linux-toradex.git/Documentation/vm, branch v4.6-rc6

mm: thp: set THP defrag by default to madvise and add a stall-free defrag option

2016-03-17T22:09:34+00:00

THP defrag is enabled by default to direct reclaim/compact but not wake
kswapd in the event of a THP allocation failure.  The problem is that
THP allocation requests potentially enter reclaim/compaction.  This
potentially incurs a severe stall that is not guaranteed to be offset by
reduced TLB misses.  While there has been considerable effort to reduce
the impact of reclaim/compaction, it is still a high cost and workloads
that should fit in memory fail to do so.  Specifically, a simple
anon/file streaming workload will enter direct reclaim on NUMA at least
even though the working set size is 80% of RAM.  It's been years and
it's time to throw in the towel.

First, this patch defines THP defrag as follows;

 madvise: A failed allocation will direct reclaim/compact if the application requests it
 never:   Neither reclaim/compact nor wake kswapd
 defer:   A failed allocation will wake kswapd/kcompactd
 always:  A failed allocation will direct reclaim/compact (historical behaviour)
          khugepaged defrag will enter direct/reclaim but not wake kswapd.

Next it sets the default defrag option to be "madvise" to only enter
direct reclaim/compaction for applications that specifically requested
it.

Lastly, it removes a check from the page allocator slowpath that is
related to __GFP_THISNODE to allow "defer" to work.  The callers that
really cares are slub/slab and they are updated accordingly.  The slab
one may be surprising because it also corrects a comment as kswapd was
never woken up by that path.

This means that a THP fault will no longer stall for most applications
by default and the ideal for most users that get THP if they are
immediately available.  There are still options for users that prefer a
stall at startup of a new application by either restoring historical
behaviour with "always" or pick a half-way point with "defer" where
kswapd does some of the work in the background and wakes kcompactd if
necessary.  THP defrag for khugepaged remains enabled and will enter
direct/reclaim but no wakeup kswapd or kcompactd.

After this patch a THP allocation failure will quickly fallback and rely
on khugepaged to recover the situation at some time in the future.  In
some cases, this will reduce THP usage but the benefit of THP is hard to
measure and not a universal win where as a stall to reclaim/compaction
is definitely measurable and can be painful.

The first test for this is using "usemem" to read a large file and write
a large anonymous mapping (to avoid the zero page) multiple times.  The
total size of the mappings is 80% of RAM and the benchmark simply
measures how long it takes to complete.  It uses multiple threads to see
if that is a factor.  On UMA, the performance is almost identical so is
not reported but on NUMA, we see this

usemem
                                   4.4.0                 4.4.0
                          kcompactd-v1r1         nodefrag-v1r3
Amean    System-1       102.86 (  0.00%)       46.81 ( 54.50%)
Amean    System-4        37.85 (  0.00%)       34.02 ( 10.12%)
Amean    System-7        48.12 (  0.00%)       46.89 (  2.56%)
Amean    System-12       51.98 (  0.00%)       56.96 ( -9.57%)
Amean    System-21       80.16 (  0.00%)       79.05 (  1.39%)
Amean    System-30      110.71 (  0.00%)      107.17 (  3.20%)
Amean    System-48      127.98 (  0.00%)      124.83 (  2.46%)
Amean    Elapsd-1       185.84 (  0.00%)      105.51 ( 43.23%)
Amean    Elapsd-4        26.19 (  0.00%)       25.58 (  2.33%)
Amean    Elapsd-7        21.65 (  0.00%)       21.62 (  0.16%)
Amean    Elapsd-12       18.58 (  0.00%)       17.94 (  3.43%)
Amean    Elapsd-21       17.53 (  0.00%)       16.60 (  5.33%)
Amean    Elapsd-30       17.45 (  0.00%)       17.13 (  1.84%)
Amean    Elapsd-48       15.40 (  0.00%)       15.27 (  0.82%)

For a single thread, the benchmark completes 43.23% faster with this
patch applied with smaller benefits as the thread increases.  Similar,
notice the large reduction in most cases in system CPU usage.  The
overall CPU time is

               4.4.0       4.4.0
        kcompactd-v1r1 nodefrag-v1r3
User        10357.65    10438.33
System       3988.88     3543.94
Elapsed      2203.01     1634.41

Which is substantial. Now, the reclaim figures

                                 4.4.0       4.4.0
                          kcompactd-v1r1nodefrag-v1r3
Minor Faults                 128458477   278352931
Major Faults                   2174976         225
Swap Ins                      16904701           0
Swap Outs                     17359627           0
Allocation stalls                43611           0
DMA allocs                           0           0
DMA32 allocs                  19832646    19448017
Normal allocs                614488453   580941839
Movable allocs                       0           0
Direct pages scanned          24163800           0
Kswapd pages scanned                 0           0
Kswapd pages reclaimed               0           0
Direct pages reclaimed        20691346           0
Compaction stalls                42263           0
Compaction success                 938           0
Compaction failures              41325           0

This patch eliminates almost all swapping and direct reclaim activity.
There is still overhead but it's from NUMA balancing which does not
identify that it's pointless trying to do anything with this workload.

