Merge branch 'i2c/for-5.4' of git://git./linux/kernel/git/wsa/linux

Pull i2c updates from Wolfram Sang:

 - new driver for ICY, an Amiga Zorro card :)

 - axxia driver gained slave mode support, NXP driver gained ACPI

 - the slave EEPROM backend gained 16 bit address support

 - and lots of regular driver updates and reworks

* 'i2c/for-5.4' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux: (52 commits)
  i2c: tegra: Move suspend handling to NOIRQ phase
  i2c: imx: ACPI support for NXP i2c controller
  i2c: uniphier(-f): remove all dev_dbg()
  i2c: uniphier(-f): use devm_platform_ioremap_resource()
  i2c: slave-eeprom: Add comment about address handling
  i2c: exynos5: Remove IRQF_ONESHOT
  i2c: stm32f7: Make structure stm32f7_i2c_algo constant
  i2c: cht-wc: drop check because i2c_unregister_device() is NULL safe
  i2c-eeprom_slave: Add support for more eeprom models
  i2c: fsi: Add of_put_node() before break
  i2c: synquacer: Make synquacer_i2c_ops constant
  i2c: hix5hd2: Remove IRQF_ONESHOT
  i2c: i801: Use iTCO version 6 in Cannon Lake PCH and beyond
  watchdog: iTCO: Add support for Cannon Lake PCH iTCO
  i2c: iproc: Make bcm_iproc_i2c_quirks constant
  i2c: iproc: Add full name of devicetree node to adapter name
  i2c: piix4: Add ACPI support
  i2c: piix4: Fix probing of reserved ports on AMD Family 16h Model 30h
  i2c: ocores: use request_any_context_irq() to register IRQ handler
  i2c: designware: Fix optional reset error handling
  ...

Merge tag 'sound-fix-5.4-rc1' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A few small remaining wrap-up for this merge window.

  Most of patches are device-specific (HD-audio and USB-audio quirks,
  FireWire, pcm316a, fsl, rsnd, Atmel, and TI fixes), while there is a
  simple fix (actually two commits) for ASoC core"

* tag 'sound-fix-5.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: usb-audio: Add DSD support for EVGA NU Audio
  ALSA: hda - Add laptop imic fixup for ASUS M9V laptop
  ASoC: ti: fix SND_SOC_DM365_VOICE_CODEC dependencies
  ASoC: pcm3168a: The codec does not support S32_LE
  ASoC: core: use list_del_init and move it back to soc_cleanup_component
  ALSA: hda/realtek - PCI quirk for Medion E4254
  ALSA: hda - Apply AMD controller workaround for Raven platform
  ASoC: rsnd: do error check after rsnd_channel_normalization()
  ASoC: atmel_ssc_dai: Remove wrong spinlock usage
  ASoC: core: delete component->card_list in soc_remove_component only
  ASoC: fsl_sai: Fix noise when using EDMA
  ALSA: usb-audio: Add Hiby device family to quirks for native DSD support
  ALSA: hda/realtek - Fix alienware headset mic
  ALSA: dice: fix wrong packet parameter for Alesis iO26

Merge tag 'for-5.4/io_uring-2019-09-24' of git://git.kernel.dk/linux-block

Pull more io_uring updates from Jens Axboe:
 "A collection of later fixes and additions, that weren't quite ready
  for pushing out with the initial pull request.

  This contains:

   - Fix potential use-after-free of shadow requests (Jackie)

   - Fix potential OOM crash in request allocation (Jackie)

   - kmalloc+memcpy -> kmemdup cleanup (Jackie)

   - Fix poll crash regression (me)

   - Fix SQ thread not being nice and giving up CPU for !PREEMPT (me)

   - Add support for timeouts, making it easier to do epoll_wait()
     conversions, for instance (me)

   - Ensure io_uring works without f_ops->read_iter() and
     f_ops->write_iter() (me)"

* tag 'for-5.4/io_uring-2019-09-24' of git://git.kernel.dk/linux-block:
  io_uring: correctly handle non ->{read,write}_iter() file_operations
  io_uring: IORING_OP_TIMEOUT support
  io_uring: use cond_resched() in sqthread
  io_uring: fix potential crash issue due to io_get_req failure
  io_uring: ensure poll commands clear ->sqe
  io_uring: fix use-after-free of shadow_req
  io_uring: use kmemdup instead of kmalloc and memcpy

Merge tag 'for-5.4/post-2019-09-24' of git://git.kernel.dk/linux-block

Pull more block updates from Jens Axboe:
 "Some later additions that weren't quite done for the first pull
  request, and also a few fixes that have arrived since.

  This contains:

   - Kill silly pktcdvd warning on attempting to register a non-scsi
     passthrough device (me)

   - Use symbolic constants for the block t10 protection types, and
     switch to handling it in core rather than in the drivers (Max)

   - libahci platform missing node put fix (Nishka)

   - Small series of fixes for BFQ (Paolo)

   - Fix possible nbd crash (Xiubo)"

* tag 'for-5.4/post-2019-09-24' of git://git.kernel.dk/linux-block:
  block: drop device references in bsg_queue_rq()
  block: t10-pi: fix -Wswitch warning
  pktcdvd: remove warning on attempting to register non-passthrough dev
  ata: libahci_platform: Add of_node_put() before loop exit
  nbd: fix possible page fault for nbd disk
  nbd: rename the runtime flags as NBD_RT_ prefixed
  block, bfq: push up injection only after setting service time
  block, bfq: increase update frequency of inject limit
  block, bfq: reduce upper bound for inject limit to max_rq_in_driver+1
  block, bfq: update inject limit only after injection occurred
  block: centralize PI remapping logic to the block layer
  block: use symbolic constants for t10_pi type

Merge branch 'akpm' (patches from Andrew)

Merge updates from Andrew Morton:

 - a few hot fixes

 - ocfs2 updates

 - almost all of -mm (slab-generic, slab, slub, kmemleak, kasan,
   cleanups, debug, pagecache, memcg, gup, pagemap, memory-hotplug,
   sparsemem, vmalloc, initialization, z3fold, compaction, mempolicy,
   oom-kill, hugetlb, migration, thp, mmap, madvise, shmem, zswap,
   zsmalloc)

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (132 commits)
  mm/zsmalloc.c: fix a -Wunused-function warning
  zswap: do not map same object twice
  zswap: use movable memory if zpool support allocate movable memory
  zpool: add malloc_support_movable to zpool_driver
  shmem: fix obsolete comment in shmem_getpage_gfp()
  mm/madvise: reduce code duplication in error handling paths
  mm: mmap: increase sockets maximum memory size pgoff for 32bits
  mm/mmap.c: refine find_vma_prev() with rb_last()
  riscv: make mmap allocation top-down by default
  mips: use generic mmap top-down layout and brk randomization
  mips: replace arch specific way to determine 32bit task with generic version
  mips: adjust brk randomization offset to fit generic version
  mips: use STACK_TOP when computing mmap base address
  mips: properly account for stack randomization and stack guard gap
  arm: use generic mmap top-down layout and brk randomization
  arm: use STACK_TOP when computing mmap base address
  arm: properly account for stack randomization and stack guard gap
  arm64, mm: make randomization selected by generic topdown mmap layout
  arm64, mm: move generic mmap layout functions to mm
  arm64: consider stack randomization for mmap base only when necessary
  ...

mm/zsmalloc.c: fix a -Wunused-function warning

set_zspage_inuse() was introduced in the commit 4f42047bbde0 ("zsmalloc:
use accessor") but all the users of it were removed later by the commits,

bdb0af7ca8f0 ("zsmalloc: factor page chain functionality out")
3783689a1aa8 ("zsmalloc: introduce zspage structure")

so the function can be safely removed now.

Link: http://lkml.kernel.org/r/1568658408-19374-1-git-send-email-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zswap: do not map same object twice

zswap_writeback_entry() maps a handle to read swpentry first, and
then in the most common case it would map the same handle again.
This is ok when zbud is the backend since its mapping callback is
plain and simple, but it slows things down for z3fold.

Since there's hardly a point in unmapping a handle _that_ fast as
zswap_writeback_entry() does when it reads swpentry, the
suggestion is to keep the handle mapped till the end.

Link: http://lkml.kernel.org/r/20190916004640.b453167d3556c4093af4cf7d@gmail.com
Signed-off-by: Vitaly Wool <vitalywool@gmail.com>
Reviewed-by: Dan Streetman <ddstreet@ieee.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Vitaly Wool <vitalywool@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zswap: use movable memory if zpool support allocate movable memory

This is the third version that was updated according to the comments from
Sergey Senozhatsky https://lkml.org/lkml/2019/5/29/73 and Shakeel Butt
https://lkml.org/lkml/2019/6/4/973

zswap compresses swap pages into a dynamically allocated RAM-based memory
pool.  The memory pool should be zbud, z3fold or zsmalloc.  All of them
will allocate unmovable pages.  It will increase the number of unmovable
page blocks that will bad for anti-fragment.

zsmalloc support page migration if request movable page:
        handle = zs_malloc(zram->mem_pool, comp_len,
                GFP_NOIO | __GFP_HIGHMEM |
                __GFP_MOVABLE);

And commit "zpool: Add malloc_support_movable to zpool_driver" add
zpool_malloc_support_movable check malloc_support_movable to make sure if
a zpool support allocate movable memory.

This commit let zswap allocate block with gfp
__GFP_HIGHMEM | __GFP_MOVABLE if zpool support allocate movable memory.

Following part is test log in a pc that has 8G memory and 2G swap.

Without this commit:
~# echo lz4 > /sys/module/zswap/parameters/compressor
~# echo zsmalloc > /sys/module/zswap/parameters/zpool
~# echo 1 > /sys/module/zswap/parameters/enabled
~# swapon /swapfile
~# cd /home/teawater/kernel/vm-scalability/
/home/teawater/kernel/vm-scalability# export unit_size=$((9 * 1024 * 1024 * 1024))
/home/teawater/kernel/vm-scalability# ./case-anon-w-seq
2717908992 bytes / 4826062 usecs = 549973 KB/s
2717908992 bytes / 4864201 usecs = 545661 KB/s
2717908992 bytes / 4867015 usecs = 545346 KB/s
2717908992 bytes / 4915485 usecs = 539968 KB/s
397853 usecs to free memory
357820 usecs to free memory
421333 usecs to free memory
420454 usecs to free memory
/home/teawater/kernel/vm-scalability# cat /proc/pagetypeinfo
Page block order: 9
Pages per block:  512

Free pages count per migrate type at order       0      1      2      3      4      5      6      7      8      9     10
Node    0, zone      DMA, type    Unmovable      1      1      1      0      2      1      1      0      1      0      0
Node    0, zone      DMA, type      Movable      0      0      0      0      0      0      0      0      0      1      3
Node    0, zone      DMA, type  Reclaimable      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone      DMA, type   HighAtomic      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone      DMA, type          CMA      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone      DMA, type      Isolate      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone    DMA32, type    Unmovable      6      5      8      6      6      5      4      1      1      1      0
Node    0, zone    DMA32, type      Movable     25     20     20     19     22     15     14     11     11      5    767
Node    0, zone    DMA32, type  Reclaimable      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone    DMA32, type   HighAtomic      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone    DMA32, type          CMA      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone    DMA32, type      Isolate      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone   Normal, type    Unmovable   4753   5588   5159   4613   3712   2520   1448    594    188     11      0
Node    0, zone   Normal, type      Movable     16      3    457   2648   2143   1435    860    459    223    224    296
Node    0, zone   Normal, type  Reclaimable      0      0     44     38     11      2      0      0      0      0      0
Node    0, zone   Normal, type   HighAtomic      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone   Normal, type          CMA      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone   Normal, type      Isolate      0      0      0      0      0      0      0      0      0      0      0

Number of blocks type     Unmovable      Movable  Reclaimable   HighAtomic          CMA      Isolate
Node 0, zone      DMA            1            7            0            0            0            0
Node 0, zone    DMA32            4         1652            0            0            0            0
Node 0, zone   Normal          931         1485           15            0            0            0

