- partial page not discarded

Test 3: running on kernel with this patch:

  [root@caspar ~]# uname -r
  4.15.0-rc6.patched+
  [root@caspar ~]# ./test_fadvise test1 1024
  length of pages: 1
  pages[0]: 0    # <-- partial page and EOF, discarded
  [root@caspar ~]# ./test_fadvise test2 8192
  length of pages: 2
  pages[0]: 0
  pages[1]: 0
  [root@caspar ~]# ./test_fadvise test3 10240
  length of pages: 3
  pages[0]: 0
  pages[1]: 0
  pages[2]: 0    # <-- partial page and EOF, discarded

[akpm@linux-foundation.org: tweak code comment]
Link: http://lkml.kernel.org/r/5222da9ee20e1695eaabb69f631f200d6e6b8876.1515132470.git.jinli.zjl@alibaba-inc.com
Signed-off-by: shidao.ytt <shidao.ytt@alibaba-inc.com>
Signed-off-by: Caspar Zhang <jinli.zjl@alibaba-inc.com>
Reviewed-by: Oliver Yang <zhiche.yy@alibaba-inc.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: pin address_space before dereferencing it while isolating an LRU page

Minchan Kim asked the following question -- what locks protects
address_space destroying when race happens between inode trauncation and
__isolate_lru_page? Jan Kara clarified by describing the race as follows

CPU1                                            CPU2

truncate(inode)                                 __isolate_lru_page()
  ...
  truncate_inode_page(mapping, page);
    delete_from_page_cache(page)
      spin_lock_irqsave(&mapping->tree_lock, flags);
        __delete_from_page_cache(page, NULL)
          page_cache_tree_delete(..)
            ...                                   mapping = page_mapping(page);
            page->mapping = NULL;
            ...
      spin_unlock_irqrestore(&mapping->tree_lock, flags);
      page_cache_free_page(mapping, page)
        put_page(page)
          if (put_page_testzero(page)) -> false
- inode now has no pages and can be freed including embedded address_space

                                                  if (mapping && !mapping->a_ops->migratepage)
- we've dereferenced mapping which is potentially already free.

The race is theoretically possible but unlikely.  Before the
delete_from_page_cache, truncate_cleanup_page is called so the page is
likely to be !PageDirty or PageWriteback which gets skipped by the only
caller that checks the mappping in __isolate_lru_page.  Even if the race
occurs, a substantial amount of work has to happen during a tiny window
with no preemption but it could potentially be done using a virtual
machine to artifically slow one CPU or halt it during the critical
window.

This patch should eliminate the race with truncation by try-locking the
page before derefencing mapping and aborting if the lock was not
acquired.  There was a suggestion from Huang Ying to use RCU as a
side-effect to prevent mapping being freed.  However, I do not like the
solution as it's an unconventional means of preserving a mapping and
it's not a context where rcu_read_lock is obviously protecting rcu data.

Link: http://lkml.kernel.org/r/20180104102512.2qos3h5vqzeisrek@techsingularity.net
Fixes: c82449352854 ("mm: compaction: make isolate_lru_page() filter-aware again")
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

userfaultfd: convert to use anon_inode_getfd()

Nothing actually calls userfaultfd_file_create() besides the
userfaultfd() system call itself.  So simplify things by folding it into
the system call and using anon_inode_getfd() instead of
anon_inode_getfile().  Do the same in resolve_userfault_fork() as well.

This removes over 50 lines with no change in functionality.

Link: http://lkml.kernel.org/r/20171229212403.22800-1-ebiggers3@gmail.com
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memfd-test: run fuse test on hugetlb backend memory

Link: http://lkml.kernel.org/r/20171107122800.25517-10-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memfd-test: move common code to a shared unit

The memfd & fuse tests will share more common code in the following
commits to test hugetlb support.

Link: http://lkml.kernel.org/r/20171107122800.25517-9-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memfd-test: add 'memfd-hugetlb:' prefix when testing hugetlbfs

Link: http://lkml.kernel.org/r/20171107122800.25517-8-marcandre.lureau@redhat.com
Suggested-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memfd-test: test hugetlbfs sealing

Remove most of the special-casing of hugetlbfs now that sealing is
supported.

Link: http://lkml.kernel.org/r/20171107122800.25517-7-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

shmem: add sealing support to hugetlb-backed memfd

Adapt add_seals()/get_seals() to work with hugetbfs-backed memory.

Teach memfd_create() to allow sealing operations on MFD_HUGETLB.

Link: http://lkml.kernel.org/r/20171107122800.25517-6-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: implement memfd sealing

Implements memfd sealing, similar to shmem:
 - WRITE: deny fallocate(PUNCH_HOLE). mmap() write is denied in
   memfd_add_seals(). write() doesn't exist for hugetlbfs.
 - SHRINK: added similar check as shmem_setattr()
 - GROW: added similar check as shmem_setattr() & shmem_fallocate()

Except write() operation that doesn't exist with hugetlbfs, that should
make sealing as close as it can be to shmem support.

Link: http://lkml.kernel.org/r/20171107122800.25517-5-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: expose hugetlbfs_inode_info in header

hugetlbfs inode information will need to be accessed by code in
mm/shmem.c for file sealing operations.  Move inode information
definition from .c file to header for needed access.

Link: http://lkml.kernel.org/r/20171107122800.25517-4-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

shmem: rename functions that are memfd-related

Those functions are called for memfd files, backed by shmem or hugetlb
(the next patches will handle hugetlb).

Link: http://lkml.kernel.org/r/20171107122800.25517-3-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

shmem: unexport shmem_add_seals()/shmem_get_seals()

Patch series "memfd: add sealing to hugetlb-backed memory", v3.

Recently, Mike Kravetz added hugetlbfs support to memfd.  However, he
didn't add sealing support.  One of the reasons to use memfd is to have
shared memory sealing when doing IPC or sharing memory with another
process with some extra safety.  qemu uses shared memory & hugetables
with vhost-user (used by dpdk), so it is reasonable to use memfd now
instead for convenience and security reasons.

This patch (of 9):

The functions are called through shmem_fcntl() only.  And no danger in
removing the EXPORTs as the routines only work with shmem file structs.

Link: http://lkml.kernel.org/r/20171107122800.25517-2-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove reference to PG_buddy

PG_buddy doesn't exist any more.  It's called PageBuddy now.

Link: http://lkml.kernel.org/r/20171220155552.15884-9-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: document how to use struct page

Be really explicit about what bits / bytes are reserved for users that
want to store extra information about the pages they allocate.

Link: http://lkml.kernel.org/r/20171220155552.15884-8-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: store compound_dtor / compound_order as bytes

Neither of these values get even close to 256; compound_dtor is
currently at a maximum of 3, and compound_order can't be over 64.  No
machine has inefficient access to bytes since EV5, and while those are
still supported, we don't optimise for them any more.  This does not
shrink struct page, but it removes an ifdef and frees up 2-6 bytes for
future use.

diff of pahole output:

  struct callback_head callback_head;      /*    32    16 */
  struct {
  long unsigned int compound_head; /*    32     8 */
- unsigned int compound_dtor;      /*    40     4 */
- unsigned int compound_order;     /*    44     4 */
+ unsigned char compound_dtor;     /*    40     1 */
+ unsigned char compound_order;    /*    41     1 */
  };                                       /*    32    16 */
  };                                               /*    32    16 */
  union {

[mawilcox@microsoft.com: add comment]
Link: http://lkml.kernel.org/r/20171221000144.GB2980@bombadil.infradead.org
Link: http://lkml.kernel.org/r/20171220155552.15884-7-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: introduce _slub_counter_t

Instead of putting the ifdef in the middle of the definition of struct
page, pull it forward to the rest of the ifdeffery around the SLUB
cmpxchg_double optimisation.

Link: http://lkml.kernel.org/r/20171220155552.15884-6-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: improve comment on page->mapping

The comment on page->mapping is terse, and out of date (it does not
mention the possibility of PAGE_MAPPING_MOVABLE).  Instead, point the
interested reader to page-flags.h where there is a much better comment.

Link: http://lkml.kernel.org/r/20171220155552.15884-5-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove misleading alignment claims

The "third double word block" isn't on 32-bit systems.  The layout looks
like this:

unsigned long flags;
struct address_space *mapping
pgoff_t index;
atomic_t _mapcount;
atomic_t _refcount;

which is 32 bytes on 64-bit, but 20 bytes on 32-bit.  Nobody is trying to
use the fact that it's double-word aligned today, so just remove the
misleading claims.

Link: http://lkml.kernel.org/r/20171220155552.15884-4-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: de-indent struct page

I found the struct { union { struct { union { struct { } } } } } layout
rather confusing.  Fortunately, there is an easier way to write this.

The innermost union is of four things which are the size of an int, so
the ones which are used by slab/slob/slub can be pulled up two levels to
be in the outermost union with 'counters'.  That leaves us with struct {
union { struct { atomic_t; atomic_t; } } } which has the same layout,
but is easier to read.