I also tried the thpscale benchmark which forces a corner case where
compaction can be used heavily and measures the latency of whether base
or huge pages were used

thpscale Fault Latencies
                                       4.4.0                 4.4.0
                              kcompactd-v1r1         nodefrag-v1r3
Amean    fault-base-1      5288.84 (  0.00%)     2817.12 ( 46.73%)
Amean    fault-base-3      6365.53 (  0.00%)     3499.11 ( 45.03%)
Amean    fault-base-5      6526.19 (  0.00%)     4363.06 ( 33.15%)
Amean    fault-base-7      7142.25 (  0.00%)     4858.08 ( 31.98%)
Amean    fault-base-12    13827.64 (  0.00%)    10292.11 ( 25.57%)
Amean    fault-base-18    18235.07 (  0.00%)    13788.84 ( 24.38%)
Amean    fault-base-24    21597.80 (  0.00%)    24388.03 (-12.92%)
Amean    fault-base-30    26754.15 (  0.00%)    19700.55 ( 26.36%)
Amean    fault-base-32    26784.94 (  0.00%)    19513.57 ( 27.15%)
Amean    fault-huge-1      4223.96 (  0.00%)     2178.57 ( 48.42%)
Amean    fault-huge-3      2194.77 (  0.00%)     2149.74 (  2.05%)
Amean    fault-huge-5      2569.60 (  0.00%)     2346.95 (  8.66%)
Amean    fault-huge-7      3612.69 (  0.00%)     2997.70 ( 17.02%)
Amean    fault-huge-12     3301.75 (  0.00%)     6727.02 (-103.74%)
Amean    fault-huge-18     6696.47 (  0.00%)     6685.72 (  0.16%)
Amean    fault-huge-24     8000.72 (  0.00%)     9311.43 (-16.38%)
Amean    fault-huge-30    13305.55 (  0.00%)     9750.45 ( 26.72%)
Amean    fault-huge-32     9981.71 (  0.00%)    10316.06 ( -3.35%)

The average time to fault pages is substantially reduced in the majority
of caseds but with the obvious caveat that fewer THPs are actually used
in this adverse workload

                                   4.4.0                 4.4.0
                          kcompactd-v1r1         nodefrag-v1r3
Percentage huge-1         0.71 (  0.00%)       14.04 (1865.22%)
Percentage huge-3        10.77 (  0.00%)       33.05 (206.85%)
Percentage huge-5        60.39 (  0.00%)       38.51 (-36.23%)
Percentage huge-7        45.97 (  0.00%)       34.57 (-24.79%)
Percentage huge-12       68.12 (  0.00%)       40.07 (-41.17%)
Percentage huge-18       64.93 (  0.00%)       47.82 (-26.35%)
Percentage huge-24       62.69 (  0.00%)       44.23 (-29.44%)
Percentage huge-30       43.49 (  0.00%)       55.38 ( 27.34%)
Percentage huge-32       50.72 (  0.00%)       51.90 (  2.35%)

                                 4.4.0       4.4.0
                          kcompactd-v1r1nodefrag-v1r3
Minor Faults                  37429143    47564000
Major Faults                      1916        1558
Swap Ins                          1466        1079
Swap Outs                      2936863      149626
Allocation stalls                62510           3
DMA allocs                           0           0
DMA32 allocs                   6566458     6401314
Normal allocs                216361697   216538171
Movable allocs                       0           0
Direct pages scanned          25977580       17998
Kswapd pages scanned                 0     3638931
Kswapd pages reclaimed               0      207236
Direct pages reclaimed         8833714          88
Compaction stalls               103349           5
Compaction success                 270           4
Compaction failures             103079           1

Note again that while this does swap as it's an aggressive workload, the
direct relcim activity and allocation stalls is substantially reduced.
There is some kswapd activity but ftrace showed that the kswapd activity
was due to normal wakeups from 4K pages being allocated.
Compaction-related stalls and activity are almost eliminated.