With this commit:
~# echo lz4 > /sys/module/zswap/parameters/compressor
~# echo zsmalloc > /sys/module/zswap/parameters/zpool
~# echo 1 > /sys/module/zswap/parameters/enabled
~# swapon /swapfile
~# cd /home/teawater/kernel/vm-scalability/
/home/teawater/kernel/vm-scalability# export unit_size=$((9 * 1024 * 1024 * 1024))
/home/teawater/kernel/vm-scalability# ./case-anon-w-seq
2717908992 bytes / 4689240 usecs = 566020 KB/s
2717908992 bytes / 4760605 usecs = 557535 KB/s
2717908992 bytes / 4803621 usecs = 552543 KB/s
2717908992 bytes / 5069828 usecs = 523530 KB/s
431546 usecs to free memory
383397 usecs to free memory
456454 usecs to free memory
224487 usecs to free memory
/home/teawater/kernel/vm-scalability# cat /proc/pagetypeinfo
Page block order: 9
Pages per block:  512

Free pages count per migrate type at order       0      1      2      3      4      5      6      7      8      9     10
Node    0, zone      DMA, type    Unmovable      1      1      1      0      2      1      1      0      1      0      0
Node    0, zone      DMA, type      Movable      0      0      0      0      0      0      0      0      0      1      3
Node    0, zone      DMA, type  Reclaimable      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone      DMA, type   HighAtomic      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone      DMA, type          CMA      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone      DMA, type      Isolate      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone    DMA32, type    Unmovable     10      8     10      9     10      4      3      2      3      0      0
Node    0, zone    DMA32, type      Movable     18     12     14     16     16     11      9      5      5      6    775
Node    0, zone    DMA32, type  Reclaimable      0      0      0      0      0      0      0      0      0      0      1
Node    0, zone    DMA32, type   HighAtomic      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone    DMA32, type          CMA      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone    DMA32, type      Isolate      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone   Normal, type    Unmovable   2669   1236    452    118     37     14      4      1      2      3      0
Node    0, zone   Normal, type      Movable   3850   6086   5274   4327   3510   2494   1520    934    438    220    470
Node    0, zone   Normal, type  Reclaimable     56     93    155    124     47     31     17      7      3      0      0
Node    0, zone   Normal, type   HighAtomic      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone   Normal, type          CMA      0      0      0      0      0      0      0      0      0      0      0
Node    0, zone   Normal, type      Isolate      0      0      0      0      0      0      0      0      0      0      0

Number of blocks type     Unmovable      Movable  Reclaimable   HighAtomic          CMA      Isolate
Node 0, zone      DMA            1            7            0            0            0            0
Node 0, zone    DMA32            4         1650            2            0            0            0
Node 0, zone   Normal           79         2326           26            0            0            0

You can see that the number of unmovable page blocks is decreased
when the kernel has this commit.

Link: http://lkml.kernel.org/r/20190605100630.13293-2-teawaterz@linux.alibaba.com
Signed-off-by: Hui Zhu <teawaterz@linux.alibaba.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Vitaly Wool <vitalywool@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zpool: add malloc_support_movable to zpool_driver

As a zpool_driver, zsmalloc can allocate movable memory because it support
migate pages.  But zbud and z3fold cannot allocate movable memory.

Add malloc_support_movable to zpool_driver.  If a zpool_driver support
allocate movable memory, set it to true.  And add
zpool_malloc_support_movable check malloc_support_movable to make sure if
a zpool support allocate movable memory.

Link: http://lkml.kernel.org/r/20190605100630.13293-1-teawaterz@linux.alibaba.com
Signed-off-by: Hui Zhu <teawaterz@linux.alibaba.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Vitaly Wool <vitalywool@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

shmem: fix obsolete comment in shmem_getpage_gfp()

Replace "fault_mm" with "vmf" in code comment because commit cfda05267f7b
("userfaultfd: shmem: add userfaultfd hook for shared memory faults") has
changed the prototpye of shmem_getpage_gfp() - pass vmf instead of
fault_mm to the function.

Before:
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
struct page **pagep, enum sgp_type sgp,
gfp_t gfp, struct mm_struct *fault_mm, int *fault_type);
After:
static int shmem_getpage_gfp(struct inode *inode, pgoff_t index,
struct page **pagep, enum sgp_type sgp,
gfp_t gfp, struct vm_area_struct *vma,
struct vm_fault *vmf, vm_fault_t *fault_type);

Link: http://lkml.kernel.org/r/20190816100204.9781-1-miles.chen@mediatek.com
Signed-off-by: Miles Chen <miles.chen@mediatek.com>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/madvise: reduce code duplication in error handling paths

madvise_behavior() converts -ENOMEM to -EAGAIN in several places using
identical code.

Move that code to a common error handling path.

No functional changes.

Link: http://lkml.kernel.org/r/1564640896-1210-1-git-send-email-rppt@linux.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Pankaj Gupta <pagupta@redhat.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: mmap: increase sockets maximum memory size pgoff for 32bits

The AF_XDP sockets umem mapping interface uses XDP_UMEM_PGOFF_FILL_RING
and XDP_UMEM_PGOFF_COMPLETION_RING offsets.  These offsets are
established already and are part of the configuration interface.

But for 32-bit systems, using AF_XDP socket configuration, these values
are too large to pass the maximum allowed file size verification.  The
offsets can be tuned off, but instead of changing the existing
interface, let's extend the max allowed file size for sockets.

No one has been using this until this patch with 32 bits as without
this fix af_xdp sockets can't be used at all, so it unblocks af_xdp
socket usage for 32bit systems.

All list of mmap cbs for sockets was verified for side effects and all
of them contain dummy cb - sock_no_mmap() at this moment, except the
following:

xsk_mmap() - it's what this fix is needed for.
tcp_mmap() - doesn't have obvious issues with pgoff - no any references on it.
packet_mmap() - return -EINVAL if it's even set.

Link: http://lkml.kernel.org/r/20190812124326.32146-1-ivan.khoronzhuk@linaro.org
Signed-off-by: Ivan Khoronzhuk <ivan.khoronzhuk@linaro.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Magnus Karlsson <magnus.karlsson@intel.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/mmap.c: refine find_vma_prev() with rb_last()

When addr is out of range of the whole rb_tree, pprev will point to the
right-most node.  rb_tree facility already provides a helper function,
rb_last(), to do this task.  We can leverage this instead of
reimplementing it.

This patch refines find_vma_prev() with rb_last() to make it a little
nicer to read.

[akpm@linux-foundation.org: little cleanup, per Vlastimil]
Link: http://lkml.kernel.org/r/20190809001928.4950-1-richardw.yang@linux.intel.com
Signed-off-by: Wei Yang <richardw.yang@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Michal Hocko <mhocko@suse.com>
Cc: 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>

riscv: make mmap allocation top-down by default

In order to avoid wasting user address space by using bottom-up mmap
allocation scheme, prefer top-down scheme when possible.

Before:
root@qemuriscv64:~# cat /proc/self/maps
00010000-00016000 r-xp 00000000 fe:00 6389       /bin/cat.coreutils
00016000-00017000 r--p 00005000 fe:00 6389       /bin/cat.coreutils
00017000-00018000 rw-p 00006000 fe:00 6389       /bin/cat.coreutils
00018000-00039000 rw-p 00000000 00:00 0          [heap]
1555556000-155556d000 r-xp 00000000 fe:00 7193   /lib/ld-2.28.so
155556d000-155556e000 r--p 00016000 fe:00 7193   /lib/ld-2.28.so
155556e000-155556f000 rw-p 00017000 fe:00 7193   /lib/ld-2.28.so
155556f000-1555570000 rw-p 00000000 00:00 0
1555570000-1555572000 r-xp 00000000 00:00 0      [vdso]
1555574000-1555576000 rw-p 00000000 00:00 0
1555576000-1555674000 r-xp 00000000 fe:00 7187   /lib/libc-2.28.so
1555674000-1555678000 r--p 000fd000 fe:00 7187   /lib/libc-2.28.so
1555678000-155567a000 rw-p 00101000 fe:00 7187   /lib/libc-2.28.so
155567a000-15556a0000 rw-p 00000000 00:00 0
3fffb90000-3fffbb1000 rw-p 00000000 00:00 0      [stack]

After:
root@qemuriscv64:~# cat /proc/self/maps
00010000-00016000 r-xp 00000000 fe:00 6389       /bin/cat.coreutils
00016000-00017000 r--p 00005000 fe:00 6389       /bin/cat.coreutils
00017000-00018000 rw-p 00006000 fe:00 6389       /bin/cat.coreutils
2de81000-2dea2000 rw-p 00000000 00:00 0          [heap]
3ff7eb6000-3ff7ed8000 rw-p 00000000 00:00 0
3ff7ed8000-3ff7fd6000 r-xp 00000000 fe:00 7187   /lib/libc-2.28.so
3ff7fd6000-3ff7fda000 r--p 000fd000 fe:00 7187   /lib/libc-2.28.so
3ff7fda000-3ff7fdc000 rw-p 00101000 fe:00 7187   /lib/libc-2.28.so
3ff7fdc000-3ff7fe2000 rw-p 00000000 00:00 0
3ff7fe4000-3ff7fe6000 r-xp 00000000 00:00 0      [vdso]
3ff7fe6000-3ff7ffd000 r-xp 00000000 fe:00 7193   /lib/ld-2.28.so
3ff7ffd000-3ff7ffe000 r--p 00016000 fe:00 7193   /lib/ld-2.28.so
3ff7ffe000-3ff7fff000 rw-p 00017000 fe:00 7193   /lib/ld-2.28.so
3ff7fff000-3ff8000000 rw-p 00000000 00:00 0
3fff888000-3fff8a9000 rw-p 00000000 00:00 0      [stack]

[alex@ghiti.fr: v6]
Link: http://lkml.kernel.org/r/20190808061756.19712-15-alex@ghiti.fr
Link: http://lkml.kernel.org/r/20190730055113.23635-15-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Acked-by: Paul Walmsley <paul.walmsley@sifive.com> [arch/riscv]
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mips: use generic mmap top-down layout and brk randomization

mips uses a top-down layout by default that exactly fits the generic
functions, so get rid of arch specific code and use the generic version by
selecting ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT.

As ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT selects ARCH_HAS_ELF_RANDOMIZE,
use the generic version of arch_randomize_brk since it also fits.  Note
that this commit also removes the possibility for mips to have elf
randomization and no MMU: without MMU, the security added by randomization
is worth nothing.

Link: http://lkml.kernel.org/r/20190730055113.23635-14-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Paul Burton <paul.burton@mips.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mips: replace arch specific way to determine 32bit task with generic version

Mips uses TASK_IS_32BIT_ADDR to determine if a task is 32bit, but this
define is mips specific and other arches do not have it: instead, use
!IS_ENABLED(CONFIG_64BIT) || is_compat_task() condition.

Link: http://lkml.kernel.org/r/20190730055113.23635-13-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Paul Burton <paul.burton@mips.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mips: adjust brk randomization offset to fit generic version

This commit simply bumps up to 32MB and 1GB the random offset of brk,
compared to 8MB and 256MB, for 32bit and 64bit respectively.

Link: http://lkml.kernel.org/r/20190730055113.23635-12-alex@ghiti.fr
Suggested-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Paul Burton <paul.burton@mips.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mips: use STACK_TOP when computing mmap base address

mmap base address must be computed wrt stack top address, using TASK_SIZE
is wrong since STACK_TOP and TASK_SIZE are not equivalent.

Link: http://lkml.kernel.org/r/20190730055113.23635-11-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Paul Burton <paul.burton@mips.com>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mips: properly account for stack randomization and stack guard gap

This commit takes care of stack randomization and stack guard gap when
computing mmap base address and checks if the task asked for
randomization.  This fixes the problem uncovered and not fixed for arm
here: https://lkml.kernel.org/r/20170622200033.25714-1-riel@redhat.com

Link: http://lkml.kernel.org/r/20190730055113.23635-10-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Kees Cook <keescook@chromium.org>
Acked-by: Paul Burton <paul.burton@mips.com>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm: use generic mmap top-down layout and brk randomization

arm uses a top-down mmap layout by default that exactly fits the generic
functions, so get rid of arch specific code and use the generic version by
selecting ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT.

As ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT selects ARCH_HAS_ELF_RANDOMIZE,
use the generic version of arch_randomize_brk since it also fits.  Note
that this commit also removes the possibility for arm to have elf
randomization and no MMU: without MMU, the security added by randomization
is worth nothing.

Note that it is safe to remove STACK_RND_MASK since it matches the default
value.

Link: http://lkml.kernel.org/r/20190730055113.23635-9-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm: use STACK_TOP when computing mmap base address

mmap base address must be computed wrt stack top address, using TASK_SIZE
is wrong since STACK_TOP and TASK_SIZE are not equivalent.