Output from the current git version of pahole, diffed with -uw to ignore
the whitespace changes from the indentation:

  }; /*    16     8 */
  union {
  long unsigned int  counters; /*    24     8 */
- struct {
- union {
- atomic_t _mapcount; /*    24     4 */
  unsigned int active; /*    24     4 */
  struct {
  unsigned int inuse:16; /*    24:16  4 */
@@ -21,7 +18,8 @@
  unsigned int frozen:1; /*    24: 0  4 */
  }; /*    24     4 */
  int units; /*    24     4 */
- }; /*    24     4 */
+ struct {
+ atomic_t   _mapcount; /*    24     4 */
  atomic_t   _refcount; /*    28     4 */
  }; /*    24     8 */
  }; /*    24     8 */

Link: http://lkml.kernel.org/r/20171220155552.15884-3-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: align struct page more aesthetically

Patch series "Restructure struct page", v2.

This series does not attempt any grand restructuring.  Instead, it cures
the worst of the indentitis, fixes the documentation and reduces the
ifdeffery.  The only layout change is compound_dtor and compound_order
are each reduced to one byte.

This patch (of 8):

Instead of an ifdef block at the end of the struct, which needed its own
comment, define _struct_page_alignment up at the top where it fits
nicely with the existing comment.

Link: http://lkml.kernel.org/r/20171220155552.15884-2-willy@infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc: simplify shrinker init/destroy

Structure zs_pool has special flag to indicate success of shrinker
initialization.  unregister_shrinker() has improved and can detect by
itself whether actual deinitialization should be performed or not, so
extra flag becomes redundant.

[akpm@linux-foundation.org: update comment (Aliaksei), remove unneeded cast]
Link: http://lkml.kernel.org/r/1513680552-9798-1-git-send-email-akaraliou.dev@gmail.com
Signed-off-by: Aliaksei Karaliou <akaraliou.dev@gmail.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom: avoid reaping only for mm's with blockable invalidate callbacks

This uses the new annotation to determine if an mm has mmu notifiers
with blockable invalidate range callbacks to avoid oom reaping.
Otherwise, the callbacks are used around unmap_page_range().

Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1712141330120.74052@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Dimitri Sivanich <sivanich@hpe.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Doug Ledford <dledford@redhat.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Mike Marciniszyn <mike.marciniszyn@intel.com>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Jérôme Glisse <jglisse@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, mmu_notifier: annotate mmu notifiers with blockable invalidate callbacks

Commit 4d4bbd8526a8 ("mm, oom_reaper: skip mm structs with mmu
notifiers") prevented the oom reaper from unmapping private anonymous
memory with the oom reaper when the oom victim mm had mmu notifiers
registered.

The rationale is that doing mmu_notifier_invalidate_range_{start,end}()
around the unmap_page_range(), which is needed, can block and the oom
killer will stall forever waiting for the victim to exit, which may not
be possible without reaping.

That concern is real, but only true for mmu notifiers that have
blockable invalidate_range_{start,end}() callbacks.  This patch adds a
"flags" field to mmu notifier ops that can set a bit to indicate that
these callbacks do not block.

The implementation is steered toward an expensive slowpath, such as
after the oom reaper has grabbed mm->mmap_sem of a still alive oom
victim.

[rientjes@google.com: mmu_notifier_invalidate_range_end() can also call the invalidate_range() must not block, fix comment]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1801091339570.240101@chino.kir.corp.google.com
[akpm@linux-foundation.org: make mm_has_blockable_invalidate_notifiers() return bool, use rwsem_is_locked()]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1712141329500.74052@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Dimitri Sivanich <sivanich@hpe.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Doug Ledford <dledford@redhat.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Mike Marciniszyn <mike.marciniszyn@intel.com>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Jérôme Glisse <jglisse@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-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: thp: use down_read_trylock() in khugepaged to avoid long block

In the current design, khugepaged needs to acquire mmap_sem before
scanning an mm.  But in some corner cases, khugepaged may scan a process
which is modifying its memory mapping, so khugepaged blocks in
uninterruptible state.  But the process might hold the mmap_sem for a
long time when modifying a huge memory space and it may trigger the
below khugepaged hung issue:

  INFO: task khugepaged:270 blocked for more than 120 seconds.
  Tainted: G E 4.9.65-006.ali3000.alios7.x86_64 #1
  "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
  khugepaged D 0 270 2 0x00000000 
  ffff883f3deae4c0 0000000000000000 ffff883f610596c0 ffff883f7d359440
  ffff883f63818000 ffffc90019adfc78 ffffffff817079a5 d67e5aa8c1860a64
  0000000000000246 ffff883f7d359440 ffffc90019adfc88 ffff883f610596c0
  Call Trace:
    schedule+0x36/0x80
    rwsem_down_read_failed+0xf0/0x150
    call_rwsem_down_read_failed+0x18/0x30
    down_read+0x20/0x40
    khugepaged+0x476/0x11d0
    kthread+0xe6/0x100
    ret_from_fork+0x25/0x30

So it sounds pointless to just block khugepaged waiting for the
semaphore so replace down_read() with down_read_trylock() to move to
scan the next mm quickly instead of just blocking on the semaphore so
that other processes can get more chances to install THP.  Then
khugepaged can come back to scan the skipped mm when it has finished the
current round full_scan.

And it appears that the change can improve khugepaged efficiency a
little bit.

Below is the test result when running LTP on a 24 cores 4GB memory 2
nodes NUMA VM:

                                    pristine          w/ trylock
  full_scan                         197               187
  pages_collapsed                   21                26
  thp_fault_alloc                   40818             44466
  thp_fault_fallback                18413             16679
  thp_collapse_alloc                21                150
  thp_collapse_alloc_failed         14                16
  thp_file_alloc                    369               369

[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: tweak comment]
[arnd@arndb.de: avoid uninitialized variable use]
Link: http://lkml.kernel.org/r/20171215125129.2948634-1-arnd@arndb.de
Link: http://lkml.kernel.org/r/1513281203-54878-1-git-send-email-yang.s@alibaba-inc.com
Signed-off-by: Yang Shi <yang.s@alibaba-inc.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/thp: remove pmd_huge_split_prepare()

Instead of marking the pmd ready for split, invalidate the pmd.  This
should take care of powerpc requirement.  Only side effect is that we
mark the pmd invalid early.  This can result in us blocking access to
the page a bit longer if we race against a thp split.

[kirill.shutemov@linux.intel.com: rebased, dirty THP once]
Link: http://lkml.kernel.org/r/20171213105756.69879-13-kirill.shutemov@linux.intel.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Daney <david.daney@cavium.com>
Cc: David Miller <davem@davemloft.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nitin Gupta <nitin.m.gupta@oracle.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.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: use updated pmdp_invalidate() interface to track dirty/accessed bits

Use the modifed pmdp_invalidate() that returns the previous value of pmd
to transfer dirty and accessed bits.

Link: http://lkml.kernel.org/r/20171213105756.69879-12-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Daney <david.daney@cavium.com>
Cc: David Miller <davem@davemloft.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nitin Gupta <nitin.m.gupta@oracle.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.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: do not lose dirty and accessed bits in pmdp_invalidate()

Vlastimil noted that pmdp_invalidate() is not atomic and we can lose
dirty and access bits if CPU sets them after pmdp dereference, but
before set_pmd_at().

The patch change pmdp_invalidate() to make the entry non-present
atomically and return previous value of the entry.  This value can be
used to check if CPU set dirty/accessed bits under us.

The race window is very small and I haven't seen any reports that can be
attributed to the bug.  For this reason, I don't think backporting to
stable trees needed.

Link: http://lkml.kernel.org/r/20171213105756.69879-11-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Daney <david.daney@cavium.com>
Cc: David Miller <davem@davemloft.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nitin Gupta <nitin.m.gupta@oracle.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

x86/mm: provide pmdp_establish() helper

We need an atomic way to setup pmd page table entry, avoiding races with
CPU setting dirty/accessed bits.  This is required to implement
pmdp_invalidate() that doesn't lose these bits.

On PAE we can avoid expensive cmpxchg8b for cases when new page table
entry is not present.  If it's present, fallback to cpmxchg loop.

[akpm@linux-foundation.org: add missing `do' to do-while loop]
Link: http://lkml.kernel.org/r/20171213105756.69879-10-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sparc64: update pmdp_invalidate() to return old pmd value

It's required to avoid losing dirty and accessed bits.

[akpm@linux-foundation.org: add a `do' to the do-while loop]
Link: http://lkml.kernel.org/r/20171213105756.69879-9-kirill.shutemov@linux.intel.com
Signed-off-by: Nitin Gupta <nitin.m.gupta@oracle.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: David Miller <davem@davemloft.net>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

s390/mm: modify pmdp_invalidate to return old value.

It's required to avoid losing dirty and accessed bits.

Link: http://lkml.kernel.org/r/20171213105756.69879-8-kirill.shutemov@linux.intel.com
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

powerpc/mm: update pmdp_invalidate to return old pmd value

It's required to avoid losing dirty and accessed bits.

Link: http://lkml.kernel.org/r/20171213105756.69879-7-kirill.shutemov@linux.intel.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mips: use generic_pmdp_establish as pmdp_establish

MIPS doesn't support hardware dirty/accessed bits.
generic_pmdp_establish() is suitable in this case.

Link: http://lkml.kernel.org/r/20171213105756.69879-6-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: David Daney <david.daney@cavium.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64: provide pmdp_establish() helper

We need an atomic way to setup pmd page table entry, avoiding races with
CPU setting dirty/accessed bits.  This is required to implement
pmdp_invalidate() that doesn't lose these bits.

Link: http://lkml.kernel.org/r/20171213105756.69879-5-kirill.shutemov@linux.intel.com
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm/mm: provide pmdp_establish() helper

ARM LPAE doesn't have hardware dirty/accessed bits.

generic_pmdp_establish() is the right implementation of pmdp_establish
for this case.

Link: http://lkml.kernel.org/r/20171213105756.69879-4-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arc: use generic_pmdp_establish as pmdp_establish

ARC doesn't support hardware dirty/accessed bits.
generic_pmdp_establish() is suitable in this case.

Link: http://lkml.kernel.org/r/20171213105756.69879-3-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

asm-generic: provide generic_pmdp_establish()

Patch series "Do not lose dirty bit on THP pages", v4.

Vlastimil noted that pmdp_invalidate() is not atomic and we can lose
dirty and access bits if CPU sets them after pmdp dereference, but
before set_pmd_at().

The bug can lead to data loss, but the race window is tiny and I haven't
seen any reports that suggested that it happens in reality.  So I don't
think it worth sending it to stable.

Unfortunately, there's no way to address the issue in a generic way.  We
need to fix all architectures that support THP one-by-one.

All architectures that have THP supported have to provide atomic
pmdp_invalidate() that returns previous value.

If generic implementation of pmdp_invalidate() is used, architecture
needs to provide atomic pmdp_estabish().

pmdp_estabish() is not used out-side generic implementation of
pmdp_invalidate() so far, but I think this can change in the future.

This patch (of 12):

This is an implementation of pmdp_establish() that is only suitable for
an architecture that doesn't have hardware dirty/accessed bits.  In this
case we can't race with CPU which sets these bits and non-atomic
approach is fine.

Link: http://lkml.kernel.org/r/20171213105756.69879-2-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Daney <david.daney@cavium.com>
Cc: David Miller <davem@davemloft.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Nitin Gupta <nitin.m.gupta@oracle.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: get 7% more pages in a pagevec

We don't have to use an entire 'long' for the number of elements in the
pagevec; we know it's a number between 0 and 14 (now 15).  So we can
store it in a char, and then the bool packs next to it and we still have
two or six bytes of padding for more elements in the header.  That gives
us space to cram in an extra page.

Link: http://lkml.kernel.org/r/20171206022521.GM26021@bombadil.infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.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: add unmap_mapping_pages()

Several users of unmap_mapping_range() would prefer to express their
range in pages rather than bytes.  Unfortuately, on a 32-bit kernel, you
have to remember to cast your page number to a 64-bit type before
shifting it, and four places in the current tree didn't remember to do
that.  That's a sign of a bad interface.

Conveniently, unmap_mapping_range() actually converts from bytes into
pages, so hoist the guts of unmap_mapping_range() into a new function
unmap_mapping_pages() and convert the callers which want to use pages.

Link: http://lkml.kernel.org/r/20171206142627.GD32044@bombadil.infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Reported-by: "zhangyi (F)" <yi.zhang@huawei.com>
Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, userfaultfd, THP: avoid waiting when PMD under THP migration

If THP migration is enabled, for a VMA handled by userfaultfd, consider
the following situation,

  do_page_fault()
    __do_huge_pmd_anonymous_page()
     handle_userfault()
       userfault_msg()
         /* a huge page is allocated and mapped at fault address */
         /* the huge page is under migration, leaves migration entry
            in page table */
       userfaultfd_must_wait()
         /* return true because !pmd_present() */
       /* may wait in loop until fatal signal */

That is, it may be possible for userfaultfd_must_wait() encounters a PMD
entry which is !pmd_none() && !pmd_present().  In the current
implementation, we will wait for such PMD entries, which may cause
unnecessary waiting, and potential soft lockup.

This is fixed via avoiding to wait when !pmd_none() && !pmd_present(),
only wait when pmd_none().

This may be not a problem in practice, because userfaultfd_must_wait()
is always called with mm->mmap_sem read-locked.  mremap() will
write-lock mm->mmap_sem.  And UFFDIO_COPY doesn't support to copy THP
mapping.  But the change introduced still makes the code more correct,
and makes the PMD and PTE code more consistent.

Link: http://lkml.kernel.org/r/20171207011752.3292-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@linux.vnet.ibm.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.UK>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/huge_memory.c: fix comment in __split_huge_pmd_locked

pmd_trans_splitting() was removed after THP refcounting redesign,
therefore related comment should be updated.

Link: http://lkml.kernel.org/r/1512625745-59451-1-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: memory_hotplug: remove second __nr_to_section in register_page_bootmem_info_section()

In register_page_bootmem_info_section() we call __nr_to_section() in
order to get the mem_section struct at the beginning of the function.
Since we already got it, there is no need for a second call to
__nr_to_section().

Link: http://lkml.kernel.org/r/20171207102914.GA12396@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@techadventures.net>
Acked-by: Michal Hocko <mhocko@suse.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>

fs/proc/task_mmu.c: do not show VmExe bigger than total executable virtual memory

If start_code / end_code pointers are screwed then "VmExe" could be
bigger than total executable virtual memory and "VmLib" becomes
negative:

  VmExe:   294320 kB
  VmLib: 18446744073709327564 kB

VmExe and VmLib documented as text segment and shared library code size.

Now their sum will be always equal to mm->exec_vm which sums size of
executable and not writable and not stack areas.

I've seen this for huge (>2Gb) statically linked binary which has whole
world inside.  For it start_code ..  end_code range also covers one of
rodata sections.  Probably this is bug in customized linker, elf loader
or both.

Anyway CONFIG_CHECKPOINT_RESTORE allows to change these pointers, thus
we cannot trust them without validation.

Link: http://lkml.kernel.org/r/150728955451.743749.11276392315459539583.stgit@buzz
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
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: update comment describing tlb_gather_mmu

The comment describes @fullmm argument, but the function has no such
parameter.

Update the comment to match the code and convert it to kernel-doc
markup.

Link: http://lkml.kernel.org/r/1512394531-2264-1-git-send-email-rppt@linux.vnet.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug.c: remove unnecesary check from register_page_bootmem_info_section()

When we call register_page_bootmem_info_section() having
CONFIG_SPARSEMEM_VMEMMAP enabled, we check if the pfn is valid.

This check is redundant as we already checked this in
register_page_bootmem_info_node() before calling
register_page_bootmem_info_section(), so let's get rid of it.

Link: http://lkml.kernel.org/r/20171205143422.GA31458@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@techadventures.net>
Acked-by: Michal Hocko <mhocko@suse.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, hugetlb: remove hugepages_treat_as_movable sysctl

hugepages_treat_as_movable has been introduced by 396faf0303d2 ("Allow
huge page allocations to use GFP_HIGH_MOVABLE") to allow hugetlb
allocations from ZONE_MOVABLE even when hugetlb pages were not
migrateable.  The purpose of the movable zone was different at the time.
It aimed at reducing memory fragmentation and hugetlb pages being long
lived and large werre not contributing to the fragmentation so it was
acceptable to use the zone back then.

Things have changed though and the primary purpose of the zone became
migratability guarantee.  If we allow non migrateable hugetlb pages to
be in ZONE_MOVABLE memory hotplug might fail to offline the memory.

Remove the knob and only rely on hugepage_migration_supported to allow
movable zones.

Mel said:

: Primarily it was aimed at allowing the hugetlb pool to safely shrink with
: the ability to grow it again.  The use case was for batched jobs, some of
: which needed huge pages and others that did not but didn't want the memory
: useless pinned in the huge pages pool.
:
: I suspect that more users rely on THP than hugetlbfs for flexible use of
: huge pages with fallback options so I think that removing the option
: should be ok.

Link: http://lkml.kernel.org/r/20171003072619.8654-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Alexandru Moise <00moses.alexander00@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Alexandru Moise <00moses.alexander00@gmail.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove unused pgdat_reclaimable_pages()

Remove unused function pgdat_reclaimable_pages() and
node_page_state_snapshot() which becomes unused as well.

Link: http://lkml.kernel.org/r/20171122094416.26019-1-jack@suse.cz
Signed-off-by: Jan Kara <jack@suse.cz>
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/interval_tree.c: use vma_pages() helper

Use vma_pages function on vma object instead of explicit computation.

  mm/interval_tree.c:21:27-33: WARNING: Consider using vma_pages helper

Generated by: scripts/coccinelle/api/vma_pages.cocci

Link: http://lkml.kernel.org/r/1511364410-13499-1-git-send-email-gomonovych@gmail.com
Signed-off-by: Vasyl Gomonovych <gomonovych@gmail.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Davidlohr Bueso <dbueso@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

selftests/vm: move 128TB mmap boundary test to generic directory

Architectures like PPC64 support mmap hint address based large address
space selection.  This test can be run on those architectures too.  Move
the test from the x86 selftests to selftest/vm so that other
architectures can use it too.

We also add a few new test scenarios in this patch.  We do test a few
boundary conditions before we do a high address mmap.  PPC64 uses the
address limit to validate the address in the fault path.  We had bugs in
this area w.r.t SLB fault handling before we updated the addess limit.

We also touch the allocated space to make sure we don't have any bugs in
the fault handling path.

[akpm@linux-foundation.org: restore tools/testing/selftests/vm/Makefile alpha ordering]
Link: http://lkml.kernel.org/r/20171123165226.32582-1-aneesh.kumar@linux.vnet.ibm.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: do not stall register_shrinker()

Shakeel Butt reported he has observed in production systems that the job
loader gets stuck for 10s of seconds while doing a mount operation.  It
turns out that it was stuck in register_shrinker() because some
unrelated job was under memory pressure and was spending time in
shrink_slab().  Machines have a lot of shrinkers registered and jobs
under memory pressure have to traverse all of those memcg-aware
shrinkers and affect unrelated jobs which want to register their own
shrinkers.

To solve the issue, this patch simply bails out slab shrinking if it is
found that someone wants to register a shrinker in parallel.  A downside
is it could cause unfair shrinking between shrinkers.  However, it
should be rare and we can add compilcated logic if we find it's not
enough.

[akpm@linux-foundation.org: tweak code comment]
Link: http://lkml.kernel.org/r/20171115005602.GB23810@bbox
Link: http://lkml.kernel.org/r/1511481899-20335-1-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Shakeel Butt <shakeelb@google.com>
Reported-by: Shakeel Butt <shakeelb@google.com>
Tested-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
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/page_alloc.c: fix comment in __get_free_pages()

__get_free_pages() will return a virtual address, but it is not just a
32-bit address, for example on a 64-bit system.  And this comment really
confuses new readers of mm.

Link: http://lkml.kernel.org/r/1511780964-64864-1-git-send-email-chenjiankang1@huawei.com
Signed-off-by: Jiankang Chen <chenjiankang1@huawei.com>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yisheng Xie <xieyisheng1@huawei.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/page_owner.c: use PTR_ERR_OR_ZERO()

Fix ptr_ret.cocci warnings:

  mm/page_owner.c:639:1-3: WARNING: PTR_ERR_OR_ZERO can be used

Use PTR_ERR_OR_ZERO rather than if(IS_ERR(...)) + PTR_ERR

Generated by: scripts/coccinelle/api/ptr_ret.cocci

Link: http://lkml.kernel.org/r/1511824101-9597-1-git-send-email-gomonovych@gmail.com
Signed-off-by: Vasyl Gomonovych <gomonovych@gmail.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
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: memcontrol: fix excessive complexity in memory.stat reporting

We've seen memory.stat reads in top-level cgroups take up to fourteen
seconds during a userspace bug that created tens of thousands of ghost
cgroups pinned by lingering page cache.

Even with a more reasonable number of cgroups, aggregating memory.stat
is unnecessarily heavy.  The complexity is this:

nr_cgroups * nr_stat_items * nr_possible_cpus

where the stat items are ~70 at this point.  With 128 cgroups and 128
CPUs - decent, not enormous setups - reading the top-level memory.stat
has to aggregate over a million per-cpu counters.  This doesn't scale.

Instead of spreading the source of truth across all CPUs, use the
per-cpu counters merely to batch updates to shared atomic counters.

This is the same as the per-cpu stocks we use for charging memory to the
shared atomic page_counters, and also the way the global vmstat counters
are implemented.

Vmstat has elaborate spilling thresholds that depend on the number of
CPUs, amount of memory, and memory pressure - carefully balancing the
cost of counter updates with the amount of per-cpu error.  That's
because the vmstat counters are system-wide, but also used for decisions
inside the kernel (e.g.  NR_FREE_PAGES in the allocator).  Neither is
true for the memory controller.

Use the same static batch size we already use for page_counter updates
during charging.  The per-cpu error in the stats will be 128k, which is
an acceptable ratio of cores to memory accounting granularity.

[hannes@cmpxchg.org: fix warning in __this_cpu_xchg() calls]
Link: http://lkml.kernel.org/r/20171201135750.GB8097@cmpxchg.org
Link: http://lkml.kernel.org/r/20171103153336.24044-3-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: memcontrol: implement lruvec stat functions on top of each other

The implementation of the lruvec stat functions and their variants for
accounting through a page, or accounting from a preemptible context, are
mostly identical and needlessly repetitive.

Implement the lruvec_page functions by looking up the page's lruvec and
then using the lruvec function.

Implement the functions for preemptible contexts by disabling preemption
before calling the atomic context functions.

Link: http://lkml.kernel.org/r/20171103153336.24044-2-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: memcontrol: eliminate raw access to stat and event counters

Replace all raw 'this_cpu_' modifications of the stat and event per-cpu
counters with API functions such as mod_memcg_state().

This makes the code easier to read, but is also in preparation for the
next patch, which changes the per-cpu implementation of those counters.

Link: http://lkml.kernel.org/r/20171103153336.24044-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap.c: remove include of hardirq.h

in_atomic() has been moved to include/linux/preempt.h, and the filemap.c
doesn't use in_atomic() directly at all, so it sounds unnecessary to
include hardirq.h.

Link: http://lkml.kernel.org/r/1509985319-38633-1-git-send-email-yang.s@alibaba-inc.com
Signed-off-by: Yang Shi <yang.s@alibaba-inc.com>
Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: split deferred_init_range into initializing and freeing parts

In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator.  We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.

There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.

This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.

In addition, this patch brings the following improvements:
 - Get rid of __def_free() helper function. And simplifies loop logic by
   adding a new pfn validity check function: deferred_pfn_valid().
 - Reduces number of variables that we track. So, there is a higher
   chance that we will avoid using stack to store/load variables inside
   hot loops.
 - Enables future multi-threading of these functions: do initialization
   in multiple threads, wait for all threads to finish, do freeing part
   in multithreading.

Tested on x86 with 1T of memory to make sure no regressions are
introduced.

[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.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: use sc->priority for slab shrink targets

Previously we were using the ratio of the number of lru pages scanned to
the number of eligible lru pages to determine the number of slab objects
to scan.  The problem with this is that these two things have nothing to
do with each other, so in slab heavy work loads where there is little to
no page cache we can end up with the pages scanned being a very low
number.  This means that we reclaim next to no slab pages and waste a
lot of time reclaiming small amounts of space.

Consider the following scenario, where we have the following values and
the rest of the memory usage is in slab

  Active:            58840 kB
  Inactive:          46860 kB

Every time we do a get_scan_count() we do this

  scan = size >> sc->priority

where sc->priority starts at DEF_PRIORITY, which is 12.  The first loop
through reclaim would result in a scan target of 2 pages to 11715 total
inactive pages, and 3 pages to 14710 total active pages.  This is a
really really small target for a system that is entirely slab pages.
And this is super optimistic, this assumes we even get to scan these
pages.  We don't increment sc->nr_scanned unless we 1) isolate the page,
which assumes it's not in use, and 2) can lock the page.  Under pressure
these numbers could probably go down, I'm sure there's some random pages
from daemons that aren't actually in use, so the targets get even
smaller.

Instead use sc->priority in the same way we use it to determine scan
amounts for the lru's.  This generally equates to pages.  Consider the
following

  slab_pages = (nr_objects * object_size) / PAGE_SIZE

What we would like to do is

  scan = slab_pages >> sc->priority

but we don't know the number of slab pages each shrinker controls, only
the objects.  However say that theoretically we knew how many pages a
shrinker controlled, we'd still have to convert this to objects, which
would look like the following

  scan = shrinker_pages >> sc->priority
  scan_objects = (PAGE_SIZE / object_size) * scan

or written another way

  scan_objects = (shrinker_pages >> sc->priority) *
 (PAGE_SIZE / object_size)

which can thus be written

  scan_objects = ((shrinker_pages * PAGE_SIZE) / object_size) >>
 sc->priority

which is just

  scan_objects = nr_objects >> sc->priority

We don't need to know exactly how many pages each shrinker represents,
it's objects are all the information we need.  Making this change allows
us to place an appropriate amount of pressure on the shrinker pools for
their relative size.

Link: http://lkml.kernel.org/r/1510780549-6812-1-git-send-email-josef@toxicpanda.com
Signed-off-by: Josef Bacik <jbacik@fb.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Dave Chinner <david@fromorbit.com>
Acked-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: show total hugetlb memory consumption in /proc/meminfo

Currently we display some hugepage statistics (total, free, etc) in
/proc/meminfo, but only for default hugepage size (e.g.  2Mb).

If hugepages of different sizes are used (like 2Mb and 1Gb on x86-64),
/proc/meminfo output can be confusing, as non-default sized hugepages
are not reflected at all, and there are no signs that they are existing
and consuming system memory.

To solve this problem, let's display the total amount of memory,
consumed by hugetlb pages of all sized (both free and used).  Let's call
it "Hugetlb", and display size in kB to match generic /proc/meminfo
style.

For example, (1024 2Mb pages and 2 1Gb pages are pre-allocated):
  $ cat /proc/meminfo
  MemTotal:        8168984 kB
  MemFree:         3789276 kB
  <...>
  CmaFree:               0 kB
  HugePages_Total:    1024
  HugePages_Free:     1024
  HugePages_Rsvd:        0
  HugePages_Surp:        0
  Hugepagesize:       2048 kB
  Hugetlb:         4194304 kB
  DirectMap4k:       32632 kB
  DirectMap2M:     4161536 kB
  DirectMap1G:     6291456 kB

Also, this patch updates corresponding docs to reflect Hugetlb entry
meaning and difference between Hugetlb and HugePages_Total * Hugepagesize.

Link: http://lkml.kernel.org/r/20171115231409.12131-1-guro@fb.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: drop hotplug lock from lru_add_drain_all()

Pulling cpu hotplug locks inside the mm core function like
lru_add_drain_all just asks for problems and the recent lockdep splat
[1] just proves this.  While the usage in that particular case might be
wrong we should avoid the locking as lru_add_drain_all() is used in many
places.  It seems that this is not all that hard to achieve actually.

We have done the same thing for drain_all_pages which is analogous by
commit a459eeb7b852 ("mm, page_alloc: do not depend on cpu hotplug locks
inside the allocator").  All we have to care about is to handle

      - the work item might be executed on a different cpu in worker from
        unbound pool so it doesn't run on pinned on the cpu

      - we have to make sure that we do not race with page_alloc_cpu_dead
        calling lru_add_drain_cpu

the first part is already handled because the worker calls lru_add_drain
which disables preemption when calling lru_add_drain_cpu on the local
cpu it is draining.  The later is true because page_alloc_cpu_dead is
called on the controlling CPU after the hotplugged CPU vanished
completely.

[1] http://lkml.kernel.org/r/089e0825eec8955c1f055c83d476@google.com

[add a cpu hotplug locking interaction as per tglx]
Link: http://lkml.kernel.org/r/20171116120535.23765-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Tejun Heo <tj@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/mempolicy: add nodes_empty check in SYSC_migrate_pages

As in manpage of migrate_pages, the errno should be set to EINVAL when
none of the node IDs specified by new_nodes are on-line and allowed by
the process's current cpuset context, or none of the specified nodes
contain memory.  However, when test by following case:

new_nodes = 0;
old_nodes = 0xf;
ret = migrate_pages(pid, old_nodes, new_nodes, MAX);

The ret will be 0 and no errno is set.  As the new_nodes is empty, we
should expect EINVAL as documented.

To fix the case like above, this patch check whether target nodes AND
current task_nodes is empty, and then check whether AND
node_states[N_MEMORY] is empty.

Link: http://lkml.kernel.org/r/1510882624-44342-4-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Chris Salls <salls@cs.ucsb.edu>
Cc: Christopher Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Tan Xiaojun <tanxiaojun@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/mempolicy: fix the check of nodemask from user

As Xiaojun reported the ltp of migrate_pages01 will fail on arm64 system
which has 4 nodes[0...3], all have memory and CONFIG_NODES_SHIFT=2:

  migrate_pages01    0  TINFO  :  test_invalid_nodes
  migrate_pages01   14  TFAIL  :  migrate_pages_common.c:45: unexpected failure - returned value = 0, expected: -1
  migrate_pages01   15  TFAIL  :  migrate_pages_common.c:55: call succeeded unexpectedly

In this case the test_invalid_nodes of migrate_pages01 will call:
SYSC_migrate_pages as:

  migrate_pages(0, , {0x0000000000000001}, 64, , {0x0000000000000010}, 64) = 0

The new nodes specifies one or more node IDs that are greater than the
maximum supported node ID, however, the errno is not set to EINVAL as
expected.

As man pages of set_mempolicy[1], mbind[2], and migrate_pages[3]
mentioned, when nodemask specifies one or more node IDs that are greater
than the maximum supported node ID, the errno should set to EINVAL.
However, get_nodes only check whether the part of bits
[BITS_PER_LONG*BITS_TO_LONGS(MAX_NUMNODES), maxnode) is zero or not, and
remain [MAX_NUMNODES, BITS_PER_LONG*BITS_TO_LONGS(MAX_NUMNODES)
unchecked.

This patch is to check the bits of [MAX_NUMNODES, maxnode) in get_nodes
to let migrate_pages set the errno to EINVAL when nodemask specifies one
or more node IDs that are greater than the maximum supported node ID,
which follows the manpage's guide.

[1] http://man7.org/linux/man-pages/man2/set_mempolicy.2.html
[2] http://man7.org/linux/man-pages/man2/mbind.2.html
[3] http://man7.org/linux/man-pages/man2/migrate_pages.2.html

Link: http://lkml.kernel.org/r/1510882624-44342-3-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Reported-by: Tan Xiaojun <tanxiaojun@huawei.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Chris Salls <salls@cs.ucsb.edu>
Cc: Christopher Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/mempolicy: remove redundant check in get_nodes

We have already checked whether maxnode is a page worth of bits, by:
    maxnode > PAGE_SIZE*BITS_PER_BYTE

So no need to check it once more.

Link: http://lkml.kernel.org/r/1510882624-44342-2-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Chris Salls <salls@cs.ucsb.edu>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Tan Xiaojun <tanxiaojun@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: relax deferred struct page requirements

There is no need to have ARCH_SUPPORTS_DEFERRED_STRUCT_PAGE_INIT, as all
the page initialization code is in common code.

Also, there is no need to depend on MEMORY_HOTPLUG, as initialization
code does not really use hotplug memory functionality.  So, we can
remove this requirement as well.

This patch allows to use deferred struct page initialization on all
platforms with memblock allocator.

Tested on x86, arm64, and sparc.  Also, verified that code compiles on
PPC with CONFIG_MEMORY_HOTPLUG disabled.

Link: http://lkml.kernel.org/r/20171117014601.31606-1-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Heiko Carstens <heiko.carstens@de.ibm.com> [s390]
Reviewed-by: Khalid Aziz <khalid.aziz@oracle.com>
Acked-by: Michael Ellerman <mpe@ellerman.id.au>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Reza Arbab <arbab@linux.vnet.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zswap: same-filled pages handling

Zswap is a cache which compresses the pages that are being swapped out
and stores them into a dynamically allocated RAM-based memory pool.
Experiments have shown that around 10-20% of pages stored in zswap are
same-filled pages (i.e.  contents of the page are all same), but these
pages are handled as normal pages by compressing and allocating memory
in the pool.

This patch adds a check in zswap_frontswap_store() to identify
same-filled page before compression of the page.  If the page is a
same-filled page, set zswap_entry.length to zero, save the same-filled
value and skip the compression of the page and alloction of memory in
zpool.  In zswap_frontswap_load(), check if value of zswap_entry.length
is zero corresponding to the page to be loaded.  If zswap_entry.length
is zero, fill the page with same-filled value.  This saves the
decompression time during load.

On a ARM Quad Core 32-bit device with 1.5GB RAM by launching and
relaunching different applications, out of ~64000 pages stored in zswap,
~11000 pages were same-value filled pages (including zero-filled pages)
and ~9000 pages were zero-filled pages.

An average of 17% of pages(including zero-filled pages) in zswap are
same-value filled pages and 14% pages are zero-filled pages.  An average
of 3% of pages are same-filled non-zero pages.

The below table shows the execution time profiling with the patch.

                            Baseline    With patch  % Improvement
  -----------------------------------------------------------------
  *Zswap Store Time           26.5ms       18ms          32%
   (of same value pages)
  *Zswap Load Time
   (of same value pages)      25.5ms       13ms          49%
  -----------------------------------------------------------------

On Ubuntu PC with 2GB RAM, while executing kernel build and other test
scripts and running multimedia applications, out of 360000 pages stored
in zswap 78000(~22%) of pages were found to be same-value filled pages
(including zero-filled pages) and 64000(~17%) are zero-filled pages.  So
an average of %5 of pages are same-filled non-zero pages.

The below table shows the execution time profiling with the patch.

                            Baseline    With patch  % Improvement
  -----------------------------------------------------------------
  *Zswap Store Time           91ms        74ms           19%
   (of same value pages)
  *Zswap Load Time            50ms        7.5ms          85%
   (of same value pages)
  -----------------------------------------------------------------

*The execution times may vary with test device used.

Dan said:

: I did test this patch out this week, and I added some instrumentation to
: check the performance impact, and tested with a small program to try to
: check the best and worst cases.
:
: When doing a lot of swap where all (or almost all) pages are same-value, I
: found this patch does save both time and space, significantly.  The exact
: improvement in time and space depends on which compressor is being used,
: but roughly agrees with the numbers you listed.
:
: In the worst case situation, where all (or almost all) pages have the
: same-value *except* the final long (meaning, zswap will check each long on
: the entire page but then still have to pass the page to the compressor),
: the same-value check is around 10-15% of the total time spent in
: zswap_frontswap_store().  That's a not-insignificant amount of time, but
: it's not huge.  Considering that most systems will probably be swapping
: pages that aren't similar to the worst case (although I don't have any
: data to know that), I'd say the improvement is worth the possible
: worst-case performance impact.

[srividya.dr@samsung.com: add memset_l instead of for loop]
Link: http://lkml.kernel.org/r/20171018104832epcms5p1b2232e2236258de3d03d1344dde9fce0@epcms5p1
Signed-off-by: Srividya Desireddy <srividya.dr@samsung.com>
Acked-by: Dan Streetman <ddstreet@ieee.org>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Dinakar Reddy Pathireddy <dinakar.p@samsung.com>
Cc: SHARAN ALLUR <sharan.allur@samsung.com>
Cc: RAJIB BASU <rajib.basu@samsung.com>
Cc: JUHUN KIM <juhunkim@samsung.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Timofey Titovets <nefelim4ag@gmail.com>
Cc: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: kmemleak: remove unused hardirq.h

Preempt counter APIs have been split out, currently, hardirq.h just
includes irq_enter/exit APIs which are not used by kmemleak at all.

So, remove the unused hardirq.h.

Link: http://lkml.kernel.org/r/1510959741-31109-1-git-send-email-yang.s@alibaba-inc.com
Signed-off-by: Yang Shi <yang.s@alibaba-inc.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/sched/mm.h: uninline mmdrop_async(), etc

mmdrop_async() is only used in fork.c.  Move that and its support
functions into fork.c, uninline it all.

Quite a lot of code gets moved around to avoid forward declarations.

Cc: Ingo Molnar <mingo@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

slub: remove obsolete comments of put_cpu_partial()

Commit d6e0b7fa1186 ("slub: make dead caches discard free slabs
immediately") makes put_cpu_partial() run with preemption disabled and
interrupts disabled when calling unfreeze_partials().

The comment: "put_cpu_partial() is done without interrupts disabled and
without preemption disabled" looks obsolete, so remove it.

Link: http://lkml.kernel.org/r/1516968550-1520-1-git-send-email-miles.chen@mediatek.com
Signed-off-by: Miles Chen <miles.chen@mediatek.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/slub.c: fix wrong address during slab padding restoration

Start address calculated for slab padding restoration was wrong.  Wrong
address would point to some section before padding and could cause
corruption

Link: http://lkml.kernel.org/r/1516604578-4577-1-git-send-email-balasubramani_vivekanandan@mentor.com
Signed-off-by: Balasubramani Vivekanandan <balasubramani_vivekanandan@mentor.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/slab.c: remove redundant assignments for slab_state

slab_state is being set to "UP" in create_kmalloc_caches(), and later on
we set it again in kmem_cache_init_late(), but slab_state does not
change in the meantime.

Remove the redundant assignment from kmem_cache_init_late().

And unless I overlooked anything, the same goes for "slab_state = FULL".
slab_state is set to "FULL" in kmem_cache_init_late(), but it is later
being set again in cpucache_init(), which gets called from
do_initcall_level().  So remove the assignment from cpucache_init() as
well.

Link: http://lkml.kernel.org/r/20171215134452.GA1920@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@techadventures.net>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/slab_common.c: make calculate_alignment() static

calculate_alignment() function is only used inside slab_common.c.  So
make it static and let the compiler do more optimizations.

After this patch there's a small improvement in text and data size.

  $ gcc --version
    gcc (GCC) 7.2.1 20171128

Before:
  text    data     bss     dec      hex filename
  9890457  3828702  1212364 14931523 e3d643 vmlinux

After:
  text    data     bss     dec      hex filename
  9890437  3828670  1212364 14931471 e3d60f vmlinux

Also I fixed a style problem reported by checkpatch.

  WARNING: Missing a blank line after declarations
  #53: FILE: mm/slab_common.c:286:
  + unsigned long ralign = cache_line_size();
  + while (size <= ralign / 2)

Link: http://lkml.kernel.org/r/20171210080132.406-1-bhlee.kernel@gmail.com
Signed-off-by: Byongho Lee <bhlee.kernel@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: return error when we attempt to access a dirty bh in jbd2

We should not reuse the dirty bh in jbd2 directly due to the following
situation:

1. When removing extent rec, we will dirty the bhs of extent rec and
   truncate log at the same time, and hand them over to jbd2.

2. The bhs are submitted to jbd2 area successfully.

3. The write-back thread of device help flush the bhs to disk but
   encounter write error due to abnormal storage link.

4. After a while the storage link become normal. Truncate log flush
   worker triggered by the next space reclaiming found the dirty bh of
   truncate log and clear its 'BH_Write_EIO' and then set it uptodate in
   __ocfs2_journal_access():

   ocfs2_truncate_log_worker
     ocfs2_flush_truncate_log
       __ocfs2_flush_truncate_log
         ocfs2_replay_truncate_records
           ocfs2_journal_access_di
             __ocfs2_journal_access // here we clear io_error and set 'tl_bh' uptodata.

5. Then jbd2 will flush the bh of truncate log to disk, but the bh of
   extent rec is still in error state, and unfortunately nobody will
   take care of it.

6. At last the space of extent rec was not reduced, but truncate log
   flush worker have given it back to globalalloc. That will cause
   duplicate cluster problem which could be identified by fsck.ocfs2.

Sadly we can hardly revert this but set fs read-only in case of ruining
atomicity and consistency of space reclaim.

Link: http://lkml.kernel.org/r/5A6E8092.8090701@huawei.com
Fixes: acf8fdbe6afb ("ocfs2: do not BUG if buffer not uptodate in __ocfs2_journal_access")
Signed-off-by: Jun Piao <piaojun@huawei.com>
Reviewed-by: Yiwen Jiang <jiangyiwen@huawei.com>
Reviewed-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: unlock bh_state if bg check fails

We should unlock bh_stat if bg->bg_free_bits_count > bg->bg_bits

Link: http://lkml.kernel.org/r/1516843095-23680-1-git-send-email-ge.changwei@h3c.com
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Suggested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: nowait aio support

Return EAGAIN if any of the following checks fail for direct I/O:

 - Cannot get the related locks immediately

 - Blocks are not allocated at the write location, it will trigger block
   allocation and block IO operations.

[ghe@suse.com: v4]
Link: http://lkml.kernel.org/r/1516007283-29932-4-git-send-email-ghe@suse.com
[ghe@suse.com: v2]
Link: http://lkml.kernel.org/r/1511944612-9629-4-git-send-email-ghe@suse.com
Link: http://lkml.kernel.org/r/1511775987-841-4-git-send-email-ghe@suse.com
Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Alex Chen <alex.chen@huawei.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Cc: Jun Piao <piaojun@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: add ocfs2_overwrite_io()

Add ocfs2_overwrite_io function, which is used to judge if overwrite
allocated blocks, otherwise, the write will bring extra block allocation
overhead.

[ghe@suse.com: v3]
Link: http://lkml.kernel.org/r/1514455665-16325-3-git-send-email-ghe@suse.com
[ghe@suse.com: v2]
Link: http://lkml.kernel.org/r/1511944612-9629-3-git-send-email-ghe@suse.com
Link: http://lkml.kernel.org/r/1511775987-841-3-git-send-email-ghe@suse.com
Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Jun Piao <piaojun@huawei.com>
Cc: alex chen <alex.chen@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: add ocfs2_try_rw_lock() and ocfs2_try_inode_lock()

Patch series "ocfs2: add nowait aio support", v4.

VFS layer has introduced the non-blocking aio flag IOCB_NOWAIT, which
tells the kernel to bail out if an AIO request will block for reasons
such as file allocations, or writeback triggering, or would block while
allocating requests while performing direct I/O.

Subsequently, pwritev2/preadv2 also can leverage this part of kernel
code.  So far, ext4/xfs/btrfs have supported this feature.  Add the
related code for the ocfs2 file system.

This patch (of 3):

Add ocfs2_try_rw_lock and ocfs2_try_inode_lock functions, which will be
used in non-blocking IO scenarios.

[ghe@suse.com: v2]
Link: http://lkml.kernel.org/r/1511944612-9629-2-git-send-email-ghe@suse.com
Link: http://lkml.kernel.org/r/1511775987-841-2-git-send-email-ghe@suse.com
Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Jun Piao <piaojun@huawei.com>
Acked-by: alex chen <alex.chen@huawei.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: add trimfs lock to avoid duplicated trims in cluster

ocfs2 supports trimming the underlying disk via the fstrim command.  But
there is a problem, ocfs2 is a shared disk cluster file system, if the
user configures a scheduled fstrim job on each file system node, this
will trigger multiple nodes trimming a shared disk simultaneously, which
is very wasteful for CPU and IO consumption.  This also might negatively
affect the lifetime of poor-quality SSD devices.

So we introduce a trimfs dlm lock to communicate with each other in this
case, which will make only one fstrim command to do the trimming on a
shared disk among the cluster.  The fstrim commands from the other nodes
should wait for the first fstrim to finish and return success directly,
to avoid running the same trim on the shared disk again.

Link: http://lkml.kernel.org/r/1513228484-2084-2-git-send-email-ghe@suse.com
Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: add trimfs dlm lock resource

Introduce a new dlm lock resource, which will be used to communicate
during fstrimming of an ocfs2 device from cluster nodes.

Link: http://lkml.kernel.org/r/1513228484-2084-1-git-send-email-ghe@suse.com
Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: try to reuse extent block in dealloc without meta_alloc

A crash issue was reported by John Lightsey with a call trace as follows:

  ocfs2_split_extent+0x1ad3/0x1b40 [ocfs2]
  ocfs2_change_extent_flag+0x33a/0x470 [ocfs2]
  ocfs2_mark_extent_written+0x172/0x220 [ocfs2]
  ocfs2_dio_end_io+0x62d/0x910 [ocfs2]
  dio_complete+0x19a/0x1a0
  do_blockdev_direct_IO+0x19dd/0x1eb0
  __blockdev_direct_IO+0x43/0x50
  ocfs2_direct_IO+0x8f/0xa0 [ocfs2]
  generic_file_direct_write+0xb2/0x170
  __generic_file_write_iter+0xc3/0x1b0
  ocfs2_file_write_iter+0x4bb/0xca0 [ocfs2]
  __vfs_write+0xae/0xf0
  vfs_write+0xb8/0x1b0
  SyS_write+0x4f/0xb0
  system_call_fastpath+0x16/0x75

The BUG code told that extent tree wants to grow but no metadata was
reserved ahead of time.  From my investigation into this issue, the root
cause it that although enough metadata is not reserved, there should be
enough for following use.  Rightmost extent is merged into its left one
due to a certain times of marking extent written.  Because during
marking extent written, we got many physically continuous extents.  At
last, an empty extent showed up and the rightmost path is removed from
extent tree.

Add a new mechanism to reuse extent block cached in dealloc which were
just unlinked from extent tree to solve this crash issue.

Criteria is that during marking extents *written*, if extent rotation
and merging results in unlinking extent with growing extent tree later
without any metadata reserved ahead of time, try to reuse those extents
in dealloc in which deleted extents are cached.

Also, this patch addresses the issue John reported that ::dw_zero_count
is not calculated properly.

After applying this patch, the issue John reported was gone.  Thanks for
the reproducer provided by John.  And this patch has passed
ocfs2-test(29 cases) suite running by New H3C Group.

[ge.changwei@h3c.com: fix static checker warnning]
Link: http://lkml.kernel.org/r/63ADC13FD55D6546B7DECE290D39E373F29196AE@H3CMLB12-EX.srv.huawei-3com.com
[akpm@linux-foundation.org: brelse(NULL) is legal]
Link: http://lkml.kernel.org/r/1515479070-32653-2-git-send-email-ge.changwei@h3c.com
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Reported-by: John Lightsey <john@nixnuts.net>
Tested-by: John Lightsey <john@nixnuts.net>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: make metadata estimation accurate and clear

Current code assume that ::w_unwritten_list always has only one item on.
This is not right and hard to get understood.  So improve how to count
unwritten item.

Link: http://lkml.kernel.org/r/1515479070-32653-1-git-send-email-ge.changwei@h3c.com
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Reported-by: John Lightsey <john@nixnuts.net>
Tested-by: John Lightsey <john@nixnuts.net>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2/acl: use 'ip_xattr_sem' to protect getting extended attribute

The race between *set_acl and *get_acl will cause getting incomplete
xattr data as below:

  processA                                    processB

  ocfs2_set_acl
    ocfs2_xattr_set
      __ocfs2_xattr_set_handle

                                              ocfs2_get_acl_nolock
                                                ocfs2_xattr_get_nolock:

processB may get incomplete xattr data if processA hasn't set_acl done.

So we should use 'ip_xattr_sem' to protect getting extended attribute in
ocfs2_get_acl_nolock(), as other processes could be changing it
concurrently.

Link: http://lkml.kernel.org/r/5A5DDCFF.7030001@huawei.com
Signed-off-by: Jun Piao <piaojun@huawei.com>
Reviewed-by: Alex Chen <alex.chen@huawei.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: clean up dead code in alloc.c

Some stack variables are no longer used but still assigned.  Trim them.

Link: http://lkml.kernel.org/r/1516105069-12643-1-git-send-email-ge.changwei@h3c.com
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Reviewed-by: Jun Piao <piaojun@huawei.com>
Reviewed-by: Alex Chen <alex.chen@huawei.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2/xattr: assign errno to 'ret' in ocfs2_calc_xattr_init()

We need catch the errno returned by ocfs2_xattr_get_nolock() and assign
it to 'ret' for printing and noticing upper callers.

Link: http://lkml.kernel.org/r/5A571CAF.8050709@huawei.com
Signed-off-by: Jun Piao <piaojun@huawei.com>
Reviewed-by: Alex Chen <alex.chen@huawei.com>
Reviewed-by: Yiwen Jiang <jiangyiwen@huawei.com>
Acked-by: Gang He <ghe@suse.com>
Acked-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: try a blocking lock before return AOP_TRUNCATED_PAGE

If we can't get inode lock immediately in the function
ocfs2_inode_lock_with_page() when reading a page, we should not return
directly here, since this will lead to a softlockup problem when the
kernel is configured with CONFIG_PREEMPT is not set.  The method is to
get a blocking lock and immediately unlock before returning, this can
avoid CPU resource waste due to lots of retries, and benefits fairness
in getting lock among multiple nodes, increase efficiency in case
modifying the same file frequently from multiple nodes.

The softlockup crash (when set /proc/sys/kernel/softlockup_panic to 1)
looks like:

  Kernel panic - not syncing: softlockup: hung tasks
  CPU: 0 PID: 885 Comm: multi_mmap Tainted: G L 4.12.14-6.1-default #1
  Hardware name: Bochs Bochs, BIOS Bochs 01/01/2011
  Call Trace:
    <IRQ>
    dump_stack+0x5c/0x82
    panic+0xd5/0x21e
    watchdog_timer_fn+0x208/0x210
    __hrtimer_run_queues+0xcc/0x200
    hrtimer_interrupt+0xa6/0x1f0
    smp_apic_timer_interrupt+0x34/0x50
    apic_timer_interrupt+0x96/0xa0
    </IRQ>
   RIP: 0010:unlock_page+0x17/0x30
   RSP: 0000:ffffaf154080bc88 EFLAGS: 00000246 ORIG_RAX: ffffffffffffff10
   RAX: dead000000000100 RBX: fffff21e009f5300 RCX: 0000000000000004
   RDX: dead0000000000ff RSI: 0000000000000202 RDI: fffff21e009f5300
   RBP: 0000000000000000 R08: 0000000000000000 R09: ffffaf154080bb00
   R10: ffffaf154080bc30 R11: 0000000000000040 R12: ffff993749a39518
   R13: 0000000000000000 R14: fffff21e009f5300 R15: fffff21e009f5300
    ocfs2_inode_lock_with_page+0x25/0x30 [ocfs2]
    ocfs2_readpage+0x41/0x2d0 [ocfs2]
    filemap_fault+0x12b/0x5c0
    ocfs2_fault+0x29/0xb0 [ocfs2]
    __do_fault+0x1a/0xa0
    __handle_mm_fault+0xbe8/0x1090
    handle_mm_fault+0xaa/0x1f0
    __do_page_fault+0x235/0x4b0
    trace_do_page_fault+0x3c/0x110
    async_page_fault+0x28/0x30
   RIP: 0033:0x7fa75ded638e
   RSP: 002b:00007ffd6657db18 EFLAGS: 00010287
   RAX: 000055c7662fb700 RBX: 0000000000000001 RCX: 000055c7662fb700
   RDX: 0000000000001770 RSI: 00007fa75e909000 RDI: 000055c7662fb700
   RBP: 0000000000000003 R08: 000000000000000e R09: 0000000000000000
   R10: 0000000000000483 R11: 00007fa75ded61b0 R12: 00007fa75e90a770
   R13: 000000000000000e R14: 0000000000001770 R15: 0000000000000000

About performance improvement, we can see the testing time is reduced,
and CPU utilization decreases, the detailed data is as follows.  I ran
multi_mmap test case in ocfs2-test package in a three nodes cluster.

Before applying this patch:
    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
   2754 ocfs2te+  20   0  170248   6980   4856 D 80.73 0.341   0:18.71 multi_mmap
   1505 root      rt   0  222236 123060  97224 S 2.658 6.015   0:01.44 corosync
      5 root      20   0       0      0      0 S 1.329 0.000   0:00.19 kworker/u8:0
     95 root      20   0       0      0      0 S 1.329 0.000   0:00.25 kworker/u8:1
   2728 root      20   0       0      0      0 S 0.997 0.000   0:00.24 jbd2/sda1-33
   2721 root      20   0       0      0      0 S 0.664 0.000   0:00.07 ocfs2dc-3C8CFD4
   2750 ocfs2te+  20   0  142976   4652   3532 S 0.664 0.227   0:00.28 mpirun

  ocfs2test@tb-node2:~>multiple_run.sh -i ens3 -k ~/linux-4.4.21-69.tar.gz -o ~/ocfs2mullog -C hacluster -s pcmk -n tb-node2,tb-node1,tb-node3 -d /dev/sda1 -b 4096 -c 32768 -t multi_mmap /mnt/shared
  Tests with "-b 4096 -C 32768"
  Thu Dec 28 14:44:52 CST 2017
  multi_mmap..................................................Passed.
  Runtime 783 seconds.

After apply this patch:

    PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
   2508 ocfs2te+  20   0  170248   6804   4680 R 54.00 0.333   0:55.37 multi_mmap
    155 root      20   0       0      0      0 S 2.667 0.000   0:01.20 kworker/u8:3
     95 root      20   0       0      0      0 S 2.000 0.000   0:01.58 kworker/u8:1
   2504 ocfs2te+  20   0  142976   4604   3480 R 1.667 0.225   0:01.65 mpirun
      5 root      20   0       0      0      0 S 1.000 0.000   0:01.36 kworker/u8:0
   2482 root      20   0       0      0      0 S 1.000 0.000   0:00.86 jbd2/sda1-33
    299 root       0 -20       0      0      0 S 0.333 0.000   0:00.13 kworker/2:1H
    335 root       0 -20       0      0      0 S 0.333 0.000   0:00.17 kworker/1:1H
    535 root      20   0   12140   7268   1456 S 0.333 0.355   0:00.34 haveged
   1282 root      rt   0  222284 123108  97224 S 0.333 6.017   0:01.33 corosync

  ocfs2test@tb-node2:~>multiple_run.sh -i ens3 -k ~/linux-4.4.21-69.tar.gz -o ~/ocfs2mullog -C hacluster -s pcmk -n tb-node2,tb-node1,tb-node3 -d /dev/sda1 -b 4096 -c 32768 -t multi_mmap /mnt/shared
  Tests with "-b 4096 -C 32768"
  Thu Dec 28 15:04:12 CST 2017
  multi_mmap..................................................Passed.
  Runtime 487 seconds.

Link: http://lkml.kernel.org/r/1514447305-30814-1-git-send-email-ghe@suse.com
Fixes: 1cce4df04f37 ("ocfs2: do not lock/unlock() inode DLM lock")
Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Eric Ren <zren@suse.com>
Acked-by: alex chen <alex.chen@huawei.com>
Acked-by: piaojun <piaojun@huawei.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: return -EROFS to mount.ocfs2 if inode block is invalid

If metadata is corrupted such as 'invalid inode block', we will get
failed by calling 'mount()' and then set filesystem readonly as below:

  ocfs2_mount
    ocfs2_initialize_super
      ocfs2_init_global_system_inodes
        ocfs2_iget
          ocfs2_read_locked_inode
            ocfs2_validate_inode_block
      ocfs2_error
        ocfs2_handle_error
          ocfs2_set_ro_flag(osb, 0);  // set readonly

In this situation we need return -EROFS to 'mount.ocfs2', so that user
can fix it by fsck.  And then mount again.  In addition, 'mount.ocfs2'
should be updated correspondingly as it only return 1 for all errno.
And I will post a patch for 'mount.ocfs2' too.

Link: http://lkml.kernel.org/r/5A4302FA.2010606@huawei.com
Signed-off-by: Jun Piao <piaojun@huawei.com>
Reviewed-by: Alex Chen <alex.chen@huawei.com>
Reviewed-by: Joseph Qi <jiangqi903@gmail.com>
Reviewed-by: Changwei Ge <ge.changwei@h3c.com>
Reviewed-by: Gang He <ghe@suse.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: clean dead code in suballoc.c

Stack variable fe is no longer used, so trim it to save some CPU cycles
and stack space.

Link: http://lkml.kernel.org/r/63ADC13FD55D6546B7DECE290D39E373F1F5A8DD@H3CMLB14-EX.srv.huawei-3com.com
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Reviewed-by: Joseph Qi <jiangqi903@gmail.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: use the OCFS2_XATTR_ROOT_SIZE macro in ocfs2_reflink_xattr_header()

Use the OCFS2_XATTR_ROOT_SIZE macro improves the readability of the
code.

Link: http://lkml.kernel.org/r/5A2E2488.70301@huawei.com
Signed-off-by: Alex Chen <alex.chen@huawei.com>
Reviewed-by: Jun Piao <piaojun@huawei.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <ge.changwei@h3c.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2/cluster: close a race that fence can't be triggered

When some nodes of cluster face with TCP connection fault, ocfs2 will
pick up a quorum to continue to work and other nodes will be fenced by
resetting host.

In order to decide which node should be fenced, ocfs2 leverages
o2quo_state::qs_holds.  If that variable is reduced to zero, then a try
to decide if fence local node is performed.  However, under a specific
scenario that local node is not disconnected from others at the same
time, above method has a problem to reduce ::qs_holds to zero.

Because, o2net 90s idle timer corresponding to different nodes is
triggered one after another.

  node 2 node 3
  90s idle timer elapses
  clear ::qs_conn_bm
  set hold
40s is passed
90 idle timer elapses
clear ::qs_conn_bm
set hold
  still up timer elapses
  clear hold (NOT to zero )
  90s idle timer elapses AGAIN
still up timer elapses.
clear hold
still up timer elapses

To solve this issue, a node which has already be evicted from
::qs_conn_bm can't set hold again and again invoked from idle timer.

Link: http://lkml.kernel.org/r/63ADC13FD55D6546B7DECE290D39E373F1F3F93B@H3CMLB12-EX.srv.huawei-3com.com
Signed-off-by: Yang Zhang <zhang.yangB@h3c.com>
Signed-off-by: Changwei Ge <ge.changwei@h3c.com>
Cc: Mark Fasheh <mfasheh@versity.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: give an obvious tip for mismatched cluster names

Add an obvious error message, due to mismatched cluster names between
on-disk and in the current cluster.  We can meet this case during OCFS2
cluster migration.

If we can give the user an obvious tip for why they can not mount the
file system after migration, they can quickly fix this mismatch problem.

Second, also move printing ocfs2_fill_super() errno to the front of
ocfs2_dismount_volume(), since ocfs2_dismount_volume() will also print
its own message.

I looked through all the code of OCFS2 (include o2cb); there is not any
place which returns this error.  In fact, the function calling path
ocfs2_fill_super -> ocfs2_mount_volume -> ocfs2_dlm_init ->
dlm_new_lockspace is a very specific one.  We can use this errno to give
the user a more clear tip, since this case is a little common during
cluster migration, but the customer can quickly get the failure cause if
there is a error printed.  Also, I think it is not possible to add this
errno in the o2cb path during ocfs2_dlm_init(), since the o2cb code has
been stable for a long time.

We only print this error tip when the user uses pcmk stack, since using
the o2cb stack the user will not meet this error.

[ghe@suse.com: v2]
Link: http://lkml.kernel.org/r/1495419305-3780-1-git-send-email-ghe@suse.com
Link: http://lkml.kernel.org/r/1495089336-19312-1-git-send-email-ghe@suse.com
Signed-off-by: Gang He <ghe@suse.com>
Reviewed-by: Mark Fasheh <mfasheh@versity.com>
Acked-by: Joseph Qi <jiangqi903@gmail.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>