I also tried the stutter benchmark.  For this, I do not have figures for
NUMA but it's something that does impact UMA so I'll report what is
available

stutter
                                 4.4.0                 4.4.0
                        kcompactd-v1r1         nodefrag-v1r3
Min         mmap      7.3571 (  0.00%)      7.3438 (  0.18%)
1st-qrtle   mmap      7.5278 (  0.00%)     17.9200 (-138.05%)
2nd-qrtle   mmap      7.6818 (  0.00%)     21.6055 (-181.25%)
3rd-qrtle   mmap     11.0889 (  0.00%)     21.8881 (-97.39%)
Max-90%     mmap     27.8978 (  0.00%)     22.1632 ( 20.56%)
Max-93%     mmap     28.3202 (  0.00%)     22.3044 ( 21.24%)
Max-95%     mmap     28.5600 (  0.00%)     22.4580 ( 21.37%)
Max-99%     mmap     29.6032 (  0.00%)     25.5216 ( 13.79%)
Max         mmap   4109.7289 (  0.00%)   4813.9832 (-17.14%)
Mean        mmap     12.4474 (  0.00%)     19.3027 (-55.07%)

This benchmark is trying to fault an anonymous mapping while there is a
heavy IO load -- a scenario that desktop users used to complain about
frequently.  This shows a mix because the ideal case of mapping with THP
is not hit as often.  However, note that 99% of the mappings complete
13.79% faster.  The CPU usage here is particularly interesting

               4.4.0       4.4.0
        kcompactd-v1r1nodefrag-v1r3
User           67.50        0.99
System       1327.88       91.30
Elapsed      2079.00     2128.98

And once again we look at the reclaim figures

                                 4.4.0       4.4.0
                          kcompactd-v1r1nodefrag-v1r3
Minor Faults                 335241922  1314582827
Major Faults                       715         819
Swap Ins                             0           0
Swap Outs                            0           0
Allocation stalls               532723           0
DMA allocs                           0           0
DMA32 allocs                1822364341  1177950222
Normal allocs               1815640808  1517844854
Movable allocs                       0           0
Direct pages scanned          21892772           0
Kswapd pages scanned          20015890    41879484
Kswapd pages reclaimed        19961986    41822072
Direct pages reclaimed        21892741           0
Compaction stalls              1065755           0
Compaction success                 514           0
Compaction failures            1065241           0

Allocation stalls and all direct reclaim activity is eliminated as well
as compaction-related stalls.

THP gives impressive gains in some cases but only if they are quickly
available.  We're not going to reach the point where they are completely
free so lets take the costs out of the fast paths finally and defer the
cost to kswapd, kcompactd and khugepaged where it belongs.

Signed-off-by: Mel Gorman 
Acked-by: Rik van Riel 
Acked-by: Johannes Weiner 
Acked-by: Vlastimil Babka 
Cc: Andrea Arcangeli 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

thp, vmstats: count deferred split events

2016-03-17T22:09:34+00:00

Count how many times we put a THP in split queue.  Currently, it happens
on partial unmap of a THP.

Rapidly growing value can indicate that an application behaves
unfriendly wrt THP: often fault in huge page and then unmap part of it.
This leads to unnecessary memory fragmentation and the application may
require tuning.

The event also can help with debugging kernel [mis-]behaviour.

Signed-off-by: Kirill A. Shutemov 
Cc: Andrea Arcangeli 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm, page_owner: convert page_owner_inited to static key

2016-03-15T23:55:16+00:00

CONFIG_PAGE_OWNER attempts to impose negligible runtime overhead when
enabled during compilation, but not actually enabled during runtime by
boot param page_owner=on.  This overhead can be further reduced using
the static key mechanism, which this patch does.

Signed-off-by: Vlastimil Babka 
Acked-by: Michal Hocko 
Cc: Joonsoo Kim 
Cc: Minchan Kim 
Cc: Sasha Levin 
Cc: "Kirill A. Shutemov" 
Cc: Mel Gorman 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

slub: convert SLAB_DEBUG_FREE to SLAB_CONSISTENCY_CHECKS

2016-03-15T23:55:16+00:00

SLAB_DEBUG_FREE allows expensive consistency checks at free to be turned
on or off.  Expand its use to be able to turn off all consistency
checks.  This gives a nice speed up if you only want features such as
poisoning or tracing.