Link: http://lkml.kernel.org/r/20190730055113.23635-8-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm: properly account for stack randomization and stack guard gap

This commit takes care of stack randomization and stack guard gap when
computing mmap base address and checks if the task asked for
randomization.  This fixes the problem uncovered and not fixed for arm
here: https://lkml.kernel.org/r/20170622200033.25714-1-riel@redhat.com

Link: http://lkml.kernel.org/r/20190730055113.23635-7-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64, mm: make randomization selected by generic topdown mmap layout

This commits selects ARCH_HAS_ELF_RANDOMIZE when an arch uses the generic
topdown mmap layout functions so that this security feature is on by
default.

Note that this commit also removes the possibility for arm64 to have elf
randomization and no MMU: without MMU, the security added by randomization
is worth nothing.

Link: http://lkml.kernel.org/r/20190730055113.23635-6-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64, mm: move generic mmap layout functions to mm

arm64 handles top-down mmap layout in a way that can be easily reused by
other architectures, so make it available in mm.  It then introduces a new
config ARCH_WANT_DEFAULT_TOPDOWN_MMAP_LAYOUT that can be set by other
architectures to benefit from those functions.  Note that this new config
depends on MMU being enabled, if selected without MMU support, a warning
will be thrown.

Link: http://lkml.kernel.org/r/20190730055113.23635-5-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Suggested-by: Christoph Hellwig <hch@infradead.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64: consider stack randomization for mmap base only when necessary

Do not offset mmap base address because of stack randomization if current
task does not want randomization.  Note that x86 already implements this
behaviour.

Link: http://lkml.kernel.org/r/20190730055113.23635-4-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64: make use of is_compat_task instead of hardcoding this test

Each architecture has its own way to determine if a task is a compat task,
by using is_compat_task in arch_mmap_rnd, it allows more genericity and
then it prepares its moving to mm/.

Link: http://lkml.kernel.org/r/20190730055113.23635-3-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Paul Burton <paul.burton@mips.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, fs: move randomize_stack_top from fs to mm

Patch series "Provide generic top-down mmap layout functions", v6.

This series introduces generic functions to make top-down mmap layout
easily accessible to architectures, in particular riscv which was the
initial goal of this series.  The generic implementation was taken from
arm64 and used successively by arm, mips and finally riscv.

Note that in addition the series fixes 2 issues:

- stack randomization was taken into account even if not necessary.

- [1] fixed an issue with mmap base which did not take into account
  randomization but did not report it to arm and mips, so by moving arm64
  into a generic library, this problem is now fixed for both
  architectures.

This work is an effort to factorize architecture functions to avoid code
duplication and oversights as in [1].

[1]: https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg1429066.html

This patch (of 14):

This preparatory commit moves this function so that further introduction
of generic topdown mmap layout is contained only in mm/util.c.

Link: http://lkml.kernel.org/r/20190730055113.23635-2-alex@ghiti.fr
Signed-off-by: Alexandre Ghiti <alex@ghiti.fr>
Acked-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Cc: Russell King <linux@armlinux.org.uk>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Paul Burton <paul.burton@mips.com>
Cc: James Hogan <jhogan@kernel.org>
Cc: Palmer Dabbelt <palmer@sifive.com>
Cc: Albert Ou <aou@eecs.berkeley.edu>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uprobe: collapse THP pmd after removing all uprobes

After all uprobes are removed from the huge page (with PTE pgtable), it is
possible to collapse the pmd and benefit from THP again.  This patch does
the collapse by calling collapse_pte_mapped_thp().

Link: http://lkml.kernel.org/r/20190815164525.1848545-7-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: kbuild test robot <lkp@intel.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

khugepaged: enable collapse pmd for pte-mapped THP

khugepaged needs exclusive mmap_sem to access page table.  When it fails
to lock mmap_sem, the page will fault in as pte-mapped THP.  As the page
is already a THP, khugepaged will not handle this pmd again.

This patch enables the khugepaged to retry collapse the page table.

struct mm_slot (in khugepaged.c) is extended with an array, containing
addresses of pte-mapped THPs.  We use array here for simplicity.  We can
easily replace it with more advanced data structures when needed.

In khugepaged_scan_mm_slot(), if the mm contains pte-mapped THP, we try to
collapse the page table.

Since collapse may happen at an later time, some pages may already fault
in.  collapse_pte_mapped_thp() is added to properly handle these pages.
collapse_pte_mapped_thp() also double checks whether all ptes in this pmd
are mapping to the same THP.  This is necessary because some subpage of
the THP may be replaced, for example by uprobe.  In such cases, it is not
possible to collapse the pmd.

[kirill.shutemov@linux.intel.com: add comments for retract_page_tables()]
Link: http://lkml.kernel.org/r/20190816145443.6ard3iilytc6jlgv@box
Link: http://lkml.kernel.org/r/20190815164525.1848545-6-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

uprobe: use FOLL_SPLIT_PMD instead of FOLL_SPLIT

Use the newly added FOLL_SPLIT_PMD in uprobe.  This preserves the huge
page when the uprobe is enabled.  When the uprobe is disabled, newer
instances of the same application could still benefit from huge page.

For the next step, we will enable khugepaged to regroup the pmd, so that
existing instances of the application could also benefit from huge page
after the uprobe is disabled.

Link: http://lkml.kernel.org/r/20190815164525.1848545-5-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, thp: introduce FOLL_SPLIT_PMD

Introduce a new foll_flag: FOLL_SPLIT_PMD.  As the name says
FOLL_SPLIT_PMD splits huge pmd for given mm_struct, the underlining huge
page stays as-is.

FOLL_SPLIT_PMD is useful for cases where we need to use regular pages, but
would switch back to huge page and huge pmd on.  One of such example is
uprobe.  The following patches use FOLL_SPLIT_PMD in uprobe.

Link: http://lkml.kernel.org/r/20190815164525.1848545-4-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.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>

uprobe: use original page when all uprobes are removed

Currently, uprobe swaps the target page with a anonymous page in both
install_breakpoint() and remove_breakpoint().  When all uprobes on a page
are removed, the given mm is still using an anonymous page (not the
original page).

This patch allows uprobe to use original page when possible (all uprobes
on the page are already removed, and the original page is in page cache
and uptodate).

As suggested by Oleg, we unmap the old_page and let the original page
fault in.

Link: http://lkml.kernel.org/r/20190815164525.1848545-3-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Suggested-by: Oleg Nesterov <oleg@redhat.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move memcmp_pages() and pages_identical()

Patch series "THP aware uprobe", v13.

This patchset makes uprobe aware of THPs.

Currently, when uprobe is attached to text on THP, the page is split by
FOLL_SPLIT.  As a result, uprobe eliminates the performance benefit of
THP.

This set makes uprobe THP-aware.  Instead of FOLL_SPLIT, we introduces
FOLL_SPLIT_PMD, which only split PMD for uprobe.

After all uprobes within the THP are removed, the PTE-mapped pages are
regrouped as huge PMD.

This set (plus a few THP patches) is also available at

   https://github.com/liu-song-6/linux/tree/uprobe-thp

This patch (of 6):

Move memcmp_pages() to mm/util.c and pages_identical() to mm.h, so that we
can use them in other files.

Link: http://lkml.kernel.org/r/20190815164525.1848545-2-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matthew Wilcox <matthew.wilcox@oracle.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: thp: make deferred split shrinker memcg aware

Currently THP deferred split shrinker is not memcg aware, this may cause
premature OOM with some configuration.  For example the below test would
run into premature OOM easily:

$ cgcreate -g memory:thp
$ echo 4G > /sys/fs/cgroup/memory/thp/memory/limit_in_bytes
$ cgexec -g memory:thp transhuge-stress 4000

transhuge-stress comes from kernel selftest.

It is easy to hit OOM, but there are still a lot THP on the deferred split
queue, memcg direct reclaim can't touch them since the deferred split
shrinker is not memcg aware.

Convert deferred split shrinker memcg aware by introducing per memcg
deferred split queue.  The THP should be on either per node or per memcg
deferred split queue if it belongs to a memcg.  When the page is
immigrated to the other memcg, it will be immigrated to the target memcg's
deferred split queue too.

Reuse the second tail page's deferred_list for per memcg list since the
same THP can't be on multiple deferred split queues.

[yang.shi@linux.alibaba.com: simplify deferred split queue dereference per Kirill Tkhai]
Link: http://lkml.kernel.org/r/1566496227-84952-5-git-send-email-yang.shi@linux.alibaba.com
Link: http://lkml.kernel.org/r/1565144277-36240-5-git-send-email-yang.shi@linux.alibaba.com
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
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: shrinker: make shrinker not depend on memcg kmem

Currently shrinker is just allocated and can work when memcg kmem is
enabled.  But, THP deferred split shrinker is not slab shrinker, it
doesn't make too much sense to have such shrinker depend on memcg kmem.
It should be able to reclaim THP even though memcg kmem is disabled.

Introduce a new shrinker flag, SHRINKER_NONSLAB, for non-slab shrinker.
When memcg kmem is disabled, just such shrinkers can be called in
shrinking memcg slab.

[yang.shi@linux.alibaba.com: add comment]
Link: http://lkml.kernel.org/r/1566496227-84952-4-git-send-email-yang.shi@linux.alibaba.com
Link: http://lkml.kernel.org/r/1565144277-36240-4-git-send-email-yang.shi@linux.alibaba.com
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
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: move mem_cgroup_uncharge out of __page_cache_release()

A later patch makes THP deferred split shrinker memcg aware, but it needs
page->mem_cgroup information in THP destructor, which is called after
mem_cgroup_uncharge() now.

So move mem_cgroup_uncharge() from __page_cache_release() to compound page
destructor, which is called by both THP and other compound pages except
HugeTLB.  And call it in __put_single_page() for single order page.

Link: http://lkml.kernel.org/r/1565144277-36240-3-git-send-email-yang.shi@linux.alibaba.com
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
Suggested-by: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
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: thp: extract split_queue_* into a struct

Patch series "Make deferred split shrinker memcg aware", v6.

Currently THP deferred split shrinker is not memcg aware, this may cause
premature OOM with some configuration.  For example the below test would
run into premature OOM easily:

$ cgcreate -g memory:thp
$ echo 4G > /sys/fs/cgroup/memory/thp/memory/limit_in_bytes
$ cgexec -g memory:thp transhuge-stress 4000

transhuge-stress comes from kernel selftest.

It is easy to hit OOM, but there are still a lot THP on the deferred split
queue, memcg direct reclaim can't touch them since the deferred split
shrinker is not memcg aware.

Convert deferred split shrinker memcg aware by introducing per memcg
deferred split queue.  The THP should be on either per node or per memcg
deferred split queue if it belongs to a memcg.  When the page is
immigrated to the other memcg, it will be immigrated to the target memcg's
deferred split queue too.

Reuse the second tail page's deferred_list for per memcg list since the
same THP can't be on multiple deferred split queues.

Make deferred split shrinker not depend on memcg kmem since it is not
slab.  It doesn't make sense to not shrink THP even though memcg kmem is
disabled.

With the above change the test demonstrated above doesn't trigger OOM even
though with cgroup.memory=nokmem.

This patch (of 4):

Put split_queue, split_queue_lock and split_queue_len into a struct in
order to reduce code duplication when we convert deferred_split to memcg
aware in the later patches.

Link: http://lkml.kernel.org/r/1565144277-36240-2-git-send-email-yang.shi@linux.alibaba.com
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
Suggested-by: "Kirill A . Shutemov" <kirill.shutemov@linux.intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reviewed-by: Kirill Tkhai <ktkhai@virtuozzo.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Qian Cai <cai@lca.pw>
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,thp: avoid writes to file with THP in pagecache

In previous patch, an application could put part of its text section in
THP via madvise().  These THPs will be protected from writes when the
application is still running (TXTBSY).  However, after the application
exits, the file is available for writes.

This patch avoids writes to file THP by dropping page cache for the file
when the file is open for write.  A new counter nr_thps is added to struct
address_space.  In do_dentry_open(), if the file is open for write and
nr_thps is non-zero, we drop page cache for the whole file.

Link: http://lkml.kernel.org/r/20190801184244.3169074-8-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Reported-by: kbuild test robot <lkp@intel.com>
Acked-by: Rik van Riel <riel@surriel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm,thp: add read-only THP support for (non-shmem) FS

This patch is (hopefully) the first step to enable THP for non-shmem
filesystems.

This patch enables an application to put part of its text sections to THP
via madvise, for example:

    madvise((void *)0x600000, 0x200000, MADV_HUGEPAGE);

We tried to reuse the logic for THP on tmpfs.

Currently, write is not supported for non-shmem THP.  khugepaged will only
process vma with VM_DENYWRITE.  sys_mmap() ignores VM_DENYWRITE requests
(see ksys_mmap_pgoff).  The only way to create vma with VM_DENYWRITE is
execve().  This requirement limits non-shmem THP to text sections.

The next patch will handle writes, which would only happen when the all
the vmas with VM_DENYWRITE are unmapped.

An EXPERIMENTAL config, READ_ONLY_THP_FOR_FS, is added to gate this
feature.

[songliubraving@fb.com: fix build without CONFIG_SHMEM]
Link: http://lkml.kernel.org/r/F53407FB-96CC-42E8-9862-105C92CC2B98@fb.com
[songliubraving@fb.com: fix double unlock in collapse_file()]
Link: http://lkml.kernel.org/r/B960CBFA-8EFC-4DA4-ABC5-1977FFF2CA57@fb.com
Link: http://lkml.kernel.org/r/20190801184244.3169074-7-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Rik van Riel <riel@surriel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

khugepaged: rename collapse_shmem() and khugepaged_scan_shmem()

Next patch will add khugepaged support of non-shmem files.  This patch
renames these two functions to reflect the new functionality:

    collapse_shmem()        =>  collapse_file()
    khugepaged_scan_shmem() =>  khugepaged_scan_file()

Link: http://lkml.kernel.org/r/20190801184244.3169074-6-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Rik van Riel <riel@surriel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm,thp: stats for file backed THP

In preparation for non-shmem THP, this patch adds a few stats and exposes
them in /proc/meminfo, /sys/bus/node/devices/<node>/meminfo, and
/proc/<pid>/task/<tid>/smaps.

This patch is mostly a rewrite of Kirill A.  Shutemov's earlier version:
https://lkml.kernel.org/r/20170126115819.58875-5-kirill.shutemov@linux.intel.com/

Link: http://lkml.kernel.org/r/20190801184244.3169074-5-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Rik van Riel <riel@surriel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

filemap: update offset check in filemap_fault()

With THP, current check of offset:

    VM_BUG_ON_PAGE(page->index != offset, page);

is no longer accurate. Update it to:

    VM_BUG_ON_PAGE(page_to_pgoff(page) != offset, page);

Link: http://lkml.kernel.org/r/20190801184244.3169074-4-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Rik van Riel <riel@surriel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

filemap: check compound_head(page)->mapping in pagecache_get_page()

Similar to previous patch, pagecache_get_page() avoids race condition with
truncate by checking page->mapping == mapping.  This does not work for
compound pages.  This patch let it check compound_head(page)->mapping
instead.

Link: http://lkml.kernel.org/r/20190801184244.3169074-3-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

filemap: check compound_head(page)->mapping in filemap_fault()

Patch series "Enable THP for text section of non-shmem files", v10;

This patchset follows up discussion at LSF/MM 2019.  The motivation is to
put text section of an application in THP, and thus reduces iTLB miss rate
and improves performance.  Both Facebook and Oracle showed strong
interests to this feature.

To make reviews easier, this set aims a mininal valid product.  Current
version of the work does not have any changes to file system specific
code.  This comes with some limitations (discussed later).

This set enables an application to "hugify" its text section by simply
running something like:

          madvise(0x600000, 0x80000, MADV_HUGEPAGE);

Before this call, the /proc/<pid>/maps looks like:

    00400000-074d0000 r-xp 00000000 00:27 2006927     app

After this call, part of the text section is split out and mapped to
THP:

    00400000-00425000 r-xp 00000000 00:27 2006927     app
    00600000-00e00000 r-xp 00200000 00:27 2006927     app   <<< on THP
    00e00000-074d0000 r-xp 00a00000 00:27 2006927     app

Limitations:

1. This only works for text section (vma with VM_DENYWRITE).
2. Original limitation #2 is removed in v3.

We gated this feature with an experimental config, READ_ONLY_THP_FOR_FS.
Once we get better support on the write path, we can remove the config and
enable it by default.

Tested cases:
1. Tested with btrfs and ext4.
2. Tested with real work application (memcache like caching service).
3. Tested with "THP aware uprobe":
   https://patchwork.kernel.org/project/linux-mm/list/?series=131339

This patch (of 7):

Currently, filemap_fault() avoids race condition with truncate by checking
page->mapping == mapping.  This does not work for compound pages.  This
patch let it check compound_head(page)->mapping instead.

Link: http://lkml.kernel.org/r/20190801184244.3169074-2-songliubraving@fb.com
Signed-off-by: Song Liu <songliubraving@fb.com>
Acked-by: Rik van Riel <riel@surriel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

thp: update split_huge_page_pmd() comment

According to 78ddc5347341 ("thp: rename split_huge_page_pmd() to
split_huge_pmd()"), update related comment.

Link: http://lkml.kernel.org/r/20190731033406.185285-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: "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/migrate.c: clean up useless code in migrate_vma_collect_pmd()

Remove unused 'pfn' variable.

Link: http://lkml.kernel.org/r/1565167272-21453-1-git-send-email-kernelfans@gmail.com
Signed-off-by: Pingfan Liu <kernelfans@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Ralph Campbell <rcampbell@nvidia.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Jan Kara <jack@suse.cz>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: don't retry when pool page allocations start to fail

When allocating hugetlbfs pool pages via /proc/sys/vm/nr_hugepages, the
pages will be interleaved between all nodes of the system.  If nodes are
not equal, it is quite possible for one node to fill up before the others.
When this happens, the code still attempts to allocate pages from the
full node.  This results in calls to direct reclaim and compaction which
slow things down considerably.

When allocating pool pages, note the state of the previous allocation for
each node.  If previous allocation failed, do not use the aggressive retry
algorithm on successive attempts.  The allocation will still succeed if
there is memory available, but it will not try as hard to free up memory.

Link: http://lkml.kernel.org/r/20190806014744.15446-5-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
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, compaction: raise compaction priority after it withdrawns

Mike Kravetz reports that "hugetlb allocations could stall for minutes or
hours when should_compact_retry() would return true more often then it
should.  Specifically, this was in the case where compact_result was
COMPACT_DEFERRED and COMPACT_PARTIAL_SKIPPED and no progress was being
made."

The problem is that the compaction_withdrawn() test in
should_compact_retry() includes compaction outcomes that are only possible
on low compaction priority, and results in a retry without increasing the
priority.  This may result in furter reclaim, and more incomplete
compaction attempts.

With this patch, compaction priority is raised when possible, or
should_compact_retry() returns false.

The COMPACT_SKIPPED result doesn't really fit together with the other
outcomes in compaction_withdrawn(), as that's a result caused by
insufficient order-0 pages, not due to low compaction priority.  With this
patch, it is moved to a new compaction_needs_reclaim() function, and for
that outcome we keep the current logic of retrying if it looks like
reclaim will be able to help.

Link: http://lkml.kernel.org/r/20190806014744.15446-4-mike.kravetz@oracle.com
Reported-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Tested-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
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, reclaim: cleanup should_continue_reclaim()

After commit "mm, reclaim: make should_continue_reclaim perform dryrun
detection", closer look at the function shows, that nr_reclaimed == 0
means the function will always return false.  And since non-zero
nr_reclaimed implies non_zero nr_scanned, testing nr_scanned serves no
purpose, and so does the testing for __GFP_RETRY_MAYFAIL.

This patch thus cleans up the function to test only !nr_reclaimed upfront,
and remove the __GFP_RETRY_MAYFAIL test and nr_scanned parameter
completely.  Comment is also updated, explaining that approximating "full
LRU list has been scanned" with nr_scanned == 0 didn't really work.

Link: http://lkml.kernel.org/r/20190806014744.15446-3-mike.kravetz@oracle.com
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
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, reclaim: make should_continue_reclaim perform dryrun detection

Patch series "address hugetlb page allocation stalls", v2.

Allocation of hugetlb pages via sysctl or procfs can stall for minutes or
hours.  A simple example on a two node system with 8GB of memory is as
follows:

echo 4096 > /sys/devices/system/node/node1/hugepages/hugepages-2048kB/nr_hugepages
echo 4096 > /proc/sys/vm/nr_hugepages

Obviously, both allocation attempts will fall short of their 8GB goal.
However, one or both of these commands may stall and not be interruptible.
The issues were initially discussed in mail thread [1] and RFC code at
[2].

This series addresses the issues causing the stalls.  There are two
distinct fixes, a cleanup, and an optimization.  The reclaim patch by
Hillf and compaction patch by Vlasitmil address corner cases in their
respective areas.  hugetlb page allocation could stall due to either of
these issues.  Vlasitmil added a cleanup patch after Hillf's
modifications.  The hugetlb patch by Mike is an optimization suggested
during the debug and development process.

[1] http://lkml.kernel.org/r/d38a095e-dc39-7e82-bb76-2c9247929f07@oracle.com
[2] http://lkml.kernel.org/r/20190724175014.9935-1-mike.kravetz@oracle.com

This patch (of 4):

Address the issue of should_continue_reclaim returning true too often for
__GFP_RETRY_MAYFAIL attempts when !nr_reclaimed and nr_scanned.  This was
observed during hugetlb page allocation causing stalls for minutes or
hours.

We can stop reclaiming pages if compaction reports it can make a progress.
There might be side-effects for other high-order allocations that would
potentially benefit from reclaiming more before compaction so that they
would be faster and less likely to stall.  However, the consequences of
premature/over-reclaim are considered worse.

We can also bail out of reclaiming pages if we know that there are not
enough inactive lru pages left to satisfy the costly allocation.

We can give up reclaiming pages too if we see dryrun occur, with the
certainty of plenty of inactive pages.  IOW with dryrun detected, we are
sure we have reclaimed as many pages as we could.

Link: http://lkml.kernel.org/r/20190806014744.15446-2-mike.kravetz@oracle.com
Signed-off-by: Hillf Danton <hdanton@sina.com>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Tested-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg, kmem: deprecate kmem.limit_in_bytes

Cgroup v1 memcg controller has exposed a dedicated kmem limit to users
which turned out to be really a bad idea because there are paths which
cannot shrink the kernel memory usage enough to get below the limit (e.g.
because the accounted memory is not reclaimable).  There are cases when
the failure is even not allowed (e.g.  __GFP_NOFAIL).  This means that the
kmem limit is in excess to the hard limit without any way to shrink and
thus completely useless.  OOM killer cannot be invoked to handle the
situation because that would lead to a premature oom killing.

As a result many places might see ENOMEM returning from kmalloc and result
in unexpected errors.  E.g.  a global OOM killer when there is a lot of
free memory because ENOMEM is translated into VM_FAULT_OOM in #PF path and
therefore pagefault_out_of_memory would result in OOM killer.

Please note that the kernel memory is still accounted to the overall limit
along with the user memory so removing the kmem specific limit should
still allow to contain kernel memory consumption.  Unlike the kmem one,
though, it invokes memory reclaim and targeted memcg oom killing if
necessary.

Start the deprecation process by crying to the kernel log.  Let's see
whether there are relevant usecases and simply return to EINVAL in the
second stage if nobody complains in few releases.

[akpm@linux-foundation.org: tweak documentation text]
Link: http://lkml.kernel.org/r/20190911151612.GI4023@dhcp22.suse.cz
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Thomas Lindroth <thomas.lindroth@gmail.com>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcontrol.c: fix a -Wunused-function warning

mem_cgroup_id_get() was introduced in commit 73f576c04b94 ("mm:memcontrol:
fix cgroup creation failure after many small jobs").

Later, it no longer has any user since the commits,

1f47b61fb407 ("mm: memcontrol: fix swap counter leak on swapout from offline cgroup")
58fa2a5512d9 ("mm: memcontrol: add sanity checks for memcg->id.ref on get/put")

so safe to remove it.

Link: http://lkml.kernel.org/r/1568648453-5482-1-git-send-email-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Acked-by: Michal Hocko <mhocko@suse.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, oom: consider present pages for the node size

constrained_alloc() calculates the size of the oom domain by using
node_spanned_pages which is incorrect because this is the full range of
the physical memory range that the numa node occupies rather than the
memory that backs that range which is represented by node_present_pages.

Sparsely populated nodes (e.g.  after memory hot remove or simply sparse
due to memory layout) can have really a large difference between the two.
This shouldn't really cause any real user observable problems because the
oom calculates a ratio against totalpages and used memory cannot exceed
present pages but it is confusing and wrong from code point of view.

Link: http://lkml.kernel.org/r/20190829163443.899-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/oom_kill.c: fix oom_cpuset_eligible() comment

Commit ac311a14c682 ("oom: decouple mems_allowed from
oom_unkillable_task") changed has_intersects_mems_allowed() to
oom_cpuset_eligible(), but didn't change the comment.

Link: http://lkml.kernel.org/r/1566959929-10638-1-git-send-email-wang.yi59@zte.com.cn
Signed-off-by: Yi Wang <wang.yi59@zte.com.cn>
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/oom: add oom_score_adj and pgtables to Killed process message

For an OOM event: print oom_score_adj value for the OOM Killed process to
document what the oom score adjust value was at the time the process was
OOM Killed.  The adjustment value can be set by user code and it affects
the resulting oom_score so it is used to influence kill process selection.

When eligible tasks are not printed (sysctl oom_dump_tasks = 0) printing
this value is the only documentation of the value for the process being
killed.  Having this value on the Killed process message is useful to
document if a miscconfiguration occurred or to confirm that the
oom_score_adj configuration applies as expected.

An example which illustates both misconfiguration and validation that the
oom_score_adj was applied as expected is:

Aug 14 23:00:02 testserver kernel: Out of memory: Killed process 2692
 (systemd-udevd) total-vm:1056800kB, anon-rss:1052760kB, file-rss:4kB,
 shmem-rss:0kB pgtables:22kB oom_score_adj:1000

The systemd-udevd is a critical system application that should have an
oom_score_adj of -1000.  It was miconfigured to have a adjustment of 1000
making it a highly favored OOM kill target process.  The output documents
both the misconfiguration and the fact that the process was correctly
targeted by OOM due to the miconfiguration.  This can be quite helpful for
triage and problem determination.

The addition of the pgtables_bytes shows page table usage by the process
and is a useful measure of the memory size of the process.

Link: http://lkml.kernel.org/r/20190822173157.1569-1-echron@arista.com
Signed-off-by: Edward Chron <echron@arista.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg, oom: don't require __GFP_FS when invoking memcg OOM killer

Masoud Sharbiani noticed that commit 29ef680ae7c21110 ("memcg, oom: move
out_of_memory back to the charge path") broke memcg OOM called from
__xfs_filemap_fault() path.  It turned out that try_charge() is retrying
forever without making forward progress because mem_cgroup_oom(GFP_NOFS)
cannot invoke the OOM killer due to commit 3da88fb3bacfaa33 ("mm, oom:
move GFP_NOFS check to out_of_memory").

Allowing forced charge due to being unable to invoke memcg OOM killer will
lead to global OOM situation.  Also, just returning -ENOMEM will be risky
because OOM path is lost and some paths (e.g.  get_user_pages()) will leak
-ENOMEM.  Therefore, invoking memcg OOM killer (despite GFP_NOFS) will be
the only choice we can choose for now.

Until 29ef680ae7c21110, we were able to invoke memcg OOM killer when
GFP_KERNEL reclaim failed [1].  But since 29ef680ae7c21110, we need to
invoke memcg OOM killer when GFP_NOFS reclaim failed [2].  Although in the
past we did invoke memcg OOM killer for GFP_NOFS [3], we might get
pre-mature memcg OOM reports due to this patch.

[1]

 leaker invoked oom-killer: gfp_mask=0x6200ca(GFP_HIGHUSER_MOVABLE), nodemask=(null), order=0, oom_score_adj=0
 CPU: 0 PID: 2746 Comm: leaker Not tainted 4.18.0+ #19
 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 04/13/2018
 Call Trace:
  dump_stack+0x63/0x88
  dump_header+0x67/0x27a
  ? mem_cgroup_scan_tasks+0x91/0xf0
  oom_kill_process+0x210/0x410
  out_of_memory+0x10a/0x2c0
  mem_cgroup_out_of_memory+0x46/0x80
  mem_cgroup_oom_synchronize+0x2e4/0x310
  ? high_work_func+0x20/0x20
  pagefault_out_of_memory+0x31/0x76
  mm_fault_error+0x55/0x115
  ? handle_mm_fault+0xfd/0x220
  __do_page_fault+0x433/0x4e0
  do_page_fault+0x22/0x30
  ? page_fault+0x8/0x30
  page_fault+0x1e/0x30
 RIP: 0033:0x4009f0
 Code: 03 00 00 00 e8 71 fd ff ff 48 83 f8 ff 49 89 c6 74 74 48 89 c6 bf c0 0c 40 00 31 c0 e8 69 fd ff ff 45 85 ff 7e 21 31 c9 66 90 <41> 0f be 14 0e 01 d3 f7 c1 ff 0f 00 00 75 05 41 c6 04 0e 2a 48 83
 RSP: 002b:00007ffe29ae96f0 EFLAGS: 00010206
 RAX: 000000000000001b RBX: 0000000000000000 RCX: 0000000001ce1000
 RDX: 0000000000000000 RSI: 000000007fffffe5 RDI: 0000000000000000
 RBP: 000000000000000c R08: 0000000000000000 R09: 00007f94be09220d
 R10: 0000000000000002 R11: 0000000000000246 R12: 00000000000186a0
 R13: 0000000000000003 R14: 00007f949d845000 R15: 0000000002800000
 Task in /leaker killed as a result of limit of /leaker
 memory: usage 524288kB, limit 524288kB, failcnt 158965
 memory+swap: usage 0kB, limit 9007199254740988kB, failcnt 0
 kmem: usage 2016kB, limit 9007199254740988kB, failcnt 0
 Memory cgroup stats for /leaker: cache:844KB rss:521136KB rss_huge:0KB shmem:0KB mapped_file:0KB dirty:132KB writeback:0KB inactive_anon:0KB active_anon:521224KB inactive_file:1012KB active_file:8KB unevictable:0KB
 Memory cgroup out of memory: Kill process 2746 (leaker) score 998 or sacrifice child
 Killed process 2746 (leaker) total-vm:536704kB, anon-rss:521176kB, file-rss:1208kB, shmem-rss:0kB
 oom_reaper: reaped process 2746 (leaker), now anon-rss:0kB, file-rss:0kB, shmem-rss:0kB

[2]

 leaker invoked oom-killer: gfp_mask=0x600040(GFP_NOFS), nodemask=(null), order=0, oom_score_adj=0
 CPU: 1 PID: 2746 Comm: leaker Not tainted 4.18.0+ #20
 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 04/13/2018
 Call Trace:
  dump_stack+0x63/0x88
  dump_header+0x67/0x27a
  ? mem_cgroup_scan_tasks+0x91/0xf0
  oom_kill_process+0x210/0x410
  out_of_memory+0x109/0x2d0
  mem_cgroup_out_of_memory+0x46/0x80
  try_charge+0x58d/0x650
  ? __radix_tree_replace+0x81/0x100
  mem_cgroup_try_charge+0x7a/0x100
  __add_to_page_cache_locked+0x92/0x180
  add_to_page_cache_lru+0x4d/0xf0
  iomap_readpages_actor+0xde/0x1b0
  ? iomap_zero_range_actor+0x1d0/0x1d0
  iomap_apply+0xaf/0x130
  iomap_readpages+0x9f/0x150
  ? iomap_zero_range_actor+0x1d0/0x1d0
  xfs_vm_readpages+0x18/0x20 [xfs]
  read_pages+0x60/0x140
  __do_page_cache_readahead+0x193/0x1b0
  ondemand_readahead+0x16d/0x2c0
  page_cache_async_readahead+0x9a/0xd0
  filemap_fault+0x403/0x620
  ? alloc_set_pte+0x12c/0x540
  ? _cond_resched+0x14/0x30
  __xfs_filemap_fault+0x66/0x180 [xfs]
  xfs_filemap_fault+0x27/0x30 [xfs]
  __do_fault+0x19/0x40
  __handle_mm_fault+0x8e8/0xb60
  handle_mm_fault+0xfd/0x220
  __do_page_fault+0x238/0x4e0
  do_page_fault+0x22/0x30
  ? page_fault+0x8/0x30
  page_fault+0x1e/0x30
 RIP: 0033:0x4009f0
 Code: 03 00 00 00 e8 71 fd ff ff 48 83 f8 ff 49 89 c6 74 74 48 89 c6 bf c0 0c 40 00 31 c0 e8 69 fd ff ff 45 85 ff 7e 21 31 c9 66 90 <41> 0f be 14 0e 01 d3 f7 c1 ff 0f 00 00 75 05 41 c6 04 0e 2a 48 83
 RSP: 002b:00007ffda45c9290 EFLAGS: 00010206
 RAX: 000000000000001b RBX: 0000000000000000 RCX: 0000000001a1e000
 RDX: 0000000000000000 RSI: 000000007fffffe5 RDI: 0000000000000000
 RBP: 000000000000000c R08: 0000000000000000 R09: 00007f6d061ff20d
 R10: 0000000000000002 R11: 0000000000000246 R12: 00000000000186a0
 R13: 0000000000000003 R14: 00007f6ce59b2000 R15: 0000000002800000
 Task in /leaker killed as a result of limit of /leaker
 memory: usage 524288kB, limit 524288kB, failcnt 7221
 memory+swap: usage 0kB, limit 9007199254740988kB, failcnt 0
 kmem: usage 1944kB, limit 9007199254740988kB, failcnt 0
 Memory cgroup stats for /leaker: cache:3632KB rss:518232KB rss_huge:0KB shmem:0KB mapped_file:0KB dirty:0KB writeback:0KB inactive_anon:0KB active_anon:518408KB inactive_file:3908KB active_file:12KB unevictable:0KB
 Memory cgroup out of memory: Kill process 2746 (leaker) score 992 or sacrifice child
 Killed process 2746 (leaker) total-vm:536704kB, anon-rss:518264kB, file-rss:1188kB, shmem-rss:0kB
 oom_reaper: reaped process 2746 (leaker), now anon-rss:0kB, file-rss:0kB, shmem-rss:0kB

[3]

 leaker invoked oom-killer: gfp_mask=0x50, order=0, oom_score_adj=0
 leaker cpuset=/ mems_allowed=0
 CPU: 1 PID: 3206 Comm: leaker Not tainted 3.10.0-957.27.2.el7.x86_64 #1
 Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 04/13/2018
 Call Trace:
  [<ffffffffaf364147>] dump_stack+0x19/0x1b
  [<ffffffffaf35eb6a>] dump_header+0x90/0x229
  [<ffffffffaedbb456>] ? find_lock_task_mm+0x56/0xc0
  [<ffffffffaee32a38>] ? try_get_mem_cgroup_from_mm+0x28/0x60
  [<ffffffffaedbb904>] oom_kill_process+0x254/0x3d0
  [<ffffffffaee36c36>] mem_cgroup_oom_synchronize+0x546/0x570
  [<ffffffffaee360b0>] ? mem_cgroup_charge_common+0xc0/0xc0
  [<ffffffffaedbc194>] pagefault_out_of_memory+0x14/0x90
  [<ffffffffaf35d072>] mm_fault_error+0x6a/0x157
  [<ffffffffaf3717c8>] __do_page_fault+0x3c8/0x4f0
  [<ffffffffaf371925>] do_page_fault+0x35/0x90
  [<ffffffffaf36d768>] page_fault+0x28/0x30
 Task in /leaker killed as a result of limit of /leaker
 memory: usage 524288kB, limit 524288kB, failcnt 20628
 memory+swap: usage 524288kB, limit 9007199254740988kB, failcnt 0
 kmem: usage 0kB, limit 9007199254740988kB, failcnt 0
 Memory cgroup stats for /leaker: cache:840KB rss:523448KB rss_huge:0KB mapped_file:0KB swap:0KB inactive_anon:0KB active_anon:523448KB inactive_file:464KB active_file:376KB unevictable:0KB
 Memory cgroup out of memory: Kill process 3206 (leaker) score 970 or sacrifice child
 Killed process 3206 (leaker) total-vm:536692kB, anon-rss:523304kB, file-rss:412kB, shmem-rss:0kB

Bisected by Masoud Sharbiani.

Link: http://lkml.kernel.org/r/cbe54ed1-b6ba-a056-8899-2dc42526371d@i-love.sakura.ne.jp
Fixes: 3da88fb3bacfaa33 ("mm, oom: move GFP_NOFS check to out_of_memory") [necessary after 29ef680ae7c21110]
Signed-off-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Reported-by: Masoud Sharbiani <msharbiani@apple.com>
Tested-by: Masoud Sharbiani <msharbiani@apple.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Cc: <stable@vger.kernel.org> [4.19+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/oom_kill.c: add task UID to info message on an oom kill

In the event of an oom kill, useful information about the killed process
is printed to dmesg.  Users, especially system administrators, will find
it useful to immediately see the UID of the process.

We already print uid when dumping eligible tasks so it is not overly hard
to find that information in the oom report.  However this information is
unavailable when dumping of eligible tasks is disabled.

In the following example, abuse_the_ram is the name of a program that
attempts to iteratively allocate all available memory until it is stopped
by force.

Current message:

Out of memory: Killed process 35389 (abuse_the_ram)
total-vm:133718232kB, anon-rss:129624980kB, file-rss:0kB,
shmem-rss:0kB

Patched message:

Out of memory: Killed process 2739 (abuse_the_ram),
total-vm:133880028kB, anon-rss:129754836kB, file-rss:0kB,
shmem-rss:0kB, UID:0

[akpm@linux-foundation.org: s/UID %d/UID:%u/ in printk]
Link: http://lkml.kernel.org/r/1560362273-534-1-git-send-email-jsavitz@redhat.com
Signed-off-by: Joel Savitz <jsavitz@redhat.com>
Suggested-by: David Rientjes <rientjes@google.com>
Acked-by: Rafael Aquini <aquini@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>

mm/mempolicy.c: remove unnecessary nodemask check in kernel_migrate_pages()

1) task_nodes = cpuset_mems_allowed(current);
   -> cpuset_mems_allowed() guaranteed to return some non-empty
      subset of node_states[N_MEMORY].

2) nodes_and(*new, *new, task_nodes);
   -> after nodes_and(), the 'new' should be empty or appropriate
      nodemask(online node and with memory).

After 1) and 2), we could remove unnecessary check whether the 'new'
AND node_states[N_MEMORY] is empty.

Link: http://lkml.kernel.org/r/20190806023634.55356-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/compaction.c: remove unnecessary zone parameter in isolate_migratepages()

Like commit 40cacbcb3240 ("mm, compaction: remove unnecessary zone
parameter in some instances"), remove unnecessary zone parameter.

No functional change.

Link: http://lkml.kernel.org/r/20190806151616.21107-1-lpf.vector@gmail.com
Signed-off-by: Pengfei Li <lpf.vector@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Qian Cai <cai@lca.pw>
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: clear total_{migrate,free}_scanned before scanning a new zone

total_{migrate,free}_scanned will be added to COMPACTMIGRATE_SCANNED and
COMPACTFREE_SCANNED in compact_zone().  We should clear them before
scanning a new zone.  In the proc triggered compaction, we forgot clearing
them.

[laoar.shao@gmail.com: introduce a helper compact_zone_counters_init()]
Link: http://lkml.kernel.org/r/1563869295-25748-1-git-send-email-laoar.shao@gmail.com
[akpm@linux-foundation.org: expand compact_zone_counters_init() into its single callsite, per mhocko]
[vbabka@suse.cz: squash compact_zone() list_head init as well]
Link: http://lkml.kernel.org/r/1fb6f7da-f776-9e42-22f8-bbb79b030b98@suse.cz
[akpm@linux-foundation.org: kcompactd_do_work(): avoid unnecessary initialization of cc.zone]
Link: http://lkml.kernel.org/r/1563789275-9639-1-git-send-email-laoar.shao@gmail.com
Fixes: 7f354a548d1c ("mm, compaction: add vmstats for kcompactd work")
Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Cc: David Rientjes <rientjes@google.com>
Cc: Yafang Shao <shaoyafang@didiglobal.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

z3fold: fix memory leak in kmem cache

Currently there is a leak in init_z3fold_page() -- it allocates handles
from kmem cache even for headless pages, but then they are never used and
never freed, so eventually kmem cache may get exhausted.  This patch
provides a fix for that.

Link: http://lkml.kernel.org/r/20190917185352.44cf285d3ebd9e64548de5de@gmail.com
Signed-off-by: Vitaly Wool <vitalywool@gmail.com>
Reported-by: Markus Linnala <markus.linnala@gmail.com>
Tested-by: Markus Linnala <markus.linnala@gmail.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Henry Burns <henrywolfeburns@gmail.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: silence -Woverride-init/initializer-overrides

When compiling a kernel with W=1, there are several of those warnings due
to arm64 overriding a field on purpose.  Just disable those warnings for
both GCC and Clang of this file, so it will help dig "gems" hidden in the
W=1 warnings by reducing some noises.

mm/init-mm.c:39:2: warning: initializer overrides prior initialization
of this subobject [-Winitializer-overrides]
        INIT_MM_CONTEXT(init_mm)
        ^~~~~~~~~~~~~~~~~~~~~~~~
./arch/arm64/include/asm/mmu.h:133:9: note: expanded from macro
'INIT_MM_CONTEXT'
        .pgd = init_pg_dir,
               ^~~~~~~~~~~
mm/init-mm.c:30:10: note: previous initialization is here
        .pgd            = swapper_pg_dir,
                          ^~~~~~~~~~~~~~

Note: there is a side project trying to support explicitly allowing
specific initializer overrides in Clang, but there is no guarantee it
will happen or not.

https://github.com/ClangBuiltLinux/linux/issues/639

Link: http://lkml.kernel.org/r/1566920867-27453-1-git-send-email-cai@lca.pw
Signed-off-by: Qian Cai <cai@lca.pw>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: use CPU_BITS_NONE to initialize init_mm.cpu_bitmask

Replace open-coded bitmap array initialization of init_mm.cpu_bitmask with
neat CPU_BITS_NONE macro.

And, since init_mm.cpu_bitmask is statically set to zero, there is no way
to clear it again in start_kernel().

Link: http://lkml.kernel.org/r/1565703815-8584-1-git-send-email-rppt@linux.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/vmalloc.c: move 'area->pages' after if statement

If !area->pages statement is true where memory allocation fails, area is
freed.

In this case 'area->pages = pages' should not executed.  So move
'area->pages = pages' after if statement.

[akpm@linux-foundation.org: give area->pages the same treatment]
Link: http://lkml.kernel.org/r/20190830035716.GA190684@LGEARND20B15
Signed-off-by: Austin Kim <austindh.kim@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Uladzislau Rezki (Sony) <urezki@gmail.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Roman Penyaev <rpenyaev@suse.de>
Cc: Rick Edgecombe <rick.p.edgecombe@intel.com>
Cc: Mike Rapoport <rppt@linux.ibm.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/vmalloc: modify struct vmap_area to reduce its size

Objective
---------

The current implementation of struct vmap_area wasted space.

After applying this commit, sizeof(struct vmap_area) has been
reduced from 11 words to 8 words.

Description
-----------

1) Pack "subtree_max_size", "vm" and "purge_list".  This is no problem
   because

A) "subtree_max_size" is only used when vmap_area is in "free" tree

B) "vm" is only used when vmap_area is in "busy" tree

C) "purge_list" is only used when vmap_area is in vmap_purge_list

2) Eliminate "flags".

;Since only one flag VM_VM_AREA is being used, and the same thing can be
done by judging whether "vm" is NULL, then the "flags" can be eliminated.

Link: http://lkml.kernel.org/r/20190716152656.12255-3-lpf.vector@gmail.com
Signed-off-by: Pengfei Li <lpf.vector@gmail.com>
Suggested-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Reviewed-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sonymobile.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/vmalloc: do not keep unpurged areas in the busy tree

The busy tree can be quite big, even though the area is freed or unmapped
it still stays there until "purge" logic removes it.

1) Optimize and reduce the size of "busy" tree by removing a node from
   it right away as soon as user triggers free paths.  It is possible to
   do so, because the allocation is done using another augmented tree.

The vmalloc test driver shows the difference, for example the
"fix_size_alloc_test" is ~11% better comparing with default configuration:

sudo ./test_vmalloc.sh performance

<default>
Summary: fix_size_alloc_test loops: 1000000 avg: 993985 usec
Summary: full_fit_alloc_test loops: 1000000 avg: 973554 usec
Summary: long_busy_list_alloc_test loops: 1000000 avg: 12617652 usec
<default>

<this patch>
Summary: fix_size_alloc_test loops: 1000000 avg: 882263 usec
Summary: full_fit_alloc_test loops: 1000000 avg: 973407 usec
Summary: long_busy_list_alloc_test loops: 1000000 avg: 12593929 usec
<this patch>

2) Since the busy tree now contains allocated areas only and does not
   interfere with lazily free nodes, introduce the new function
   show_purge_info() that dumps "unpurged" areas that is propagated
   through "/proc/vmallocinfo".

3) Eliminate VM_LAZY_FREE flag.

Link: http://lkml.kernel.org/r/20190716152656.12255-2-lpf.vector@gmail.com
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Signed-off-by: Pengfei Li <lpf.vector@gmail.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sonymobile.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/sparse.c: remove NULL check in clear_hwpoisoned_pages()

There is no possibility for memmap to be NULL in the current codebase.

This check was added in commit 95a4774d055c ("memory-hotplug: update
mce_bad_pages when removing the memory") where memmap was originally
inited to NULL, and only conditionally given a value.

The code that could have passed a NULL has been removed by commit
ba72b4c8cf60 ("mm/sparsemem: support sub-section hotplug"), so there is no
longer a possibility that memmap can be NULL.

Link: http://lkml.kernel.org/r/20190829035151.20975-1-alastair@d-silva.org
Signed-off-by: Alastair D'Silva <alastair@d-silva.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Wei Yang <richard.weiyang@gmail.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Baoquan He <bhe@redhat.com>
Cc: Balbir Singh <bsingharora@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/sparse.c: don't manually decrement num_poisoned_pages

Use the function written to do it instead.

Link: http://lkml.kernel.org/r/20190827053656.32191-2-alastair@au1.ibm.com
Signed-off-by: Alastair D'Silva <alastair@d-silva.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Reviewed-by: Wei Yang <richardw.yang@linux.intel.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/sparse.c: use __nr_to_section(section_nr) to get mem_section

__pfn_to_section is defined as __nr_to_section(pfn_to_section_nr(pfn)).

Since we already get section_nr, it is not necessary to get mem_section
from start_pfn. By doing so, we reduce one redundant operation.

Link: http://lkml.kernel.org/r/20190809010242.29797-1-richardw.yang@linux.intel.com
Signed-off-by: Wei Yang <richardw.yang@linux.intel.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Tested-by: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/sparse.c: fix ALIGN() without power of 2 in sparse_buffer_alloc()

The size argument passed into sparse_buffer_alloc() has already been
aligned with PAGE_SIZE or PMD_SIZE.

If the size after aligned is not power of 2 (e.g.  0x480000), the
PTR_ALIGN() will return wrong value.  Use roundup to round sparsemap_buf
up to next multiple of size.

Link: http://lkml.kernel.org/r/20190705114826.28586-1-lecopzer.chen@mediatek.com
Signed-off-by: Lecopzer Chen <lecopzer.chen@mediatek.com>
Signed-off-by: Mark-PK Tsai <Mark-PK.Tsai@mediatek.com>
Cc: YJ Chiang <yj.chiang@mediatek.com>
Cc: Lecopzer Chen <lecopzer.chen@mediatek.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/sparse.c: fix memory leak of sparsemap_buf in aligned memory

sparse_buffer_alloc(xsize) gets the size of memory from sparsemap_buf
after being aligned with the size.  However, the size is at least
PAGE_ALIGN(sizeof(struct page) * PAGES_PER_SECTION) and usually larger
than PAGE_SIZE.

Also, sparse_buffer_fini() only frees memory between sparsemap_buf and
sparsemap_buf_end, since sparsemap_buf may be changed by PTR_ALIGN()
first, the aligned space before sparsemap_buf is wasted and no one will
touch it.

In our ARM32 platform (without SPARSEMEM_VMEMMAP)
  Sparse_buffer_init
    Reserve d359c000 - d3e9c000 (9M)
  Sparse_buffer_alloc
    Alloc   d3a00000 - d3E80000 (4.5M)
  Sparse_buffer_fini
    Free    d3e80000 - d3e9c000 (~=100k)
 The reserved memory between d359c000 - d3a00000 (~=4.4M) is unfreed.

In ARM64 platform (with SPARSEMEM_VMEMMAP)

  sparse_buffer_init
    Reserve ffffffc07d623000 - ffffffc07f623000 (32M)
  Sparse_buffer_alloc
    Alloc   ffffffc07d800000 - ffffffc07f600000 (30M)
  Sparse_buffer_fini
    Free    ffffffc07f600000 - ffffffc07f623000 (140K)
 The reserved memory between ffffffc07d623000 - ffffffc07d800000
 (~=1.9M) is unfreed.

Let's explicit free redundant aligned memory.

[arnd@arndb.de: mark sparse_buffer_free as __meminit]
Link: http://lkml.kernel.org/r/20190709185528.3251709-1-arnd@arndb.de
Link: http://lkml.kernel.org/r/20190705114730.28534-1-lecopzer.chen@mediatek.com
Signed-off-by: Lecopzer Chen <lecopzer.chen@mediatek.com>
Signed-off-by: Mark-PK Tsai <Mark-PK.Tsai@mediatek.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: YJ Chiang <yj.chiang@mediatek.com>
Cc: Lecopzer Chen <lecopzer.chen@mediatek.com>
Cc: Pavel Tatashin <pasha.tatashin@oracle.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug.c: s/is/if

Correct typo in comment.

Link: http://lkml.kernel.org/r/1568233954-3913-1-git-send-email-jrdr.linux@gmail.com
Signed-off-by: Souptick Joarder <jrdr.linux@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug: online_pages cannot be 0 in online_pages()

walk_system_ram_range() will fail with -EINVAL in case
online_pages_range() was never called (== no resource applicable in the
range).  Otherwise, we will always call online_pages_range() with nr_pages
> 0 and, therefore, have online_pages > 0.

Remove that special handling.

Link: http://lkml.kernel.org/r/20190814154109.3448-6-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Arun KS <arunks@codeaurora.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Wei Yang <richardw.yang@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug: make sure the pfn is aligned to the order when onlining

Commit a9cd410a3d29 ("mm/page_alloc.c: memory hotplug: free pages as
higher order") assumed that any PFN we get via memory resources is aligned
to to MAX_ORDER - 1, I am not convinced that is always true.  Let's play
safe, check the alignment and fallback to single pages.

akpm: warn in this situation so we get to find out if and why this ever
occurs.

[akpm@linux-foundation.org: add WARN_ON_ONCE()]
Link: http://lkml.kernel.org/r/20190814154109.3448-5-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Cc: Arun KS <arunks@codeaurora.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Wei Yang <richardw.yang@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug: simplify online_pages_range()

online_pages always corresponds to nr_pages.  Simplify the code, getting
rid of online_pages_blocks().  Add some comments.

Link: http://lkml.kernel.org/r/20190814154109.3448-4-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Arun KS <arunks@codeaurora.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Wei Yang <richardw.yang@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug: drop PageReserved() check in online_pages_range()

move_pfn_range_to_zone() will set all pages to PG_reserved via
memmap_init_zone().  The only way a page could no longer be reserved would
be if a MEM_GOING_ONLINE notifier would clear PG_reserved - which is not
done (the online_page callback is used for that purpose by e.g., Hyper-V
instead).  walk_system_ram_range() will never call online_pages_range()
with duplicate PFNs, so drop the PageReserved() check.

This seems to be a leftover from ancient times where the memmap was
initialized when adding memory and we wanted to check for already onlined
memory.

Link: http://lkml.kernel.org/r/20190814154109.3448-3-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Arun KS <arunks@codeaurora.org>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Wei Yang <richardw.yang@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug.c: use PFN_UP / PFN_DOWN in walk_system_ram_range()

Patch series "mm/memory_hotplug: online_pages() cleanups", v2.

Some cleanups (+ one fix for a special case) in the context of
online_pages().

This patch (of 5):

This makes it clearer that we will never call func() with duplicate PFNs
in case we have multiple sub-page memory resources.  All unaligned parts
of PFNs are completely discarded.

Link: http://lkml.kernel.org/r/20190814154109.3448-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Wei Yang <richardw.yang@linux.intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Borislav Petkov <bp@suse.de>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Arun KS <arunks@codeaurora.org>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug.c: prevent memory leak when reusing pgdat

When offlining a node in try_offline_node(), pgdat is not released.  So
that pgdat could be reused in hotadd_new_pgdat().  While we reallocate
pgdat->per_cpu_nodestats if this pgdat is reused.

This patch prevents the memory leak by just allocating per_cpu_nodestats
when it is a new pgdat.

Link: http://lkml.kernel.org/r/20190813020608.10194-1-richardw.yang@linux.intel.com
Signed-off-by: Wei Yang <richardw.yang@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <OSalvador@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/base/memory.c: don't store end_section_nr in memory blocks

Each memory block spans the same amount of sections/pages/bytes.  The size
is determined before the first memory block is created.  No need to store
what we can easily calculate - and the calculations even look simpler now.

Michal brought up the idea of variable-sized memory blocks.  However, if
we ever implement something like this, we will need an API compatibility
switch and reworks at various places (most code assumes a fixed memory
block size).  So let's cleanup what we have right now.

While at it, fix the variable naming in register_mem_sect_under_node() -
we no longer talk about a single section.

Link: http://lkml.kernel.org/r/20190809110200.2746-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

driver/base/memory.c: validate memory block size early

Let's validate the memory block size early, when initializing the memory
device infrastructure.  Fail hard in case the value is not suitable.

As nobody checks the return value of memory_dev_init(), turn it into a
void function and fail with a panic in all scenarios instead.  Otherwise,
we'll crash later during boot when core/drivers expect that the memory
device infrastructure (including memory_block_size_bytes()) works as
expected.

I think long term, we should move the whole memory block size
configuration (set_memory_block_size_order() and
memory_block_size_bytes()) into drivers/base/memory.c.

Link: http://lkml.kernel.org/r/20190806090142.22709-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/base/memory.c: fixup documentation of removable/phys_index/block_size_bytes

Let's rephrase to memory block terminology and add some further
clarifications.

Link: http://lkml.kernel.org/r/20190806080826.5963-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/base/node.c: simplify unregister_memory_block_under_nodes()

We don't allow to offline memory block devices that belong to multiple
numa nodes.  Therefore, such devices can never get removed.  It is
sufficient to process a single node when removing the memory block.  No
need to iterate over each and every PFN.

We already have the nid stored for each memory block.  Make sure that the
nid always has a sane value.

Please note that checking for node_online(nid) is not required.  If we
would have a memory block belonging to a node that is no longer offline,
then we would have a BUG in the node offlining code.

Link: http://lkml.kernel.org/r/20190719135244.15242-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug: remove move_pfn_range()

Let's remove this indirection.  We need the zone in the caller either way,
so let's just detect it there.  Add some documentation for
move_pfn_range_to_zone() instead.

[akpm@linux-foundation.org: restore newline, per David]
Link: http://lkml.kernel.org/r/20190724142324.3686-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: do not hash address in print_bad_pte()

Using %px to show the actual address in print_bad_pte()
to help us to debug issue.

Link: http://lkml.kernel.org/r/20190831011816.141002-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.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: consolidate pgtable_cache_init() and pgd_cache_init()

Both pgtable_cache_init() and pgd_cache_init() are used to initialize kmem
cache for page table allocations on several architectures that do not use
PAGE_SIZE tables for one or more levels of the page table hierarchy.

Most architectures do not implement these functions and use __weak default
NOP implementation of pgd_cache_init().  Since there is no such default
for pgtable_cache_init(), its empty stub is duplicated among most
architectures.

Rename the definitions of pgd_cache_init() to pgtable_cache_init() and
drop empty stubs of pgtable_cache_init().

Link: http://lkml.kernel.org/r/1566457046-22637-1-git-send-email-rppt@linux.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Will Deacon <will@kernel.org> [arm64]
Acked-by: Thomas Gleixner <tglx@linutronix.de> [x86]
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

microblaze: switch to generic version of pte allocation

The microblaze implementation of pte_alloc_one() has a provision to
allocated PTEs from high memory, but neither CONFIG_HIGHPTE nor pte_map*()
versions for suitable for HIGHPTE are defined.

Except that, microblaze version of pte_alloc_one() is identical to the
generic one as well as the implementations of pte_free() and
pte_free_kernel().

Switch microblaze to use the generic versions of these functions.  Also
remove pte_free_slow() that is not referenced anywhere in the code.

Link: http://lkml.kernel.org/r/1565690952-32158-1-git-send-email-rppt@linux.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Cc: Michal Simek <monstr@monstr.eu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sh: switch to generic version of pte allocation

The sh implementation pte_alloc_one(), pte_alloc_one_kernel(),
pte_free_kernel() and pte_free() is identical to the generic except of
lack of __GFP_ACCOUNT for the user PTEs allocation.

Switch sh to use generic version of these functions.

Link: http://lkml.kernel.org/r/1565250728-21721-4-git-send-email-rppt@linux.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ia64: switch to generic version of pte allocation

The ia64 implementation pte_alloc_one(), pte_alloc_one_kernel(),
pte_free_kernel() and pte_free() is identical to the generic except of
lack of __GFP_ACCOUNT for the user PTEs allocation.

Switch ia64 to use generic version of these functions.

Link: http://lkml.kernel.org/r/1565250728-21721-3-git-send-email-rppt@linux.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove quicklist page table caches

Patch series "mm: remove quicklist page table caches".

A while ago Nicholas proposed to remove quicklist page table caches [1].

I've rebased his patch on the curren upstream and switched ia64 and sh to
use generic versions of PTE allocation.

[1] https://lore.kernel.org/linux-mm/20190711030339.20892-1-npiggin@gmail.com

This patch (of 3):

Remove page table allocator "quicklists".  These have been around for a
long time, but have not got much traction in the last decade and are only
used on ia64 and sh architectures.

The numbers in the initial commit look interesting but probably don't
apply anymore.  If anybody wants to resurrect this it's in the git
history, but it's unhelpful to have this code and divergent allocator
behaviour for minor archs.

Also it might be better to instead make more general improvements to page
allocator if this is still so slow.

Link: http://lkml.kernel.org/r/1565250728-21721-2-git-send-email-rppt@linux.ibm.com
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Signed-off-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: release the spinlock on zap_pte_range

In our testing (camera recording), Miguel and Wei found
unmap_page_range() takes above 6ms with preemption disabled easily.
When I see that, the reason is it holds page table spinlock during
entire 512 page operation in a PMD.  6.2ms is never trivial for user
experince if RT task couldn't run in the time because it could make
frame drop or glitch audio problem.

I had a time to benchmark it via adding some trace_printk hooks between
pte_offset_map_lock and pte_unmap_unlock in zap_pte_range.  The testing
device is 2018 premium mobile device.

I can get 2ms delay rather easily to release 2M(ie, 512 pages) when the
task runs on little core even though it doesn't have any IPI and LRU
lock contention.  It's already too heavy.

If I remove activate_page, 35-40% overhead of zap_pte_range is gone so
most of overhead(about 0.7ms) comes from activate_page via
mark_page_accessed.  Thus, if there are LRU contention, that 0.7ms could
accumulate up to several ms.

So this patch adds a check for need_resched() in the loop, and a
preemption point if necessary.

Link: http://lkml.kernel.org/r/20190731061440.GC155569@google.com
Signed-off-by: Minchan Kim <minchan@kernel.org>
Reported-by: Miguel de Dios <migueldedios@google.com>
Reported-by: Wei Wang <wvw@google.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
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: remove redundant assignment of entry

Since ptent will not be changed after previous assignment of entry, it is
not necessary to do the assignment again.

Link: http://lkml.kernel.org/r/20190708082740.21111-1-richardw.yang@linux.intel.com
Signed-off-by: Wei Yang <richardw.yang@linux.intel.com>
Acked-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Will Deacon <will@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

net/xdp: convert put_page() to put_user_page*()

For pages that were retained via get_user_pages*(), release those pages
via the new put_user_page*() routines, instead of via put_page() or
release_pages().

This is part a tree-wide conversion, as described in fc1d8e7cca2d ("mm:
introduce put_user_page*(), placeholder versions").

Link: http://lkml.kernel.org/r/20190724044537.10458-4-jhubbard@nvidia.com
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Acked-by: Björn Töpel <bjorn.topel@intel.com>
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Magnus Karlsson <magnus.karlsson@intel.com>
Cc: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/gpu/drm/via: convert put_page() to put_user_page*()

For pages that were retained via get_user_pages*(), release those pages
via the new put_user_page*() routines, instead of via put_page() or
release_pages().

This is part a tree-wide conversion, as described in fc1d8e7cca2d ("mm:
introduce put_user_page*(), placeholder versions").

Also reverse the order of a comparison, in order to placate checkpatch.pl.

Link: http://lkml.kernel.org/r/20190724044537.10458-3-jhubbard@nvidia.com
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Daniel Vetter <daniel@ffwll.ch>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/gup: add make_dirty arg to put_user_pages_dirty_lock()

[11~From: John Hubbard <jhubbard@nvidia.com>
Subject: mm/gup: add make_dirty arg to put_user_pages_dirty_lock()

Patch series "mm/gup: add make_dirty arg to put_user_pages_dirty_lock()",
v3.

There are about 50+ patches in my tree [2], and I'll be sending out the
remaining ones in a few more groups:

* The block/bio related changes (Jerome mostly wrote those, but I've had
  to move stuff around extensively, and add a little code)

* mm/ changes

* other subsystem patches

* an RFC that shows the current state of the tracking patch set.  That
  can only be applied after all call sites are converted, but it's good to
  get an early look at it.

This is part a tree-wide conversion, as described in fc1d8e7cca2d ("mm:
introduce put_user_page*(), placeholder versions").

This patch (of 3):

Provide more capable variation of put_user_pages_dirty_lock(), and delete
put_user_pages_dirty().  This is based on the following:

1.  Lots of call sites become simpler if a bool is passed into
   put_user_page*(), instead of making the call site choose which
   put_user_page*() variant to call.

2.  Christoph Hellwig's observation that set_page_dirty_lock() is
   usually correct, and set_page_dirty() is usually a bug, or at least
   questionable, within a put_user_page*() calling chain.

This leads to the following API choices:

    * put_user_pages_dirty_lock(page, npages, make_dirty)

    * There is no put_user_pages_dirty(). You have to
      hand code that, in the rare case that it's
      required.

[jhubbard@nvidia.com: remove unused variable in siw_free_plist()]
Link: http://lkml.kernel.org/r/20190729074306.10368-1-jhubbard@nvidia.com
Link: http://lkml.kernel.org/r/20190724044537.10458-2-jhubbard@nvidia.com
Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jason Gunthorpe <jgg@ziepe.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: vmscan: do not share cgroup iteration between reclaimers

One of our services observed a high rate of cgroup OOM kills in the
presence of large amounts of clean cache.  Debugging showed that the
culprit is the shared cgroup iteration in page reclaim.

Under high allocation concurrency, multiple threads enter reclaim at the
same time.  Fearing overreclaim when we first switched from the single
global LRU to cgrouped LRU lists, we introduced a shared iteration state
for reclaim invocations - whether 1 or 20 reclaimers are active
concurrently, we only walk the cgroup tree once: the 1st reclaimer
reclaims the first cgroup, the second the second one etc.  With more
reclaimers than cgroups, we start another walk from the top.

This sounded reasonable at the time, but the problem is that reclaim
concurrency doesn't scale with allocation concurrency.  As reclaim
concurrency increases, the amount of memory individual reclaimers get to
scan gets smaller and smaller.  Individual reclaimers may only see one
cgroup per cycle, and that may not have much reclaimable memory.  We see
individual reclaimers declare OOM when there is plenty of reclaimable
memory available in cgroups they didn't visit.

This patch does away with the shared iterator, and every reclaimer is
allowed to scan the full cgroup tree and see all of reclaimable memory,
just like it would on a non-cgrouped system.  This way, when OOM is
declared, we know that the reclaimer actually had a chance.

To still maintain fairness in reclaim pressure, disallow cgroup reclaim
from bailing out of the tree walk early.  Kswapd and regular direct
reclaim already don't bail, so it's not clear why limit reclaim would have
to, especially since it only walks subtrees to begin with.

This change completely eliminates the OOM kills on our service, while
showing no signs of overreclaim - no increased scan rates, %sys time, or
abrupt free memory spikes.  I tested across 100 machines that have 64G of
RAM and host about 300 cgroups each.

[ It's possible overreclaim never was a *practical* issue to begin
  with - it was simply a concern we had on the mailing lists at the
  time, with no real data to back it up. But we have also added more
  bail-out conditions deeper inside reclaim (e.g. the proportional
  exit in shrink_node_memcg) since. Regardless, now we have data that
  suggests full walks are more reliable and scale just fine. ]

Link: http://lkml.kernel.org/r/20190812192316.13615-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.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: memcontrol: switch to rcu protection in drain_all_stock()

Commit 72f0184c8a00 ("mm, memcg: remove hotplug locking from try_charge")
introduced css_tryget()/css_put() calls in drain_all_stock(), which are
supposed to protect the target memory cgroup from being released during
the mem_cgroup_is_descendant() call.

However, it's not completely safe.  In theory, memcg can go away between
reading stock->cached pointer and calling css_tryget().

This can happen if drain_all_stock() races with drain_local_stock()
performed on the remote cpu as a result of a work, scheduled by the
previous invocation of drain_all_stock().

The race is a bit theoretical and there are few chances to trigger it, but
the current code looks a bit confusing, so it makes sense to fix it
anyway.  The code looks like as if css_tryget() and css_put() are used to
protect stocks drainage.  It's not necessary because stocked pages are
holding references to the cached cgroup.  And it obviously won't work for
works, scheduled on other cpus.

So, let's read the stock->cached pointer and evaluate the memory cgroup
inside a rcu read section, and get rid of css_tryget()/css_put() calls.

Link: http://lkml.kernel.org/r/20190802192241.3253165-1-guro@fb.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hillf Danton <hdanton@sina.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, memcg: throttle allocators when failing reclaim over memory.high

We're trying to use memory.high to limit workloads, but have found that
containment can frequently fail completely and cause OOM situations
outside of the cgroup.  This happens especially with swap space -- either
when none is configured, or swap is full.  These failures often also don't
have enough warning to allow one to react, whether for a human or for a
daemon monitoring PSI.

Here is output from a simple program showing how long it takes in usec
(column 2) to allocate a megabyte of anonymous memory (column 1) when a
cgroup is already beyond its memory high setting, and no swap is
available:

    [root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
    > --wait -t timeout 300 /root/mdf
    [...]
    95  1035
    96  1038
    97  1000
    98  1036
    99  1048
    100 1590
    101 1968
    102 1776
    103 1863
    104 1757
    105 1921
    106 1893
    107 1760
    108 1748
    109 1843
    110 1716
    111 1924
    112 1776
    113 1831
    114 1766
    115 1836
    116 1588
    117 1912
    118 1802
    119 1857
    120 1731
    [...]
    [System OOM in 2-3 seconds]

The delay does go up extremely marginally past the 100MB memory.high
threshold, as now we spend time scanning before returning to usermode, but
it's nowhere near enough to contain growth.  It also doesn't get worse the
more pages you have, since it only considers nr_pages.

The current situation goes against both the expectations of users of
memory.high, and our intentions as cgroup v2 developers.  In
cgroup-v2.txt, we claim that we will throttle and only under "extreme
conditions" will memory.high protection be breached.  Likewise, cgroup v2
users generally also expect that memory.high should throttle workloads as
they exceed their high threshold.  However, as seen above, this isn't
always how it works in practice -- even on banal setups like those with no
swap, or where swap has become exhausted, we can end up with memory.high
being breached and us having no weapons left in our arsenal to combat
runaway growth with, since reclaim is futile.

It's also hard for system monitoring software or users to tell how bad the
situation is, as "high" events for the memcg may in some cases be benign,
and in others be catastrophic.  The current status quo is that we fail
containment in a way that doesn't provide any advance warning that things
are about to go horribly wrong (for example, we are about to invoke the
kernel OOM killer).

This patch introduces explicit throttling when reclaim is failing to keep
memcg size contained at the memory.high setting.  It does so by applying
an exponential delay curve derived from the memcg's overage compared to
memory.high.  In the normal case where the memcg is either below or only
marginally over its memory.high setting, no throttling will be performed.

This composes well with system health monitoring and remediation, as these
allocator delays are factored into PSI's memory pressure calculations.
This both creates a mechanism system administrators or applications
consuming the PSI interface to trivially see that the memcg in question is
struggling and use that to make more reasonable decisions, and permits
them enough time to act.  Either of these can act with significantly more
nuance than that we can provide using the system OOM killer.

This is a similar idea to memory.oom_control in cgroup v1 which would put
the cgroup to sleep if the threshold was violated, but it's also
significantly improved as it results in visible memory pressure, and also
doesn't schedule indefinitely, which previously made tracing and other
introspection difficult (ie.  it's clamped at 2*HZ per allocation through
MEMCG_MAX_HIGH_DELAY_JIFFIES).

Contrast the previous results with a kernel with this patch:

    [root@ktst ~]# systemd-run -p MemoryHigh=100M -p MemorySwapMax=1 \
    > --wait -t timeout 300 /root/mdf
    [...]
    95  1002
    96  1000
    97  1002
    98  1003
    99  1000
    100 1043
    101 84724
    102 330628
    103 610511
    104 1016265
    105 1503969
    106 2391692
    107 2872061
    108 3248003
    109 4791904
    110 5759832
    111 6912509
    112 8127818
    113 9472203
    114 12287622
    115 12480079
    116 14144008
    117 15808029
    118 16384500
    119 16383242
    120 16384979
    [...]

As you can see, in the normal case, memory allocation takes around 1000
usec.  However, as we exceed our memory.high, things start to increase
exponentially, but fairly leniently at first.  Our first megabyte over
memory.high takes us 0.16 seconds, then the next is 0.46 seconds, then the
next is almost an entire second.  This gets worse until we reach our
eventual 2*HZ clamp per batch, resulting in 16 seconds per megabyte.
However, this is still making forward progress, so permits tracing or
further analysis with programs like GDB.

We use an exponential curve for our delay penalty for a few reasons:

1. We run mem_cgroup_handle_over_high to potentially do reclaim after
   we've already performed allocations, which means that temporarily
   going over memory.high by a small amount may be perfectly legitimate,
   even for compliant workloads. We don't want to unduly penalise such
   cases.
2. An exponential curve (as opposed to a static or linear delay) allows
   ramping up memory pressure stats more gradually, which can be useful
   to work out that you have set memory.high too low, without destroying
   application performance entirely.

This patch expands on earlier work by Johannes Weiner. Thanks!

[akpm@linux-foundation.org: fix max() warning]
[akpm@linux-foundation.org: fix __udivdi3 ref on 32-bit]
[akpm@linux-foundation.org: fix it even more]
[chris@chrisdown.name: fix 64-bit divide even more]
Link: http://lkml.kernel.org/r/20190723180700.GA29459@chrisdown.name
Signed-off-by: Chris Down <chris@chrisdown.name>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Roman Gushchin <guro@fb.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: page cache: store only head pages in i_pages

Transparent Huge Pages are currently stored in i_pages as pointers to
consecutive subpages.  This patch changes that to storing consecutive
pointers to the head page in preparation for storing huge pages more
efficiently in i_pages.

Large parts of this are "inspired" by Kirill's patch
https://lore.kernel.org/lkml/20170126115819.58875-2-kirill.shutemov@linux.intel.com/

Kirill and Huang Ying contributed several fixes.

[willy@infradead.org: use compound_nr, squish uninit-var warning]
Link: http://lkml.kernel.org/r/20190731210400.7419-1-willy@infradead.org
Signed-off-by: Matthew Wilcox <willy@infradead.org>
Acked-by: Jan Kara <jack@suse.cz>
Reviewed-by: Kirill Shutemov <kirill@shutemov.name>
Reviewed-by: Song Liu <songliubraving@fb.com>
Tested-by: Song Liu <songliubraving@fb.com>
Tested-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Tested-by: Qian Cai <cai@lca.pw>
Tested-by: Mikhail Gavrilov <mikhail.v.gavrilov@gmail.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Song Liu <songliubraving@fb.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap.c: rewrite mapping_needs_writeback in less fancy manner

This actually checks that writeback is needed or in progress.

Link: http://lkml.kernel.org/r/156378817069.1087.1302816672037672488.stgit@buzz
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap.c: don't initiate writeback if mapping has no dirty pages

Functions like filemap_write_and_wait_range() should do nothing if inode
has no dirty pages or pages currently under writeback.  But they anyway
construct struct writeback_control and this does some atomic operations if
CONFIG_CGROUP_WRITEBACK=y - on fast path it locks inode->i_lock and
updates state of writeback ownership, on slow path might be more work.
Current this path is safely avoided only when inode mapping has no pages.

For example generic_file_read_iter() calls filemap_write_and_wait_range()
at each O_DIRECT read - pretty hot path.

This patch skips starting new writeback if mapping has no dirty tags set.
If writeback is already in progress filemap_write_and_wait_range() will
wait for it.

Link: http://lkml.kernel.org/r/156378816804.1087.8607636317907921438.stgit@buzz
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>