Credit to Mathias Krause for the original work which inspired this
series

Signed-off-by: Laura Abbott 
Acked-by: Christoph Lameter 
Cc: Pekka Enberg 
Cc: David Rientjes 
Cc: Joonsoo Kim 
Cc: Kees Cook 
Cc: Mathias Krause 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

Merge branch 'akpm' (patches from Andrew)

2016-01-17T20:58:52+00:00

Merge second patch-bomb from Andrew Morton:

 - more MM stuff:

    - Kirill's page-flags rework

    - Kirill's now-allegedly-fixed THP rework

    - MADV_FREE implementation

    - DAX feature work (msync/fsync).  This isn't quite complete but DAX
      is new and it's good enough and the guys have a handle on what
      needs to be done - I expect this to be wrapped in the next week or
      two.

  - some vsprintf maintenance work

  - various other misc bits

* emailed patches from Andrew Morton : (145 commits)
  printk: change recursion_bug type to bool
  lib/vsprintf: factor out %pN[F] handler as netdev_bits()
  lib/vsprintf: refactor duplicate code to special_hex_number()
  printk-formats.txt: remove unimplemented %pT
  printk: help pr_debug and pr_devel to optimize out arguments
  lib/test_printf.c: test dentry printing
  lib/test_printf.c: add test for large bitmaps
  lib/test_printf.c: account for kvasprintf tests
  lib/test_printf.c: add a few number() tests
  lib/test_printf.c: test precision quirks
  lib/test_printf.c: check for out-of-bound writes
  lib/test_printf.c: don't BUG
  lib/kasprintf.c: add sanity check to kvasprintf
  lib/vsprintf.c: warn about too large precisions and field widths
  lib/vsprintf.c: help gcc make number() smaller
  lib/vsprintf.c: expand field_width to 24 bits
  lib/vsprintf.c: eliminate potential race in string()
  lib/vsprintf.c: move string() below widen_string()
  lib/vsprintf.c: pull out padding code from dentry_name()
  printk: do cond_resched() between lines while outputting to consoles
  ...

thp: update documentation

2016-01-16T01:56:32+00:00

The patch updates Documentation/vm/transhuge.txt to reflect changes in
THP design.

Signed-off-by: Kirill A. Shutemov 
Acked-by: Jerome Marchand 
Cc: Sasha Levin 
Cc: Aneesh Kumar K.V 
Cc: Andrea Arcangeli 
Cc: Hugh Dickins 
Cc: Dave Hansen 
Cc: Mel Gorman 
Cc: Rik van Riel 
Cc: Naoya Horiguchi 
Cc: Steve Capper 
Cc: Johannes Weiner 
Cc: Michal Hocko 
Cc: Christoph Lameter 
Cc: David Rientjes 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

Doc: treewide: Fix grammar "a" to "an"

2015-12-10T18:36:40+00:00

This patch fix some grammar mistake.

Signed-off-by: Masanari Iida 
Signed-off-by: Jonathan Corbet

mm: make compound_head() robust

2015-11-07T01:50:42+00:00

Hugh has pointed that compound_head() call can be unsafe in some
context. There's one example:

	CPU0					CPU1

isolate_migratepages_block()
  page_count()
    compound_head()
      !!PageTail() == true
					put_page()
					  tail->first_page = NULL
      head = tail->first_page
					alloc_pages(__GFP_COMP)
					   prep_compound_page()
					     tail->first_page = head
					     __SetPageTail(p);
      !!PageTail() == true
    

The race is pure theoretical. I don't it's possible to trigger it in
practice. But who knows.

We can fix the race by changing how encode PageTail() and compound_head()
within struct page to be able to update them in one shot.

The patch introduces page->compound_head into third double word block in
front of compound_dtor and compound_order. Bit 0 encodes PageTail() and
the rest bits are pointer to head page if bit zero is set.

The patch moves page->pmd_huge_pte out of word, just in case if an
architecture defines pgtable_t into something what can have the bit 0
set.

hugetlb_cgroup uses page->lru.next in the second tail page to store
pointer struct hugetlb_cgroup. The patch switch it to use page->private
in the second tail page instead. The space is free since ->first_page is
removed from the union.

The patch also opens possibility to remove HUGETLB_CGROUP_MIN_ORDER
limitation, since there's now space in first tail page to store struct
hugetlb_cgroup pointer. But that's out of scope of the patch.

That means page->compound_head shares storage space with:

 - page->lru.next;
 - page->next;
 - page->rcu_head.next;

That's too long list to be absolutely sure, but looks like nobody uses
bit 0 of the word.

page->rcu_head.next guaranteed[1] to have bit 0 clean as long as we use
call_rcu(), call_rcu_bh(), call_rcu_sched(), or call_srcu(). But future
call_rcu_lazy() is not allowed as it makes use of the bit and we can
get false positive PageTail().

[1] http://lkml.kernel.org/g/20150827163634.GD4029@linux.vnet.ibm.com

Signed-off-by: Kirill A. Shutemov 
Acked-by: Michal Hocko 
Reviewed-by: Andrea Arcangeli 
Cc: Hugh Dickins 
Cc: David Rientjes 
Cc: Vlastimil Babka 
Acked-by: Paul E. McKenney 
Cc: Aneesh Kumar K.V 
Cc: Andi Kleen 
Cc: Christoph Lameter 
Cc: Joonsoo Kim 
Cc: Sergey Senozhatsky 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm, page_alloc: distinguish between being unable to sleep, unwilling to sleep and avoiding waking kswapd

2015-11-07T01:50:42+00:00

__GFP_WAIT has been used to identify atomic context in callers that hold
spinlocks or are in interrupts.  They are expected to be high priority and
have access one of two watermarks lower than "min" which can be referred
to as the "atomic reserve".  __GFP_HIGH users get access to the first
lower watermark and can be called the "high priority reserve".

Over time, callers had a requirement to not block when fallback options
were available.  Some have abused __GFP_WAIT leading to a situation where
an optimisitic allocation with a fallback option can access atomic
reserves.

This patch uses __GFP_ATOMIC to identify callers that are truely atomic,
cannot sleep and have no alternative.  High priority users continue to use
__GFP_HIGH.  __GFP_DIRECT_RECLAIM identifies callers that can sleep and
are willing to enter direct reclaim.  __GFP_KSWAPD_RECLAIM to identify
callers that want to wake kswapd for background reclaim.  __GFP_WAIT is
redefined as a caller that is willing to enter direct reclaim and wake
kswapd for background reclaim.

This patch then converts a number of sites

o __GFP_ATOMIC is used by callers that are high priority and have memory
  pools for those requests. GFP_ATOMIC uses this flag.

o Callers that have a limited mempool to guarantee forward progress clear
  __GFP_DIRECT_RECLAIM but keep __GFP_KSWAPD_RECLAIM. bio allocations fall
  into this category where kswapd will still be woken but atomic reserves
  are not used as there is a one-entry mempool to guarantee progress.

o Callers that are checking if they are non-blocking should use the
  helper gfpflags_allow_blocking() where possible. This is because
  checking for __GFP_WAIT as was done historically now can trigger false
  positives. Some exceptions like dm-crypt.c exist where the code intent
  is clearer if __GFP_DIRECT_RECLAIM is used instead of the helper due to
  flag manipulations.

o Callers that built their own GFP flags instead of starting with GFP_KERNEL
  and friends now also need to specify __GFP_KSWAPD_RECLAIM.

The first key hazard to watch out for is callers that removed __GFP_WAIT
and was depending on access to atomic reserves for inconspicuous reasons.
In some cases it may be appropriate for them to use __GFP_HIGH.

The second key hazard is callers that assembled their own combination of
GFP flags instead of starting with something like GFP_KERNEL.  They may
now wish to specify __GFP_KSWAPD_RECLAIM.  It's almost certainly harmless
if it's missed in most cases as other activity will wake kswapd.

Signed-off-by: Mel Gorman 
Acked-by: Vlastimil Babka 
Acked-by: Michal Hocko 
Acked-by: Johannes Weiner 
Cc: Christoph Lameter 
Cc: David Rientjes 
Cc: Vitaly Wool 
Cc: Rik van Riel 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

mm: page migration avoid touching newpage until no going back

2015-11-06T03:34:48+00:00

We have had trouble in the past from the way in which page migration's
newpage is initialized in dribs and drabs - see commit 8bdd63809160 ("mm:
fix direct reclaim writeback regression") which proposed a cleanup.

We have no actual problem now, but I think the procedure would be clearer
(and alternative get_new_page pools safer to implement) if we assert that
newpage is not touched until we are sure that it's going to be used -
except for taking the trylock on it in __unmap_and_move().

So shift the early initializations from move_to_new_page() into
migrate_page_move_mapping(), mapping and NULL-mapping paths.  Similarly
migrate_huge_page_move_mapping(), but its NULL-mapping path can just be
deleted: you cannot reach hugetlbfs_migrate_page() with a NULL mapping.

Adjust stages 3 to 8 in the Documentation file accordingly.

Signed-off-by: Hugh Dickins 
Cc: Christoph Lameter 
Cc: "Kirill A. Shutemov" 
Cc: Rik van Riel 
Cc: Vlastimil Babka 
Cc: Davidlohr Bueso 
Cc: Oleg Nesterov 
Cc: Sasha Levin 
Cc: Dmitry Vyukov 
Cc: KOSAKI Motohiro 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds