mm, kfence: insert KFENCE hooks for SLUB

Inserts KFENCE hooks into the SLUB allocator.

To pass the originally requested size to KFENCE, add an argument
'orig_size' to slab_alloc*(). The additional argument is required to
preserve the requested original size for kmalloc() allocations, which
uses size classes (e.g. an allocation of 272 bytes will return an object
of size 512). Therefore, kmem_cache::size does not represent the
kmalloc-caller's requested size, and we must introduce the argument
'orig_size' to propagate the originally requested size to KFENCE.

Without the originally requested size, we would not be able to detect
out-of-bounds accesses for objects placed at the end of a KFENCE object
page if that object is not equal to the kmalloc-size class it was
bucketed into.

When KFENCE is disabled, there is no additional overhead, since
slab_alloc*() functions are __always_inline.

Link: https://lkml.kernel.org/r/20201103175841.3495947-6-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Signed-off-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Jann Horn <jannh@google.com>
Co-developed-by: Marco Elver <elver@google.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joern Engel <joern@purestorage.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: SeongJae Park <sjpark@amazon.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, kfence: insert KFENCE hooks for SLAB

Inserts KFENCE hooks into the SLAB allocator.

To pass the originally requested size to KFENCE, add an argument
'orig_size' to slab_alloc*(). The additional argument is required to
preserve the requested original size for kmalloc() allocations, which
uses size classes (e.g. an allocation of 272 bytes will return an object
of size 512). Therefore, kmem_cache::size does not represent the
kmalloc-caller's requested size, and we must introduce the argument
'orig_size' to propagate the originally requested size to KFENCE.

Without the originally requested size, we would not be able to detect
out-of-bounds accesses for objects placed at the end of a KFENCE object
page if that object is not equal to the kmalloc-size class it was
bucketed into.

When KFENCE is disabled, there is no additional overhead, since
slab_alloc*() functions are __always_inline.

Link: https://lkml.kernel.org/r/20201103175841.3495947-5-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Signed-off-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Co-developed-by: Marco Elver <elver@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jann Horn <jannh@google.com>
Cc: Joern Engel <joern@purestorage.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: SeongJae Park <sjpark@amazon.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kfence: use pt_regs to generate stack trace on faults

Instead of removing the fault handling portion of the stack trace based on
the fault handler's name, just use struct pt_regs directly.

Change kfence_handle_page_fault() to take a struct pt_regs, and plumb it
through to kfence_report_error() for out-of-bounds, use-after-free, or
invalid access errors, where pt_regs is used to generate the stack trace.

If the kernel is a DEBUG_KERNEL, also show registers for more information.

Link: https://lkml.kernel.org/r/20201105092133.2075331-1-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Suggested-by: Mark Rutland <mark.rutland@arm.com>
Acked-by: Mark Rutland <mark.rutland@arm.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jann Horn <jannh@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64, kfence: enable KFENCE for ARM64

Add architecture specific implementation details for KFENCE and enable
KFENCE for the arm64 architecture. In particular, this implements the
required interface in <asm/kfence.h>.

KFENCE requires that attributes for pages from its memory pool can
individually be set. Therefore, force the entire linear map to be mapped
at page granularity. Doing so may result in extra memory allocated for
page tables in case rodata=full is not set; however, currently
CONFIG_RODATA_FULL_DEFAULT_ENABLED=y is the default, and the common case
is therefore not affected by this change.

[elver@google.com: add missing copyright and description header]
Link: https://lkml.kernel.org/r/20210118092159.145934-3-elver@google.com
Link: https://lkml.kernel.org/r/20201103175841.3495947-4-elver@google.com
Signed-off-by: Alexander Potapenko <glider@google.com>
Signed-off-by: Marco Elver <elver@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Co-developed-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Jann Horn <jannh@google.com>
Reviewed-by: Mark Rutland <mark.rutland@arm.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joern Engel <joern@purestorage.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: SeongJae Park <sjpark@amazon.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

x86, kfence: enable KFENCE for x86

Add architecture specific implementation details for KFENCE and enable
KFENCE for the x86 architecture. In particular, this implements the
required interface in <asm/kfence.h> for setting up the pool and
providing helper functions for protecting and unprotecting pages.

For x86, we need to ensure that the pool uses 4K pages, which is done
using the set_memory_4k() helper function.

[elver@google.com: add missing copyright and description header]
Link: https://lkml.kernel.org/r/20210118092159.145934-2-elver@google.com
Link: https://lkml.kernel.org/r/20201103175841.3495947-3-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Signed-off-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Co-developed-by: Marco Elver <elver@google.com>
Reviewed-by: Jann Horn <jannh@google.com>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Joern Engel <joern@purestorage.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: SeongJae Park <sjpark@amazon.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add Kernel Electric-Fence infrastructure

Patch series "KFENCE: A low-overhead sampling-based memory safety error detector", v7.

This adds the Kernel Electric-Fence (KFENCE) infrastructure. KFENCE is a
low-overhead sampling-based memory safety error detector of heap
use-after-free, invalid-free, and out-of-bounds access errors.  This
series enables KFENCE for the x86 and arm64 architectures, and adds
KFENCE hooks to the SLAB and SLUB allocators.

KFENCE is designed to be enabled in production kernels, and has near
zero performance overhead. Compared to KASAN, KFENCE trades performance
for precision. The main motivation behind KFENCE's design, is that with
enough total uptime KFENCE will detect bugs in code paths not typically
exercised by non-production test workloads. One way to quickly achieve a
large enough total uptime is when the tool is deployed across a large
fleet of machines.

KFENCE objects each reside on a dedicated page, at either the left or
right page boundaries. The pages to the left and right of the object
page are "guard pages", whose attributes are changed to a protected
state, and cause page faults on any attempted access to them. Such page
faults are then intercepted by KFENCE, which handles the fault
gracefully by reporting a memory access error.

Guarded allocations are set up based on a sample interval (can be set
via kfence.sample_interval). After expiration of the sample interval,
the next allocation through the main allocator (SLAB or SLUB) returns a
guarded allocation from the KFENCE object pool. At this point, the timer
is reset, and the next allocation is set up after the expiration of the
interval.

To enable/disable a KFENCE allocation through the main allocator's
fast-path without overhead, KFENCE relies on static branches via the
static keys infrastructure. The static branch is toggled to redirect the
allocation to KFENCE.

The KFENCE memory pool is of fixed size, and if the pool is exhausted no
further KFENCE allocations occur. The default config is conservative
with only 255 objects, resulting in a pool size of 2 MiB (with 4 KiB
pages).

We have verified by running synthetic benchmarks (sysbench I/O,
hackbench) and production server-workload benchmarks that a kernel with
KFENCE (using sample intervals 100-500ms) is performance-neutral
compared to a non-KFENCE baseline kernel.

KFENCE is inspired by GWP-ASan [1], a userspace tool with similar
properties. The name "KFENCE" is a homage to the Electric Fence Malloc
Debugger [2].

For more details, see Documentation/dev-tools/kfence.rst added in the
series -- also viewable here:

https://raw.githubusercontent.com/google/kasan/kfence/Documentation/dev-tools/kfence.rst

[1] http://llvm.org/docs/GwpAsan.html
[2] https://linux.die.net/man/3/efence

This patch (of 9):

This adds the Kernel Electric-Fence (KFENCE) infrastructure. KFENCE is a
low-overhead sampling-based memory safety error detector of heap
use-after-free, invalid-free, and out-of-bounds access errors.

KFENCE is designed to be enabled in production kernels, and has near
zero performance overhead. Compared to KASAN, KFENCE trades performance
for precision. The main motivation behind KFENCE's design, is that with
enough total uptime KFENCE will detect bugs in code paths not typically
exercised by non-production test workloads. One way to quickly achieve a
large enough total uptime is when the tool is deployed across a large
fleet of machines.

KFENCE objects each reside on a dedicated page, at either the left or
right page boundaries. The pages to the left and right of the object
page are "guard pages", whose attributes are changed to a protected
state, and cause page faults on any attempted access to them. Such page
faults are then intercepted by KFENCE, which handles the fault
gracefully by reporting a memory access error. To detect out-of-bounds
writes to memory within the object's page itself, KFENCE also uses
pattern-based redzones. The following figure illustrates the page
layout:

  ---+-----------+-----------+-----------+-----------+-----------+---
     | xxxxxxxxx | O :       | xxxxxxxxx |       : O | xxxxxxxxx |
     | xxxxxxxxx | B :       | xxxxxxxxx |       : B | xxxxxxxxx |
     | x GUARD x | J : RED-  | x GUARD x | RED-  : J | x GUARD x |
     | xxxxxxxxx | E :  ZONE | xxxxxxxxx |  ZONE : E | xxxxxxxxx |
     | xxxxxxxxx | C :       | xxxxxxxxx |       : C | xxxxxxxxx |
     | xxxxxxxxx | T :       | xxxxxxxxx |       : T | xxxxxxxxx |
  ---+-----------+-----------+-----------+-----------+-----------+---

Guarded allocations are set up based on a sample interval (can be set
via kfence.sample_interval). After expiration of the sample interval, a
guarded allocation from the KFENCE object pool is returned to the main
allocator (SLAB or SLUB). At this point, the timer is reset, and the
next allocation is set up after the expiration of the interval.

To enable/disable a KFENCE allocation through the main allocator's
fast-path without overhead, KFENCE relies on static branches via the
static keys infrastructure. The static branch is toggled to redirect the
allocation to KFENCE. To date, we have verified by running synthetic
benchmarks (sysbench I/O, hackbench) that a kernel compiled with KFENCE
is performance-neutral compared to the non-KFENCE baseline.

For more details, see Documentation/dev-tools/kfence.rst (added later in
the series).

[elver@google.com: fix parameter description for kfence_object_start()]
Link: https://lkml.kernel.org/r/20201106092149.GA2851373@elver.google.com
[elver@google.com: avoid stalling work queue task without allocations]
Link: https://lkml.kernel.org/r/CADYN=9J0DQhizAGB0-jz4HOBBh+05kMBXb4c0cXMS7Qi5NAJiw@mail.gmail.com
Link: https://lkml.kernel.org/r/20201110135320.3309507-1-elver@google.com
[elver@google.com: fix potential deadlock due to wake_up()]
Link: https://lkml.kernel.org/r/000000000000c0645805b7f982e4@google.com
Link: https://lkml.kernel.org/r/20210104130749.1768991-1-elver@google.com
[elver@google.com: add option to use KFENCE without static keys]
Link: https://lkml.kernel.org/r/20210111091544.3287013-1-elver@google.com
[elver@google.com: add missing copyright and description headers]
Link: https://lkml.kernel.org/r/20210118092159.145934-1-elver@google.com
Link: https://lkml.kernel.org/r/20201103175841.3495947-2-elver@google.com
Signed-off-by: Marco Elver <elver@google.com>
Signed-off-by: Alexander Potapenko <glider@google.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: SeongJae Park <sjpark@amazon.de>
Co-developed-by: Marco Elver <elver@google.com>
Reviewed-by: Jann Horn <jannh@google.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Andrey Konovalov <andreyknvl@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christopher Lameter <cl@linux.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Eric Dumazet <edumazet@google.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Hillf Danton <hdanton@sina.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Joern Engel <joern@purestorage.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/early_ioremap.c: use __func__ instead of function name

It is better to use __func__ instead of function name.

Link: https://lkml.kernel.org/r/1611385587-4209-1-git-send-email-stephenzhangzsd@gmail.com
Signed-off-by: Stephen Zhang <stephenzhangzsd@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/backing-dev.c: use might_alloc()

Now that my little helper has landed, use it more.  On top of the existing
check this also uses lockdep through the fs_reclaim annotations.

[akpm@linux-foundation.org: include linux/sched/mm.h]

Link: https://lkml.kernel.org/r/20210113135009.3606813-2-daniel.vetter@ffwll.ch
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/dmapool: use might_alloc()

Now that my little helper has landed, use it more.  On top of the existing
check this also uses lockdep through the fs_reclaim annotations.

Link: https://lkml.kernel.org/r/20210113135009.3606813-1-daniel.vetter@ffwll.ch
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: page-flags.h: Typo fix (It -> If)

The "If" was wrongly spelled as "It".

Link: https://lkml.kernel.org/r/1608959036-91409-1-git-send-email-guoren@kernel.org
Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Alexander Duyck <alexander.h.duyck@linux.intel.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Steven Price <steven.price@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc.c: use page_private() to access page->private

It's recommended to use helper macro page_private() to access the private
field of page.  Use such helper to eliminate direct access.

Link: https://lkml.kernel.org/r/20210203091857.20017-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zsmalloc: account the number of compacted pages correctly

There exists multiple path may do zram compaction concurrently.
1. auto-compaction triggered during memory reclaim
2. userspace utils write zram<id>/compaction node

So, multiple threads may call zs_shrinker_scan/zs_compact concurrently.
But pages_compacted is a per zsmalloc pool variable and modification
of the variable is not serialized(through under class->lock).
There are two issues here:
1. the pages_compacted may not equal to total number of pages
freed(due to concurrently add).
2. zs_shrinker_scan may not return the correct number of pages
freed(issued by current shrinker).

The fix is simple:
1. account the number of pages freed in zs_compact locally.
2. use actomic variable pages_compacted to accumulate total number.

Link: https://lkml.kernel.org/r/20210202122235.26885-1-wu-yan@tcl.com
Fixes: 860c707dca155a56 ("zsmalloc: account the number of compacted pages")
Signed-off-by: Rokudo Yan <wu-yan@tcl.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zsmalloc.c: convert to use kmem_cache_zalloc in cache_alloc_zspage()

We always memset the zspage allocated via cache_alloc_zspage.  So it's
more convenient to use kmem_cache_zalloc in cache_alloc_zspage than caller
do it manually.

Link: https://lkml.kernel.org/r/20210114120032.25885-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: set the sleep_mapped to true for zbud and z3fold

zpool driver adds a flag to indicate whether the zpool driver can enter an
atomic context after mapping.  This patch sets it true for z3fold and
zbud.

Link: https://lkml.kernel.org/r/1611035683-12732-3-git-send-email-tiantao6@hisilicon.com
Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Reviewed-by: Vitaly Wool <vitaly.wool@konsulko.com>
Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reported-by: Mike Galbraith <efault@gmx.de>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Barry Song <song.bao.hua@hisilicon.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/zswap: add the flag can_sleep_mapped

Patch series "Fix the compatibility of zsmalloc and zswap".

Patch #1 adds a flag to zpool, then zswap used to determine if zpool
drivers such as zbud/z3fold/zsmalloc will enter an atomic context after
mapping.

The difference between zbud/z3fold and zsmalloc is that zsmalloc requires
an atomic context that since its map function holds a preempt-disabled,
but zbud/z3fold don't require an atomic context.  So patch #2 sets flag
sleep_mapped to true indicating that zbud/z3fold can sleep after mapping.
zsmalloc didn't support sleep after mapping, so don't set that flag to
true.

This patch (of 2):

Add a flag to zpool, named is "can_sleep_mapped", and have it set true for
zbud/z3fold, not set this flag for zsmalloc, so its default value is
false.  Then zswap could go the current path if the flag is true; and if
it's false, copy data from src to a temporary buffer, then unmap the
handle, take the mutex, process the buffer instead of src to avoid
sleeping function called from atomic context.

[natechancellor@gmail.com: add return value in zswap_frontswap_load]
Link: https://lkml.kernel.org/r/20210121214804.926843-1-natechancellor@gmail.com
[tiantao6@hisilicon.com: fix potential memory leak]
Link: https://lkml.kernel.org/r/1611538365-51811-1-git-send-email-tiantao6@hisilicon.com
[colin.king@canonical.com: fix potential uninitialized pointer read on tmp]
Link: https://lkml.kernel.org/r/20210128141728.639030-1-colin.king@canonical.com
[tiantao6@hisilicon.com: fix variable 'entry' is uninitialized when used]
Link: https://lkml.kernel.org/r/1611223030-58346-1-git-send-email-tiantao6@hisilicon.comLink:
Link: https://lkml.kernel.org/r/1611035683-12732-2-git-send-email-tiantao6@hisilicon.com
Signed-off-by: Tian Tao <tiantao6@hisilicon.com>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Vitaly Wool <vitaly.wool@konsulko.com>
Acked-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Reported-by: Mike Galbraith <efault@gmx.de>
Cc: Barry Song <song.bao.hua@hisilicon.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: zswap: clean up confusing comment

Correct wording and change one duplicated word (it) to "it is".

Link: https://lkml.kernel.org/r/20201221042848.13980-1-rdunlap@infradead.org
Fixes: 0ab0abcf5115 ("mm/zswap: refactor the get/put routines")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Weijie Yang <weijie.yang@samsung.com>
Cc: Seth Jennings <sjennings@variantweb.net>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Vitaly Wool <vitaly.wool@konsulko.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/rmap: fix potential pte_unmap on an not mapped pte

For PMD-mapped page (usually THP), pvmw->pte is NULL.  For PTE-mapped THP,
pvmw->pte is mapped.  But for HugeTLB pages, pvmw->pte is not mapped and
set to the relevant page table entry.  So in page_vma_mapped_walk_done(),
we may do pte_unmap() for HugeTLB pte which is not mapped.  Fix this by
checking pvmw->page against PageHuge before trying to do pte_unmap().

Link: https://lkml.kernel.org/r/20210127093349.39081-1-linmiaohe@huawei.com
Fixes: ace71a19cec5 ("mm: introduce page_vma_mapped_walk()")
Signed-off-by: Hongxiang Lou <louhongxiang@huawei.com>
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com>
Cc: Kees Cook <keescook@chromium.org>
Cc: Nathan Chancellor <natechancellor@gmail.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Michel Lespinasse <walken@google.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Cc: Dmitry Safonov <0x7f454c46@gmail.com>
Cc: Brian Geffon <bgeffon@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/rmap: correct obsolete comment of page_get_anon_vma()

Since commit 746b18d421da ("mm: use refcounts for page_lock_anon_vma()"),
page_lock_anon_vma() is renamed to page_get_anon_vma() and converted to
return a refcount increased anon_vma.  But it forgot to change the
relevant comment.

Link: https://lkml.kernel.org/r/20210203093215.31990-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/rmap: use page_not_mapped in try_to_unmap()

page_mapcount_is_zero() calculates accurately how many mappings a hugepage
has in order to check against 0 only.  This is a waste of cpu time.  We
can do this via page_not_mapped() to save some possible atomic_read
cycles.  Remove the function page_mapcount_is_zero() as it's not used
anymore and move page_not_mapped() above try_to_unmap() to avoid
identifier undeclared compilation error.

Link: https://lkml.kernel.org/r/20210130084904.35307-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/rmap: fix obsolete comment in __page_check_anon_rmap()

Commit 21333b2b66b8 ("ksm: no debug in page_dup_rmap()") has reverted
page_dup_rmap() to an inline atomic_inc of mapcount.  So page_dup_rmap()
does not call __page_check_anon_rmap() anymore.

Link: https://lkml.kernel.org/r/20210128110209.50857-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/rmap: remove unneeded semicolon in page_not_mapped()

Remove extra semicolon without any functional change intended.

Link: https://lkml.kernel.org/r/20210127093425.39640-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/rmap: correct some obsolete comments of anon_vma

commit 2b575eb64f7a ("mm: convert anon_vma->lock to a mutex") changed
spinlock used to serialize access to vma list to mutex.  And further, the
commit 5a505085f043 ("mm/rmap: Convert the struct anon_vma::mutex to an
rwsem") converted the mutex to an rwsem for solving scalability problem.
So replace spinlock with rwsem to make comment uptodate.

Link: https://lkml.kernel.org/r/20210123072459.25903-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Rik van Riel <riel@surriel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/mlock: stop counting mlocked pages when none vma is found

There will be no vma satisfies addr < vm_end when find_vma() returns NULL.
Thus it's meaningless to traverse the vma list below because we can't
find any vma to count mlocked pages.  Stop counting mlocked pages in this
case to save some vma list traversal cycles.

Link: https://lkml.kernel.org/r/20210204110705.17586-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
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>

virtio-mem: check against mhp_get_pluggable_range() which memory we can hotplug

Right now, we only check against MAX_PHYSMEM_BITS - but turns out there
are more restrictions of which memory we can actually hotplug, especially
om arm64 or s390x once we support them: we might receive something like
-E2BIG or -ERANGE from add_memory_driver_managed(), stopping device
operation.

So, check right when initializing the device which memory we can add,
warning the user.  Try only adding actually pluggable ranges: in the worst
case, no memory provided by our device is pluggable.

In the usual case, we expect all device memory to be pluggable, and in
corner cases only some memory at the end of the device-managed memory
region to not be pluggable.

Link: https://lkml.kernel.org/r/1612149902-7867-5-git-send-email-anshuman.khandual@arm.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Cc: teawater <teawaterz@linux.alibaba.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Pankaj Gupta <pankaj.gupta@cloud.ionos.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

s390/mm: define arch_get_mappable_range()

This overrides arch_get_mappabble_range() on s390 platform which will be
used with recently added generic framework.  It modifies the existing
range check in vmem_add_mapping() using arch_get_mappable_range().  It
also adds a VM_BUG_ON() check that would ensure that mhp_range_allowed()
has already been called on the hotplug path.

Link: https://lkml.kernel.org/r/1612149902-7867-4-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Pankaj Gupta <pankaj.gupta@cloud.ionos.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: teawater <teawaterz@linux.alibaba.com>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arm64/mm: define arch_get_mappable_range()

This overrides arch_get_mappable_range() on arm64 platform which will be
used with recently added generic framework.  It drops
inside_linear_region() and subsequent check in arch_add_memory() which are
no longer required.  It also adds a VM_BUG_ON() check that would ensure
that mhp_range_allowed() has already been called.

Link: https://lkml.kernel.org/r/1612149902-7867-3-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Pankaj Gupta <pankaj.gupta@cloud.ionos.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: teawater <teawaterz@linux.alibaba.com>
Cc: Vasily Gorbik <gor@linux.ibm.com>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug: prevalidate the address range being added with platform

Patch series "mm/memory_hotplug: Pre-validate the address range with platform", v5.

This series adds a mechanism allowing platforms to weigh in and
prevalidate incoming address range before proceeding further with the
memory hotplug.  This helps prevent potential platform errors for the
given address range, down the hotplug call chain, which inevitably fails
the hotplug itself.

This mechanism was suggested by David Hildenbrand during another
discussion with respect to a memory hotplug fix on arm64 platform.

https://lore.kernel.org/linux-arm-kernel/1600332402-30123-1-git-send-email-anshuman.khandual@arm.com/

This mechanism focuses on the addressibility aspect and not [sub] section
alignment aspect.  Hence check_hotplug_memory_range() and check_pfn_span()
have been left unchanged.

This patch (of 4):

This introduces mhp_range_allowed() which can be called in various memory
hotplug paths to prevalidate the address range which is being added, with
the platform.  Then mhp_range_allowed() calls mhp_get_pluggable_range()
which provides applicable address range depending on whether linear
mapping is required or not.  For ranges that require linear mapping, it
calls a new arch callback arch_get_mappable_range() which the platform can
override.  So the new callback, in turn provides the platform an
opportunity to configure acceptable memory hotplug address ranges in case
there are constraints.

This mechanism will help prevent platform specific errors deep down during
hotplug calls.  This drops now redundant
check_hotplug_memory_addressable() check in __add_pages() but instead adds
a VM_BUG_ON() check which would ensure that the range has been validated
with mhp_range_allowed() earlier in the call chain.  Besides
mhp_get_pluggable_range() also can be used by potential memory hotplug
callers to avail the allowed physical range which would go through on a
given platform.

This does not really add any new range check in generic memory hotplug but
instead compensates for lost checks in arch_add_memory() where applicable
and check_hotplug_memory_addressable(), with unified mhp_range_allowed().

[akpm@linux-foundation.org: make pagemap_range() return -EINVAL when mhp_range_allowed() fails]

Link: https://lkml.kernel.org/r/1612149902-7867-1-git-send-email-anshuman.khandual@arm.com
Link: https://lkml.kernel.org/r/1612149902-7867-2-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Suggested-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Vasily Gorbik <gor@linux.ibm.com> # s390
Cc: Will Deacon <will@kernel.org>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Pankaj Gupta <pankaj.gupta@cloud.ionos.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: teawater <teawaterz@linux.alibaba.com>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Documentation: sysfs/memory: clarify some memory block device properties

In commit 53cdc1cb29e8 ("drivers/base/memory.c: indicate all memory blocks
as removable") we changed the output of the "removable" property of memory
devices to return "1" if and only if the kernel supports memory offlining.

Let's update documentation, stating that the interface is legacy.  Also
update documentation of the "state" property and "valid_zones" properties.

Link: https://lkml.kernel.org/r/20210201181347.13262-3-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: Dave Hansen <dave.hansen@intel.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Cc: Ilya Dryomov <idryomov@gmail.com>
Cc: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Cc: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/base/memory: don't store phys_device in memory blocks

No need to store the value for each and every memory block, as we can
easily query the value at runtime.  Reshuffle the members to optimize the
memory layout.  Also, let's clarify what the interface once was used for
and why it's legacy nowadays.

"phys_device" was used on s390x in older versions of lsmem[2]/chmem[3],
back when they were still part of s390x-tools.  They were later replaced
by the variants in linux-utils.  For example, RHEL6 and RHEL7 contain
lsmem/chmem from s390-utils.  RHEL8 switched to versions from util-linux
on s390x [4].

"phys_device" was added with sysfs support for memory hotplug in commit
3947be1969a9 ("[PATCH] memory hotplug: sysfs and add/remove functions") in
2005.  It always returned 0.

s390x started returning something != 0 on some setups (if sclp.rzm is set
by HW) in 2010 via commit 57b552ba0b2f ("memory hotplug/s390: set
phys_device").

For s390x, it allowed for identifying which memory block devices belong to
the same storage increment (RZM).  Only if all memory block devices
comprising a single storage increment were offline, the memory could
actually be removed in the hypervisor.

Since commit e5d709bb5fb7 ("s390/memory hotplug: provide
memory_block_size_bytes() function") in 2013 a memory block device spans
at least one storage increment - which is why the interface isn't really
helpful/used anymore (except by old lsmem/chmem tools).

There were once RFC patches to make use of "phys_device" in ACPI context;
however, the underlying problem could be solved using different interfaces
[1].

[1] https://patchwork.kernel.org/patch/2163871/
[2] https://github.com/ibm-s390-tools/s390-tools/blob/v2.1.0/zconf/lsmem
[3] https://github.com/ibm-s390-tools/s390-tools/blob/v2.1.0/zconf/chmem
[4] https://bugzilla.redhat.com/show_bug.cgi?id=1504134

Link: https://lkml.kernel.org/r/20210201181347.13262-2-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: Dave Hansen <dave.hansen@intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Cc: Ilya Dryomov <idryomov@gmail.com>
Cc: Vaibhav Jain <vaibhav@linux.ibm.com>
Cc: Tom Rix <trix@redhat.com>
Cc: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug: use helper function zone_end_pfn() to get end_pfn

Commit 108bcc96ef70 ("mm: add & use zone_end_pfn() and zone_spans_pfn()")
introduced the helper zone_end_pfn() to calculate the zone end pfn.  But
update_pgdat_span() forgot to use it.

Use this helper and rename local variable zone_end_pfn to end_pfn to avoid
a naming conflict with the existing zone_end_pfn().

Link: https://lkml.kernel.org/r/20210127093211.37714-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
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/memory_hotplug: MEMHP_MERGE_RESOURCE -> MHP_MERGE_RESOURCE

Let's make "MEMHP_MERGE_RESOURCE" consistent with "MHP_NONE", "mhp_t" and
"mhp_flags".  As discussed recently [1], "mhp" is our internal acronym for
memory hotplug now.

[1] https://lore.kernel.org/linux-mm/c37de2d0-28a1-4f7d-f944-cfd7d81c334d@redhat.com/

Link: https://lkml.kernel.org/r/20210126115829.10909-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Wei Liu <wei.liu@kernel.org>
Reviewed-by: Pankaj Gupta <pankaj.gupta@cloud.ionos.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Juergen Gross <jgross@suse.com>
Cc: Stefano Stabellini <sstabellini@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memory_hotplug: rename all existing 'memhp' into 'mhp'

This renames all 'memhp' instances to 'mhp' except for memhp_default_state
for being a kernel command line option.  This is just a clean up and
should not cause a functional change.  Let's make it consistent rater than
mixing the two prefixes.  In preparation for more users of the 'mhp'
terminology.

Link: https://lkml.kernel.org/r/1611554093-27316-1-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Suggested-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: "Rafael J. Wysocki" <rafael@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: fix memory_failure() handling of dax-namespace metadata

Given 'struct dev_pagemap' spans both data pages and metadata pages be
careful to consult the altmap if present to delineate metadata.  In fact
the pfn_first() helper already identifies the first valid data pfn, so
export that helper for other code paths via pgmap_pfn_valid().

Other usage of get_dev_pagemap() are not a concern because those are
operating on known data pfns having been looked up by get_user_pages().
I.e.  metadata pfns are never user mapped.

Link: https://lkml.kernel.org/r/161058501758.1840162.4239831989762604527.stgit@dwillia2-desk3.amr.corp.intel.com
Fixes: 6100e34b2526 ("mm, memory_failure: Teach memory_failure() about dev_pagemap pages")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Qian Cai <cai@lca.pw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: teach pfn_to_online_page() about ZONE_DEVICE section collisions

While pfn_to_online_page() is able to determine pfn_valid() at subsection
granularity it is not able to reliably determine if a given pfn is also
online if the section is mixes ZONE_{NORMAL,MOVABLE} with ZONE_DEVICE.
This means that pfn_to_online_page() may return invalid @page objects.
For example with a memory map like:

100000000-1fbffffff : System RAM
  142000000-143002e16 : Kernel code
  143200000-143713fff : Kernel rodata
  143800000-143b15b7f : Kernel data
  144227000-144ffffff : Kernel bss
1fc000000-2fbffffff : Persistent Memory (legacy)
  1fc000000-2fbffffff : namespace0.0

This command:

echo 0x1fc000000 > /sys/devices/system/memory/soft_offline_page

...succeeds when it should fail.  When it succeeds it touches an
uninitialized page and may crash or cause other damage (see
dissolve_free_huge_page()).

While the memory map above is contrived via the memmap=ss!nn kernel
command line option, the collision happens in practice on shipping
platforms.  The memory controller resources that decode spans of physical
address space are a limited resource.  One technique platform-firmware
uses to conserve those resources is to share a decoder across 2 devices to
keep the address range contiguous.  Unfortunately the unit of operation of
a decoder is 64MiB while the Linux section size is 128MiB.  This results
in situations where, without subsection hotplug memory mappings with
different lifetimes collide into one object that can only express one
lifetime.

Update move_pfn_range_to_zone() to flag (SECTION_TAINT_ZONE_DEVICE) a
section that mixes ZONE_DEVICE pfns with other online pfns.  With
SECTION_TAINT_ZONE_DEVICE to delineate, pfn_to_online_page() can fall back
to a slow-path check for ZONE_DEVICE pfns in an online section.  In the
fast path online_section() for a full ZONE_DEVICE section returns false.

Because the collision case is rare, and for simplicity, the
SECTION_TAINT_ZONE_DEVICE flag is never cleared once set.

[dan.j.williams@intel.com: fix CONFIG_ZONE_DEVICE=n build]
Link: https://lkml.kernel.org/r/CAPcyv4iX+7LAgAeSqx7Zw-Zd=ZV9gBv8Bo7oTbwCOOqJoZ3+Yg@mail.gmail.com
Link: https://lkml.kernel.org/r/161058500675.1840162.7887862152161279354.stgit@dwillia2-desk3.amr.corp.intel.com
Fixes: ba72b4c8cf60 ("mm/sparsemem: support sub-section hotplug")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: Michal Hocko <mhocko@suse.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reported-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Qian Cai <cai@lca.pw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: teach pfn_to_online_page() to consider subsection validity

pfn_to_online_page is primarily used to filter out offline or fully
uninitialized pages.  pfn_valid resp.  online_section_nr have a coarse
per memory section granularity.  If a section shared with a partially
offline memory (e.g.  part of ZONE_DEVICE) then pfn_to_online_page
would lead to a false positive on some pfns.  Fix this by adding
pfn_section_valid check which is subsection aware.

[mhocko@kernel.org: changelog rewrite]

Link: https://lkml.kernel.org/r/161058500148.1840162.4365921007820501696.stgit@dwillia2-desk3.amr.corp.intel.com
Fixes: b13bc35193d9 ("mm/hotplug: invalid PFNs from pfn_to_online_page()")
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: David Hildenbrand <david@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move pfn_to_online_page() out of line

Patch series "mm: Fix pfn_to_online_page() with respect to ZONE_DEVICE", v4.

A pfn-walker that uses pfn_to_online_page() may inadvertently translate a
pfn as online and in the page allocator, when it is offline managed by a
ZONE_DEVICE mapping (details in Patch 3: ("mm: Teach pfn_to_online_page()
about ZONE_DEVICE section collisions")).

The 2 proposals under consideration are teach pfn_to_online_page() to be
precise in the presence of mixed-zone sections, or teach the memory-add
code to drop the System RAM associated with ZONE_DEVICE collisions.  In
order to not regress memory capacity by a few 10s to 100s of MiB the
approach taken in this set is to add precision to pfn_to_online_page().

In the course of validating pfn_to_online_page() a couple other fixes
fell out:

1/ soft_offline_page() fails to drop the reference taken in the
   madvise(..., MADV_SOFT_OFFLINE) case.

2/ memory_failure() uses get_dev_pagemap() to lookup ZONE_DEVICE pages,
   however that mapping may contain data pages and metadata raw pfns.
   Introduce pgmap_pfn_valid() to delineate the 2 types and fail the
   handling of raw metadata pfns.

This patch (of 4);

pfn_to_online_page() is already too large to be a macro or an inline
function.  In anticipation of further logic changes / growth, move it out
of line.

No functional change, just code movement.

Link: https://lkml.kernel.org/r/161058499000.1840162.702316708443239771.stgit@dwillia2-desk3.amr.corp.intel.com
Link: https://lkml.kernel.org/r/161058499608.1840162.10165648147615238793.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Reported-by: Michal Hocko <mhocko@kernel.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Qian Cai <cai@lca.pw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/vmstat.c: erase latency in vmstat_shepherd

Many 100us+ latencies have been deteceted in vmstat_shepherd() on CPX
platform which has 208 logic cpus.  And vmstat_shepherd is queued every
second, which could make the case worse.

Add schedule point in vmstat_shepherd() to erase the latency.

Link: https://lkml.kernel.org/r/20210111035526.1511-1-benbjiang@tencent.com
Signed-off-by: Jiang Biao <benbjiang@tencent.com>
Reported-by: Bin Lai <robinlai@tencent.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: vmstat: add some comments on internal storage of byte items

Byte-accounted items are used for slab object accounting at the cgroup
level, because the objects in a slab page can belong to different cgroups.
At the global level these items always change in multiples of whole slab
pages.  The vmstat code exploits this and stores these items as pages
internally, which allows for more compact per-cpu data.

This optimization isn't self-evident from the asserts and the division in
the stat update functions.  Provide the reader with some context.

Link: https://lkml.kernel.org/r/20210202184411.118614-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: vmstat: fix NOHZ wakeups for node stat changes

On NOHZ, the periodic vmstat flushers on each CPU can go to sleep and
won't wake up until stat changes are detected in the per-cpu deltas of the
zone vmstat counters.

In commit 75ef71840539 ("mm, vmstat: add infrastructure for per-node
vmstats") per-node counters were introduced, and subsequently most stats
were moved from the zone to the node level.  However, the node counters
weren't added to the NOHZ wakeup detection.

In theory this can cause per-cpu errors to remain in the user-reported
stats indefinitely.  In practice this only affects a handful of sub
counters (file_mapped, dirty and writeback e.g.) because other page state
changes at the node level likely involve a change at the zone level as
well (alloc and free, lru ops).  Also, nobody has complained.

Fix it up for completeness: wake up vmstat refreshing on node changes.
Also remove the BUILD_BUG_ONs that assert counter size; we haven't relied
on it since we added sizeof() to the range calculation in commit
13c9aaf7fa01 ("mm/vmstat.c: fix NUMA statistics updates").

Link: https://lkml.kernel.org/r/20210202184342.118513-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: cma: print region name on failure

Print the name of the CMA region for convenience.  This is useful
information to have when cma_alloc() fails.

[pdaly@codeaurora.org: print the "count" variable]
Link: https://lkml.kernel.org/r/20210209142414.12768-1-georgi.djakov@linaro.org
Link: https://lkml.kernel.org/r/20210208115200.20286-1-georgi.djakov@linaro.org
Signed-off-by: Patrick Daly <pdaly@codeaurora.org>
Signed-off-by: Georgi Djakov <georgi.djakov@linaro.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/page_alloc: count CMA pages per zone and print them in /proc/zoneinfo

Let's count the number of CMA pages per zone and print them in
/proc/zoneinfo.

Having access to the total number of CMA pages per zone is helpful for
debugging purposes to know where exactly the CMA pages ended up, and to
figure out how many pages of a zone might behave differently, even after
some of these pages might already have been allocated.

As one example, CMA pages part of a kernel zone cannot be used for
ordinary kernel allocations but instead behave more like ZONE_MOVABLE.

For now, we are only able to get the global nr+free cma pages from
/proc/meminfo and the free cma pages per zone from /proc/zoneinfo.

Example after this patch when booting a 6 GiB QEMU VM with
"hugetlb_cma=2G":
  # cat /proc/zoneinfo | grep cma
          cma      0
        nr_free_cma  0
          cma      0
        nr_free_cma  0
          cma      524288
        nr_free_cma  493016
          cma      0
          cma      0
  # cat /proc/meminfo | grep Cma
  CmaTotal:        2097152 kB
  CmaFree:         1972064 kB

Note: We print even without CONFIG_CMA, just like "nr_free_cma"; this way,
      one can be sure when spotting "cma 0", that there are definetly no
      CMA pages located in a zone.

[david@redhat.com: v2]
Link: https://lkml.kernel.org/r/20210128164533.18566-1-david@redhat.com
[david@redhat.com: v3]
Link: https://lkml.kernel.org/r/20210129113451.22085-1-david@redhat.com
Link: https://lkml.kernel.org/r/20210127101813.6370-3-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/cma: expose all pages to the buddy if activation of an area fails

Right now, if activation fails, we might already have exposed some pages
to the buddy for CMA use (although they will never get actually used by
CMA), and some pages won't be exposed to the buddy at all.

Let's check for "single zone" early and on error, don't expose any pages
for CMA use - instead, expose them to the buddy available for any use.
Simply call free_reserved_page() on every single page - easier than going
via free_reserved_area(), converting back and forth between pfns and virt
addresses.

In addition, make sure to fixup totalcma_pages properly.

Example: 6 GiB QEMU VM with "... hugetlb_cma=2G movablecore=20% ...":
  [    0.006891] hugetlb_cma: reserve 2048 MiB, up to 2048 MiB per node
  [    0.006893] cma: Reserved 2048 MiB at 0x0000000100000000
  [    0.006893] hugetlb_cma: reserved 2048 MiB on node 0
  ...
  [    0.175433] cma: CMA area hugetlb0 could not be activated

Before this patch:
  # cat /proc/meminfo
  MemTotal:        5867348 kB
  MemFree:         5692808 kB
  MemAvailable:    5542516 kB
  ...
  CmaTotal:        2097152 kB
  CmaFree:         1884160 kB

After this patch:
  # cat /proc/meminfo
  MemTotal:        6077308 kB
  MemFree:         5904208 kB
  MemAvailable:    5747968 kB
  ...
  CmaTotal:              0 kB
  CmaFree:               0 kB

Note: cma_init_reserved_mem() makes sure that we always cover full
pageblocks / MAX_ORDER - 1 pages.

Link: https://lkml.kernel.org/r/20210127101813.6370-2-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "Peter Zijlstra (Intel)" <peterz@infradead.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Wei Yang <richard.weiyang@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: cma: allocate cma areas bottom-up

Currently cma areas without a fixed base are allocated close to the end of
the node.  This placement is sub-optimal because of compaction: it brings
pages into the cma area.  In particular, it can bring in hot executable
pages, even if there is a plenty of free memory on the machine.  This
results in cma allocation failures.

Instead let's place cma areas close to the beginning of a node.  In this
case the compaction will help to free cma areas, resulting in better cma
allocation success rates.

If there is enough memory let's try to allocate bottom-up starting with
4GB to exclude any possible interference with DMA32.  On smaller machines
or in a case of a failure, stick with the old behavior.

16GB vm, 2GB cma area:
With this patch:
[    0.000000] Command line: root=/dev/vda3 rootflags=subvol=/root systemd.unified_cgroup_hierarchy=1 enforcing=0 console=ttyS0,115200 hugetlb_cma=2G
[    0.002928] hugetlb_cma: reserve 2048 MiB, up to 2048 MiB per node
[    0.002930] cma: Reserved 2048 MiB at 0x0000000100000000
[    0.002931] hugetlb_cma: reserved 2048 MiB on node 0

Without this patch:
[    0.000000] Command line: root=/dev/vda3 rootflags=subvol=/root systemd.unified_cgroup_hierarchy=1 enforcing=0 console=ttyS0,115200 hugetlb_cma=2G
[    0.002930] hugetlb_cma: reserve 2048 MiB, up to 2048 MiB per node
[    0.002933] cma: Reserved 2048 MiB at 0x00000003c0000000
[    0.002934] hugetlb_cma: reserved 2048 MiB on node 0

v2:
  - switched to memblock_set_bottom_up(true), by Mike
  - start with 4GB, by Mike

[guro@fb.com: whitespace fix, per Mike]
Link: https://lkml.kernel.org/r/20201221170551.GB3428478@carbon.DHCP.thefacebook.com
[guro@fb.com: fix 32-bit warnings]
Link: https://lkml.kernel.org/r/20201223163537.GA4011967@carbon.DHCP.thefacebook.com
[guro@fb.com: fix 32-bit systems]
[akpm@linux-foundation.org: build fix]

Link: https://lkml.kernel.org/r/20201217201214.3414100-1-guro@fb.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Reviewed-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Wonhyuk Yang <vvghjk1234@gmail.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Rik van Riel <riel@surriel.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,shmem,thp: limit shmem THP allocations to requested zones

Hugh pointed out that the gma500 driver uses shmem pages, but needs to
limit them to the DMA32 zone.  Ensure the allocations resulting from the
gfp_mask returned by limit_gfp_mask use the zone flags that were
originally passed to shmem_getpage_gfp.

Link: https://lkml.kernel.org/r/20210224121016.1314ed6d@imladris.surriel.com
Signed-off-by: Rik van Riel <riel@surriel.com>
Suggested-by: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Xu Yu <xuyu@linux.alibaba.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm,thp,shmem: make khugepaged obey tmpfs mount flags

Currently if thp enabled=[madvise], mounting a tmpfs filesystem with
huge=always and mmapping files from that tmpfs does not result in
khugepaged collapsing those mappings, despite the mount flag indicating
that it should.

Fix that by breaking up the blocks of tests in hugepage_vma_check a little
bit, and testing things in the correct order.

Link: https://lkml.kernel.org/r/20201124194925.623931-4-riel@surriel.com
Fixes: c2231020ea7b ("mm: thp: register mm for khugepaged when merging vma for shmem")
Signed-off-by: Rik van Riel <riel@surriel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Xu Yu <xuyu@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm,thp,shm: limit gfp mask to no more than specified

Matthew Wilcox pointed out that the i915 driver opportunistically
allocates tmpfs memory, but will happily reclaim some of its pool if no
memory is available.

Make sure the gfp mask used to opportunistically allocate a THP is always
at least as restrictive as the original gfp mask.

Link: https://lkml.kernel.org/r/20201124194925.623931-3-riel@surriel.com
Signed-off-by: Rik van Riel <riel@surriel.com>
Suggested-by: Matthew Wilcox <willy@infradead.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Xu Yu <xuyu@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm,thp,shmem: limit shmem THP alloc gfp_mask

Patch series "mm,thp,shm: limit shmem THP alloc gfp_mask", v6.

The allocation flags of anonymous transparent huge pages can be controlled
through the files in /sys/kernel/mm/transparent_hugepage/defrag, which can
help the system from getting bogged down in the page reclaim and
compaction code when many THPs are getting allocated simultaneously.

However, the gfp_mask for shmem THP allocations were not limited by those
configuration settings, and some workloads ended up with all CPUs stuck on
the LRU lock in the page reclaim code, trying to allocate dozens of THPs
simultaneously.

This patch applies the same configurated limitation of THPs to shmem
hugepage allocations, to prevent that from happening.

This way a THP defrag setting of "never" or "defer+madvise" will result in
quick allocation failures without direct reclaim when no 2MB free pages
are available.

With this patch applied, THP allocations for tmpfs will be a little more
aggressive than today for files mmapped with MADV_HUGEPAGE, and a little
less aggressive for files that are not mmapped or mapped without that
flag.

This patch (of 4):

The allocation flags of anonymous transparent huge pages can be controlled
through the files in /sys/kernel/mm/transparent_hugepage/defrag, which can
help the system from getting bogged down in the page reclaim and
compaction code when many THPs are getting allocated simultaneously.

However, the gfp_mask for shmem THP allocations were not limited by those
configuration settings, and some workloads ended up with all CPUs stuck on
the LRU lock in the page reclaim code, trying to allocate dozens of THPs
simultaneously.

This patch applies the same configurated limitation of THPs to shmem
hugepage allocations, to prevent that from happening.

Controlling the gfp_mask of THP allocations through the knobs in sysfs
allows users to determine the balance between how aggressively the system
tries to allocate THPs at fault time, and how much the application may end
up stalling attempting those allocations.

This way a THP defrag setting of "never" or "defer+madvise" will result in
quick allocation failures without direct reclaim when no 2MB free pages
are available.

With this patch applied, THP allocations for tmpfs will be a little more
aggressive than today for files mmapped with MADV_HUGEPAGE, and a little
less aggressive for files that are not mmapped or mapped without that
flag.

Link: https://lkml.kernel.org/r/20201124194925.623931-1-riel@surriel.com
Link: https://lkml.kernel.org/r/20201124194925.623931-2-riel@surriel.com
Signed-off-by: Rik van Riel <riel@surriel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Xu Yu <xuyu@linux.alibaba.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
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: remove pagevec_lookup_entries

pagevec_lookup_entries() is now just a wrapper around find_get_entries()
so remove it and convert all its callers.

Link: https://lkml.kernel.org/r/20201112212641.27837-15-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: pass pvec directly to find_get_entries

All callers of find_get_entries() use a pvec, so pass it directly instead
of manipulating it in the caller.

Link: https://lkml.kernel.org/r/20201112212641.27837-14-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove nr_entries parameter from pagevec_lookup_entries

All callers want to fetch the full size of the pvec.

Link: https://lkml.kernel.org/r/20201112212641.27837-13-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add an 'end' parameter to pagevec_lookup_entries

Simplifies the callers and uses the existing functionality in
find_get_entries().  We can also drop the final argument of
truncate_exceptional_pvec_entries() and simplify the logic in that
function.

Link: https://lkml.kernel.org/r/20201112212641.27837-12-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add an 'end' parameter to find_get_entries

This simplifies the callers and leads to a more efficient implementation
since the XArray has this functionality already.

Link: https://lkml.kernel.org/r/20201112212641.27837-11-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add and use find_lock_entries

We have three functions (shmem_undo_range(), truncate_inode_pages_range()
and invalidate_mapping_pages()) which want exactly this function, so add
it to filemap.c.  Before this patch, shmem_undo_range() would split any
compound page which overlaps either end of the range being punched in both
the first and second loops through the address space.  After this patch,
that functionality is left for the second loop, which is arguably more
appropriate since the first loop is supposed to run through all the pages
quickly, and splitting a page can sleep.

[willy@infradead.org: add assertion]
Link: https://lkml.kernel.org/r/20201124041507.28996-3-willy@infradead.org
Link: https://lkml.kernel.org/r/20201112212641.27837-10-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

iomap: use mapping_seek_hole_data

Enhance mapping_seek_hole_data() to handle partially uptodate pages and
convert the iomap seek code to call it.

Link: https://lkml.kernel.org/r/20201112212641.27837-9-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap: add mapping_seek_hole_data

Rewrite shmem_seek_hole_data() and move it to filemap.c.

[willy@infradead.org: don't put an xa_is_value() page]
Link: https://lkml.kernel.org/r/20201124041507.28996-4-willy@infradead.org
Link: https://lkml.kernel.org/r/20201112212641.27837-8-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap: add helper for finding pages

There is a lot of common code in find_get_entries(),
find_get_pages_range() and find_get_pages_range_tag().  Factor out
find_get_entry() which simplifies all three functions.

[willy@infradead.org: remove VM_BUG_ON_PAGE()]
Link: https://lkml.kernel.org/r/20201124041507.28996-2-willy@infradead.orgLink:
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap: rename find_get_entry to mapping_get_entry

find_get_entry doesn't "find" anything.  It returns the entry at a
particular index.

Link: https://lkml.kernel.org/r/20201112212641.27837-6-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add FGP_ENTRY

The functionality of find_lock_entry() and find_get_entry() can be
provided by pagecache_get_page(), which lets us delete find_lock_entry()
and make find_get_entry() static.

Link: https://lkml.kernel.org/r/20201112212641.27837-5-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swap: optimise get_shadow_from_swap_cache

There's no need to get a reference to the page, just load the entry and
see if it's a shadow entry.

Link: https://lkml.kernel.org/r/20201112212641.27837-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/shmem: use pagevec_lookup in shmem_unlock_mapping

The comment shows that the reason for using find_get_entries() is now
stale; find_get_pages() will not return 0 if it hits a consecutive run of
swap entries, and I don't believe it has since 2011.  pagevec_lookup() is
a simpler function to use than find_get_pages(), so use it instead.

Link: https://lkml.kernel.org/r/20201112212641.27837-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: make pagecache tagged lookups return only head pages

Patch series "Overhaul multi-page lookups for THP", v4.

This THP prep patchset changes several page cache iteration APIs to only
return head pages.

 - It's only possible to tag head pages in the page cache, so only
   return head pages, not all their subpages.
 - Factor a lot of common code out of the various batch lookup routines
 - Add mapping_seek_hole_data()
 - Unify find_get_entries() and pagevec_lookup_entries()
 - Make find_get_entries only return head pages, like find_get_entry().

These are only loosely connected, but they seem to make sense together as
a series.

This patch (of 14):

Pagecache tags are used for dirty page writeback.  Since dirtiness is
tracked on a per-THP basis, we only want to return the head page rather
than each subpage of a tagged page.  All the filesystems which use huge
pages today are in-memory, so there are no tagged huge pages today.

Link: https://lkml.kernel.org/r/20201112212641.27837-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Cc: Dave Chinner <dchinner@redhat.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>

Merge tag 'kbuild-v5.12' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild updates from Masahiro Yamada:

 - Fix false-positive build warnings for ARCH=ia64 builds

 - Optimize dictionary size for module compression with xz

 - Check the compiler and linker versions in Kconfig

 - Fix misuse of extra-y

 - Support DWARF v5 debug info

 - Clamp SUBLEVEL to 255 because stable releases 4.4.x and 4.9.x
   exceeded the limit

 - Add generic syscall{tbl,hdr}.sh for cleanups across arches

 - Minor cleanups of genksyms

 - Minor cleanups of Kconfig

* tag 'kbuild-v5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild: (38 commits)
  initramfs: Remove redundant dependency of RD_ZSTD on BLK_DEV_INITRD
  kbuild: remove deprecated 'always' and 'hostprogs-y/m'
  kbuild: parse C= and M= before changing the working directory
  kbuild: reuse this-makefile to define abs_srctree
  kconfig: unify rule of config, menuconfig, nconfig, gconfig, xconfig
  kconfig: omit --oldaskconfig option for 'make config'
  kconfig: fix 'invalid option' for help option
  kconfig: remove dead code in conf_askvalue()
  kconfig: clean up nested if-conditionals in check_conf()
  kconfig: Remove duplicate call to sym_get_string_value()
  Makefile: Remove # characters from compiler string
  Makefile: reuse CC_VERSION_TEXT
  kbuild: check the minimum linker version in Kconfig
  kbuild: remove ld-version macro
  scripts: add generic syscallhdr.sh
  scripts: add generic syscalltbl.sh
  arch: syscalls: remove $(srctree)/ prefix from syscall tables
  arch: syscalls: add missing FORCE and fix 'targets' to make if_changed work
  gen_compile_commands: prune some directories
  kbuild: simplify access to the kernel's version
  ...

Merge tag 'ext4_for_linus' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 updates from Ted Ts'o:
 "Miscellaneous ext4 cleanups and bug fixes. Pretty boring this cycle..."

* tag 'ext4_for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: add .kunitconfig fragment to enable ext4-specific tests
  ext: EXT4_KUNIT_TESTS should depend on EXT4_FS instead of selecting it
  ext4: reset retry counter when ext4_alloc_file_blocks() makes progress
  ext4: fix potential htree index checksum corruption
  ext4: factor out htree rep invariant check
  ext4: Change list_for_each* to list_for_each_entry*
  ext4: don't try to processed freed blocks until mballoc is initialized
  ext4: use DEFINE_MUTEX() for mutex lock

Merge tag 'pci-v5.12-changes' of git://git./linux/kernel/git/helgaas/pci

Pull PCI updates from Bjorn Helgaas:
 "Enumeration:
   - Remove unnecessary locking around _OSC (Bjorn Helgaas)
   - Clarify message about _OSC failure (Bjorn Helgaas)
   - Remove notification of PCIe bandwidth changes (Bjorn Helgaas)
   - Tidy checking of syscall user config accessors (Heiner Kallweit)

  Resource management:
   - Decline to resize resources if boot config must be preserved (Ard
     Biesheuvel)
   - Fix pci_register_io_range() memory leak (Geert Uytterhoeven)

  Error handling (Keith Busch):
   - Clear error status from the correct device
   - Retain error recovery status so drivers can use it after reset
   - Log the type of Port (Root or Switch Downstream) that we reset
   - Always request a reset for Downstream Ports in frozen state

  Endpoint framework and NTB (Kishon Vijay Abraham I):
   - Make *_get_first_free_bar() take into account 64 bit BAR
   - Add helper API to get the 'next' unreserved BAR
   - Make *_free_bar() return error codes on failure
   - Remove unused pci_epf_match_device()
   - Add support to associate secondary EPC with EPF
   - Add support in configfs to associate two EPCs with EPF
   - Add pci_epc_ops to map MSI IRQ
   - Add pci_epf_ops to expose function-specific attrs
   - Allow user to create sub-directory of 'EPF Device' directory
   - Implement ->msi_map_irq() ops for cadence
   - Configure LM_EP_FUNC_CFG based on epc->function_num_map for cadence
   - Add EP function driver to provide NTB functionality
   - Add support for EPF PCI Non-Transparent Bridge
   - Add specification for PCI NTB function device
   - Add PCI endpoint NTB function user guide
   - Add configfs binding documentation for pci-ntb endpoint function

  Broadcom STB PCIe controller driver:
   - Add support for BCM4908 and external PERST# signal controller
     (Rafał Miłecki)

  Cadence PCIe controller driver:
   - Retrain Link to work around Gen2 training defect (Nadeem Athani)
   - Fix merge botch in cdns_pcie_host_map_dma_ranges() (Krzysztof
     Wilczyński)

  Freescale Layerscape PCIe controller driver:
   - Add LX2160A rev2 EP mode support (Hou Zhiqiang)
   - Convert to builtin_platform_driver() (Michael Walle)

  MediaTek PCIe controller driver:
   - Fix OF node reference leak (Krzysztof Wilczyński)

  Microchip PolarFlare PCIe controller driver:
   - Add Microchip PolarFire PCIe controller driver (Daire McNamara)

  Qualcomm PCIe controller driver:
   - Use PHY_REFCLK_USE_PAD only for ipq8064 (Ansuel Smith)
   - Add support for ddrss_sf_tbu clock for sm8250 (Dmitry Baryshkov)

  Renesas R-Car PCIe controller driver:
   - Drop PCIE_RCAR config option (Lad Prabhakar)
   - Always allocate MSI addresses in 32bit space (Marek Vasut)

  Rockchip PCIe controller driver:
   - Add FriendlyARM NanoPi M4B DT binding (Chen-Yu Tsai)
   - Make 'ep-gpios' DT property optional (Chen-Yu Tsai)

  Synopsys DesignWare PCIe controller driver:
   - Work around ECRC configuration hardware defect (Vidya Sagar)
   - Drop support for config space in DT 'ranges' (Rob Herring)
   - Change size to u64 for EP outbound iATU (Shradha Todi)
   - Add upper limit address for outbound iATU (Shradha Todi)
   - Make dw_pcie ops optional (Jisheng Zhang)
   - Remove unnecessary dw_pcie_ops from al driver (Jisheng Zhang)

  Xilinx Versal CPM PCIe controller driver:
   - Fix OF node reference leak (Pan Bian)

  Miscellaneous:
   - Remove tango host controller driver (Arnd Bergmann)
   - Remove IRQ handler & data together (altera-msi, brcmstb, dwc)
     (Martin Kaiser)
   - Fix xgene-msi race in installing chained IRQ handler (Martin
     Kaiser)
   - Apply CONFIG_PCI_DEBUG to entire drivers/pci hierarchy (Junhao He)
   - Fix pci-bridge-emul array overruns (Russell King)
   - Remove obsolete uses of WARN_ON(in_interrupt()) (Sebastian Andrzej
     Siewior)"

* tag 'pci-v5.12-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci: (69 commits)
  PCI: qcom: Use PHY_REFCLK_USE_PAD only for ipq8064
  PCI: qcom: Add support for ddrss_sf_tbu clock
  dt-bindings: PCI: qcom: Document ddrss_sf_tbu clock for sm8250
  PCI: al: Remove useless dw_pcie_ops
  PCI: dwc: Don't assume the ops in dw_pcie always exist
  PCI: dwc: Add upper limit address for outbound iATU
  PCI: dwc: Change size to u64 for EP outbound iATU
  PCI: dwc: Drop support for config space in 'ranges'
  PCI: layerscape: Convert to builtin_platform_driver()
  PCI: layerscape: Add LX2160A rev2 EP mode support
  dt-bindings: PCI: layerscape: Add LX2160A rev2 compatible strings
  PCI: dwc: Work around ECRC configuration issue
  PCI/portdrv: Report reset for frozen channel
  PCI/AER: Specify the type of Port that was reset
  PCI/ERR: Retain status from error notification
  PCI/AER: Clear AER status from Root Port when resetting Downstream Port
  PCI/ERR: Clear status of the reporting device
  dt-bindings: arm: rockchip: Add FriendlyARM NanoPi M4B
  PCI: rockchip: Make 'ep-gpios' DT property optional
  Documentation: PCI: Add PCI endpoint NTB function user guide
  ...

Merge tag 'nds32-for-linux-5.12' of git://git./linux/kernel/git/greentime/linux

Pull nds32 updates from Greentime Hu:
 "Code clean-up and refinement"

* tag 'nds32-for-linux-5.12' of git://git.kernel.org/pub/scm/linux/kernel/git/greentime/linux:
  nds32: Fix bogus reference to <asm/procinfo.h>
  nds32: use get_kernel_nofault in dump_mem
  nds32: remove dump_instr
  nds32: configs: Cleanup CONFIG_CROSS_COMPILE
  nds32: Replace <linux/clk-provider.h> by <linux/of_clk.h>

nds32: Fix bogus reference to <asm/procinfo.h>

Andestech(nds32) never had <asm/procinfo.h>.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Greentime Hu <green.hu@gmail.com>
Signed-off-by: Greentime Hu <green.hu@gmail.com>

nds32: use get_kernel_nofault in dump_mem

Use the proper get_kernel_nofault helper to access an unsafe kernel
pointer without faulting instead of playing with set_fs and get_user.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Nick Hu <nickhu@andestech.com>
Acked-by: Greentime Hu <green.hu@gmail.com>
Signed-off-by: Greentime Hu <green.hu@gmail.com>

nds32: remove dump_instr

dump_inst has a return before actually doing anything, so just drop the
whole thing.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Nick Hu <nickhu@andestech.com>
Acked-by: Greentime Hu <green.hu@gmail.com>
Signed-off-by: Greentime Hu <green.hu@gmail.com>

nds32: configs: Cleanup CONFIG_CROSS_COMPILE

CONFIG_CROSS_COMPILE is gone since commit f1089c92da79 ("kbuild: remove
CONFIG_CROSS_COMPILE support").

Signed-off-by: Krzysztof Kozlowski <krzk@kernel.org>
Acked-by: Greentime Hu <green.hu@gmail.com>
Signed-off-by: Greentime Hu <green.hu@gmail.com>

nds32: Replace <linux/clk-provider.h> by <linux/of_clk.h>

The Andes platform code is not a clock provider, and just needs to call
of_clk_init().

Hence it can include <linux/of_clk.h> instead of <linux/clk-provider.h>.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Acked-by: Greentime Hu <green.hu@gmail.com>
Reviewed-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greentime Hu <green.hu@gmail.com>

Merge tag 'x86-entry-2021-02-24' of git://git./linux/kernel/git/tip/tip

Pull x86 irq entry updates from Thomas Gleixner:
 "The irq stack switching was moved out of the ASM entry code in course
  of the entry code consolidation. It ended up being suboptimal in
  various ways.

  This reworks the X86 irq stack handling:

   - Make the stack switching inline so the stackpointer manipulation is
     not longer at an easy to find place.

   - Get rid of the unnecessary indirect call.

   - Avoid the double stack switching in interrupt return and reuse the
     interrupt stack for softirq handling.

   - A objtool fix for CONFIG_FRAME_POINTER=y builds where it got
     confused about the stack pointer manipulation"

* tag 'x86-entry-2021-02-24' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  objtool: Fix stack-swizzle for FRAME_POINTER=y
  um: Enforce the usage of asm-generic/softirq_stack.h
  x86/softirq/64: Inline do_softirq_own_stack()
  softirq: Move do_softirq_own_stack() to generic asm header
  softirq: Move __ARCH_HAS_DO_SOFTIRQ to Kconfig
  x86: Select CONFIG_HAVE_IRQ_EXIT_ON_IRQ_STACK
  x86/softirq: Remove indirection in do_softirq_own_stack()
  x86/entry: Use run_sysvec_on_irqstack_cond() for XEN upcall
  x86/entry: Convert device interrupts to inline stack switching
  x86/entry: Convert system vectors to irq stack macro
  x86/irq: Provide macro for inlining irq stack switching
  x86/apic: Split out spurious handling code
  x86/irq/64: Adjust the per CPU irq stack pointer by 8
  x86/irq: Sanitize irq stack tracking
  x86/entry: Fix instrumentation annotation

Merge branch 'akpm' (patches from Andrew)

Merge misc updates from Andrew Morton:
 "A few small subsystems and some of MM.

  172 patches.

  Subsystems affected by this patch series: hexagon, scripts, ntfs,
  ocfs2, vfs, and mm (slab-generic, slab, slub, debug, pagecache, swap,
  memcg, pagemap, mprotect, mremap, page-reporting, vmalloc, kasan,
  pagealloc, memory-failure, hugetlb, vmscan, z3fold, compaction,
  mempolicy, oom-kill, hugetlbfs, and migration)"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>: (172 commits)
  mm/migrate: remove unneeded semicolons
  hugetlbfs: remove unneeded return value of hugetlb_vmtruncate()
  hugetlbfs: fix some comment typos
  hugetlbfs: correct some obsolete comments about inode i_mutex
  hugetlbfs: make hugepage size conversion more readable
  hugetlbfs: remove meaningless variable avoid_reserve
  hugetlbfs: correct obsolete function name in hugetlbfs_read_iter()
  hugetlbfs: use helper macro default_hstate in init_hugetlbfs_fs
  hugetlbfs: remove useless BUG_ON(!inode) in hugetlbfs_setattr()
  hugetlbfs: remove special hugetlbfs_set_page_dirty()
  mm/hugetlb: change hugetlb_reserve_pages() to type bool
  mm, oom: fix a comment in dump_task()
  mm/mempolicy: use helper range_in_vma() in queue_pages_test_walk()
  numa balancing: migrate on fault among multiple bound nodes
  mm, compaction: make fast_isolate_freepages() stay within zone
  mm/compaction: fix misbehaviors of fast_find_migrateblock()
  mm/compaction: correct deferral logic for proactive compaction
  mm/compaction: remove duplicated VM_BUG_ON_PAGE !PageLocked
  mm/compaction: remove rcu_read_lock during page compaction
  z3fold: simplify the zhdr initialization code in init_z3fold_page()
  ...

mm/migrate: remove unneeded semicolons

Remove superfluous semicolons after function definitions.

Link: https://lkml.kernel.org/r/20210115110131.2359683-1-cy.fan@huawei.com
Signed-off-by: Chengyang Fan <cy.fan@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: remove unneeded return value of hugetlb_vmtruncate()

The function hugetlb_vmtruncate() is guaranteed to always success since
commit 7aa91e104028 ("hugetlb: allow extending ftruncate on hugetlbfs").
So we should remove the unneeded return value which is always 0.

Link: https://lkml.kernel.org/r/20210208084637.47789-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
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>

hugetlbfs: fix some comment typos

Fix typos reserv to reserve, minimim to minimum. No functional change
intended.

Link: https://lkml.kernel.org/r/20210130092351.28072-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: correct some obsolete comments about inode i_mutex

Since commit 9902af79c01a ("parallel lookups: actual switch to rwsem"),
i_mutex of inode is converted to i_rwsem. So replace i_mutex with i_rwsem
to make comments up to date.

Link: https://lkml.kernel.org/r/20210127093111.36672-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: make hugepage size conversion more readable

The calculation 1U << (h->order + PAGE_SHIFT - 10) is actually equal to
(PAGE_SHIFT << (h->order)) >> 10.  So we can make it more readable by
replace it with huge_page_size(h) >> 10.

Link: https://lkml.kernel.org/r/20210122083141.24548-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: remove meaningless variable avoid_reserve

The variable avoid_reserve is meaningless because we never changed its
value and just passed it to alloc_huge_page().  So remove it to make code
more clear that in hugetlbfs_fallocate, we never avoid reserve when alloc
hugepage yet.  Also add a comment offered by Mike Kravetz to explain this.

Link: https://lkml.kernel.org/r/20210120071508.9078-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: correct obsolete function name in hugetlbfs_read_iter()

Since commit 36e789144267 ("kill do_generic_mapping_read"), the function
do_generic_mapping_read() is renamed to do_generic_file_read(). And then
commit 47c27bc46946 ("fs: pass iocb to do_generic_file_read") renamed it
to generic_file_buffered_read(). So replace do_generic_mapping_read() with
generic_file_buffered_read() to keep comment uptodate.

Link: https://lkml.kernel.org/r/20210118063210.47118-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: use helper macro default_hstate in init_hugetlbfs_fs

Since commit e5ff215941d5 ("hugetlb: multiple hstates for multiple page
sizes"), we can use macro default_hstate to get the struct hstate which we
use by default.  But init_hugetlbfs_fs() forgot to use it.

Link: https://lkml.kernel.org/r/20210116091827.20982-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: remove useless BUG_ON(!inode) in hugetlbfs_setattr()

When we reach here with inode = NULL, we should have crashed as inode has
already been dereferenced via hstate_inode.  So this BUG_ON(!inode) does
not take effect and should be removed.

Link: https://lkml.kernel.org/r/20210118110700.52506-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: remove special hugetlbfs_set_page_dirty()

Matthew Wilcox noticed that hugetlbfs_set_page_dirty always returns 0.
Instead, it should return 1 or 0 depending on the previous state of the
dirty bit.  In addition, the call to compound_head is redundant as it is
also performed in calling routine set_page_dirty.

Replace the hugetlbfs specific routine hugetlbfs_set_page_dirty with
__set_page_dirty_no_writeback as it addresses both of these issues.

Link: https://lkml.kernel.org/r/20201221192542.15732-2-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Suggested-by: Matthew Wilcox <willy@infradead.org>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Davidlohr Bueso <dave@stgolabs.net>
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/hugetlb: change hugetlb_reserve_pages() to type bool

While reviewing a bug in hugetlb_reserve_pages, it was noticed that all
callers ignore the return value.  Any failure is considered an ENOMEM
error by the callers.

Change the function to be of type bool.  The function will return true if
the reservation was successful, false otherwise.  Callers currently assume
a zero return code indicates success.  Change the callers to look for true
to indicate success.  No functional change, only code cleanup.

Link: https://lkml.kernel.org/r/20201221192542.15732-1-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, oom: fix a comment in dump_task()

If p is a kthread, it will be checked in oom_unkillable_task() so
we can delete the corresponding comment.

Link: https://lkml.kernel.org/r/20210125133006.7242-1-tangyizhou@huawei.com
Signed-off-by: Tang Yizhou <tangyizhou@huawei.com>
Acked-by: David Rientjes <rientjes@google.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/mempolicy: use helper range_in_vma() in queue_pages_test_walk()

The helper range_in_vma() is introduced via commit 017b1660df89 ("mm:
migration: fix migration of huge PMD shared pages"). But we forgot to
use it in queue_pages_test_walk().

Link: https://lkml.kernel.org/r/20210130091352.20220-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
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>

numa balancing: migrate on fault among multiple bound nodes

Now, NUMA balancing can only optimize the page placement among the NUMA
nodes if the default memory policy is used.  Because the memory policy
specified explicitly should take precedence.  But this seems too strict in
some situations.  For example, on a system with 4 NUMA nodes, if the
memory of an application is bound to the node 0 and 1, NUMA balancing can
potentially migrate the pages between the node 0 and 1 to reduce
cross-node accessing without breaking the explicit memory binding policy.

So in this patch, we add MPOL_F_NUMA_BALANCING mode flag to
set_mempolicy() when mode is MPOL_BIND.  With the flag specified, NUMA
balancing will be enabled within the thread to optimize the page placement
within the constrains of the specified memory binding policy.  With the
newly added flag, the NUMA balancing control mechanism becomes,

 - sysctl knob numa_balancing can enable/disable the NUMA balancing
   globally.

 - even if sysctl numa_balancing is enabled, the NUMA balancing will be
   disabled for the memory areas or applications with the explicit
   memory policy by default.

 - MPOL_F_NUMA_BALANCING can be used to enable the NUMA balancing for
   the applications when specifying the explicit memory policy
   (MPOL_BIND).

Various page placement optimization based on the NUMA balancing can be
done with these flags.  As the first step, in this patch, if the memory of
the application is bound to multiple nodes (MPOL_BIND), and in the hint
page fault handler the accessing node are in the policy nodemask, the page
will be tried to be migrated to the accessing node to reduce the
cross-node accessing.

If the newly added MPOL_F_NUMA_BALANCING flag is specified by an
application on an old kernel version without its support, set_mempolicy()
will return -1 and errno will be set to EINVAL.  The application can use
this behavior to run on both old and new kernel versions.

And if the MPOL_F_NUMA_BALANCING flag is specified for the mode other than
MPOL_BIND, set_mempolicy() will return -1 and errno will be set to EINVAL
as before.  Because we don't support optimization based on the NUMA
balancing for these modes.

In the previous version of the patch, we tried to reuse MPOL_MF_LAZY for
mbind().  But that flag is tied to MPOL_MF_MOVE.*, so it seems not a good
API/ABI for the purpose of the patch.

And because it's not clear whether it's necessary to enable NUMA balancing
for a specific memory area inside an application, so we only add the flag
at the thread level (set_mempolicy()) instead of the memory area level
(mbind()).  We can do that when it become necessary.

To test the patch, we run a test case as follows on a 4-node machine with
192 GB memory (48 GB per node).

1. Change pmbench memory accessing benchmark to call set_mempolicy()
   to bind its memory to node 1 and 3 and enable NUMA balancing.  Some
   related code snippets are as follows,

     #include <numaif.h>
     #include <numa.h>

struct bitmask *bmp;
int ret;

bmp = numa_parse_nodestring("1,3");
ret = set_mempolicy(MPOL_BIND | MPOL_F_NUMA_BALANCING,
    bmp->maskp, bmp->size + 1);
/* If MPOL_F_NUMA_BALANCING isn't supported, fall back to MPOL_BIND */
if (ret < 0 && errno == EINVAL)
ret = set_mempolicy(MPOL_BIND, bmp->maskp, bmp->size + 1);
if (ret < 0) {
perror("Failed to call set_mempolicy");
exit(-1);
}

2. Run a memory eater on node 3 to use 40 GB memory before running pmbench.

3. Run pmbench with 64 processes, the working-set size of each process
   is 640 MB, so the total working-set size is 64 * 640 MB = 40 GB.  The
   CPU and the memory (as in step 1.) of all pmbench processes is bound
   to node 1 and 3. So, after CPU usage is balanced, some pmbench
   processes run on the CPUs of the node 3 will access the memory of
   the node 1.

4. After the pmbench processes run for 100 seconds, kill the memory
   eater.  Now it's possible for some pmbench processes to migrate
   their pages from node 1 to node 3 to reduce cross-node accessing.

Test results show that, with the patch, the pages can be migrated from
node 1 to node 3 after killing the memory eater, and the pmbench score
can increase about 17.5%.

Link: https://lkml.kernel.org/r/20210120061235.148637-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, compaction: make fast_isolate_freepages() stay within zone

Compaction always operates on pages from a single given zone when
isolating both pages to migrate and freepages.  Pageblock boundaries are
intersected with zone boundaries to be safe in case zone starts or ends in
the middle of pageblock.  The use of pageblock_pfn_to_page() protects
against non-contiguous pageblocks.

The functions fast_isolate_freepages() and fast_isolate_around() don't
currently protect the fast freepage isolation thoroughly enough against
these corner cases, and can result in freepage isolation operate outside
of zone boundaries:

 - in fast_isolate_freepages() if we get a pfn from the first pageblock
   of a zone that starts in the middle of that pageblock, 'highest' can
   be a pfn outside of the zone.

   If we fail to isolate anything in this function, we may then call
   fast_isolate_around() on a pfn outside of the zone and there
   effectively do a set_pageblock_skip(page_to_pfn(highest)) which may
   currently hit a VM_BUG_ON() in some configurations

 - fast_isolate_around() checks only the zone end boundary and not
   beginning, nor that the pageblock is contiguous (with
   pageblock_pfn_to_page()) so it's possible that we end up calling
   isolate_freepages_block() on a range of pfn's from two different
   zones and end up e.g. isolating freepages under the wrong zone's
   lock.

This patch should fix the above issues.

Link: https://lkml.kernel.org/r/20210217173300.6394-1-vbabka@suse.cz
Fixes: 5a811889de10 ("mm, compaction: use free lists to quickly locate a migration target")
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/compaction: fix misbehaviors of fast_find_migrateblock()

In the fast_find_migrateblock(), it iterates ocer the freelist to find the
proper pageblock.  But there are some misbehaviors.

First, if the page we found is equal to cc->migrate_pfn, it is considered
that we didn't find a suitable pageblock.  Secondly, if the loop was
terminated because order is less than PAGE_ALLOC_COSTLY_ORDER, it could be
considered that we found a suitable one.  Thirdly, if the skip bit is set
on the page block and we goto continue, it doesn't check nr_scanned.
Fourthly, if the page block's skip bit is set, it checks that page block
is the last of list, which is unnecessary.

Link: https://lkml.kernel.org/r/20210128130411.6125-1-vvghjk1234@gmail.com
Fixes: 70b44595eafe9 ("mm, compaction: use free lists to quickly locate a migration source")
Signed-off-by: Wonhyuk Yang <vvghjk1234@gmail.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
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/compaction: correct deferral logic for proactive compaction

should_proactive_compact_node() returns true when sum of the weighted
fragmentation score of all the zones in the node is greater than the
wmark_high of compaction, which then triggers the proactive compaction
that operates on the individual zones of the node.  But proactive
compaction runs on the zone only when its weighted fragmentation score
is greater than wmark_low(=wmark_high - 10).

This means that the sum of the weighted fragmentation scores of all the
zones can exceed the wmark_high but individual weighted fragmentation zone
scores can still be less than wmark_low which makes the unnecessary
trigger of the proactive compaction only to return doing nothing.

Issue with the return of proactive compaction with out even trying is its
deferral.  It is simply deferred for 1 << COMPACT_MAX_DEFER_SHIFT if the
scores across the proactive compaction is same, thinking that compaction
didn't make any progress but in reality it didn't even try.  With the
delay between successive retries for proactive compaction is 500msec, it
can result into the deferral for ~30sec with out even trying the proactive
compaction.

Test scenario is that: compaction_proactiveness=50 thus the wmark_low = 50
and wmark_high = 60.  System have 2 zones(Normal and Movable) with sizes
5GB and 6GB respectively.  After opening some apps on the android, the
weighted fragmentation scores of these zones are 47 and 49 respectively.
Since the sum of these fragmentation scores are above the wmark_high which
triggers the proactive compaction and there since the individual zones
weighted fragmentation scores are below wmark_low, it returns without
trying the proactive compaction.  As a result the weighted fragmentation
scores of the zones are still 47 and 49 which makes the existing logic to
defer the compaction thinking that noprogress is made across the
compaction.

Fix this by checking just zone fragmentation score, not the weighted, in
__compact_finished() and use the zones weighted fragmentation score in
fragmentation_score_node().  In the test case above, If the weighted
average of is above wmark_high, then individual score (not adjusted) of
atleast one zone has to be above wmark_high.  Thus it avoids the
unnecessary trigger and deferrals of the proactive compaction.

Link: https://lkml.kernel.org/r/1610989938-31374-1-git-send-email-charante@codeaurora.org
Signed-off-by: Charan Teja Reddy <charante@codeaurora.org>
Suggested-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Khalid Aziz <khalid.aziz@oracle.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Nitin Gupta <ngupta@nitingupta.dev>
Cc: Vinayak Menon <vinmenon@codeaurora.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/compaction: remove duplicated VM_BUG_ON_PAGE !PageLocked

The VM_BUG_ON_PAGE(!PageLocked(page), page) is also done in PageMovable.
Remove this explicitly one.

Link: https://lkml.kernel.org/r/20210109081420.46030-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
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/compaction: remove rcu_read_lock during page compaction

isolate_migratepages_block() used rcu_read_lock() with the intention of
safeguarding against the mem_cgroup being destroyed concurrently; but
its TestClearPageLRU already protects against that.  Delete the
unnecessary rcu_read_lock() and _unlock().

Hugh Dickins helped on commit log polishing, Thanks!

Link: https://lkml.kernel.org/r/1608614453-10739-3-git-send-email-alex.shi@linux.alibaba.com
Signed-off-by: Alex Shi <alex.shi@linux.alibaba.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

z3fold: simplify the zhdr initialization code in init_z3fold_page()

We can simplify the zhdr initialization by memset() the zhdr first
instead of set struct member to zero one by one.  This would also make
code more compact and clear.

Link: https://lkml.kernel.org/r/20210120085851.16159-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vitaly Wool <vitaly.wool@konsulko.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

z3fold: remove unused attribute for release_z3fold_page

Since commit dcf5aedb24f8 ("z3fold: stricter locking and more careful
reclaim"), release_z3fold_page() is used again.  So we can drop the
unused attribute safely.

Link: https://lkml.kernel.org/r/20210120084008.58432-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vitaly Wool <vitaly.wool@konsulko.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/vmscan: restore zone_reclaim_mode ABI

I went to go add a new RECLAIM_* mode for the zone_reclaim_mode sysctl.
Like a good kernel developer, I also went to go update the
documentation.  I noticed that the bits in the documentation didn't
match the bits in the #defines.

The VM never explicitly checks the RECLAIM_ZONE bit.  The bit is,
however implicitly checked when checking 'node_reclaim_mode==0'.  The
RECLAIM_ZONE #define was removed in a cleanup.  That, by itself is fine.

But, when the bit was removed (bit 0) the _other_ bit locations also got
changed.  That's not OK because the bit values are documented to mean
one specific thing.  Users surely do not expect the meaning to change
from kernel to kernel.

The end result is that if someone had a script that did:

sysctl vm.zone_reclaim_mode=1

it would have gone from enabling node reclaim for clean unmapped pages
to writing out pages during node reclaim after the commit in question.
That's not great.

Put the bits back the way they were and add a comment so something like
this is a bit harder to do again.  Update the documentation to make it
clear that the first bit is ignored.

Link: https://lkml.kernel.org/r/20210219172555.FF0CDF23@viggo.jf.intel.com
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Fixes: 648b5cf368e0 ("mm/vmscan: remove unused RECLAIM_OFF/RECLAIM_ZONE")
Reviewed-by: Ben Widawsky <ben.widawsky@intel.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Daniel Wagner <dwagner@suse.de>
Cc: "Tobin C. Harding" <tobin@kernel.org>
Cc: Christoph Lameter <cl@linux.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Qian Cai <cai@lca.pw>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: fix uninitialized subpool pointer

Gerald Schaefer reported a panic on s390 in hugepage_subpool_put_pages()
with linux-next 5.12.0-20210222.
Call trace:
  hugepage_subpool_put_pages.part.0+0x2c/0x138
  __free_huge_page+0xce/0x310
  alloc_pool_huge_page+0x102/0x120
  set_max_huge_pages+0x13e/0x350
  hugetlb_sysctl_handler_common+0xd8/0x110
  hugetlb_sysctl_handler+0x48/0x58
  proc_sys_call_handler+0x138/0x238
  new_sync_write+0x10e/0x198
  vfs_write.part.0+0x12c/0x238
  ksys_write+0x68/0xf8
  do_syscall+0x82/0xd0
  __do_syscall+0xb4/0xc8
  system_call+0x72/0x98

This is a result of the change which moved the hugetlb page subpool
pointer from page->private to page[1]->private.  When new pages are
allocated from the buddy allocator, the private field of the head
page will be cleared, but the private field of subpages is not modified.
Therefore, old values may remain.

Fix by initializing hugetlb page subpool pointer in prep_new_huge_page().

Link: https://lkml.kernel.org/r/20210223215544.313871-1-mike.kravetz@oracle.com
Fixes: f1280272ae4d ("hugetlb: use page.private for hugetlb specific page flags")
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reported-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Heiko Carstens <hca@linux.ibm.com>
Cc: Sven Schnelle <svens@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/hugetlb.h: add synchronization information for new hugetlb specific flags

Add comments, no functional change.

Link: https://lkml.kernel.org/r/62a80585-2a73-10cc-4a2d-5721540d4ad2@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.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>

hugetlb: convert PageHugeFreed to HPageFreed flag

Use new hugetlb specific HPageFreed flag to replace the PageHugeFreed
interfaces.

Link: https://lkml.kernel.org/r/20210122195231.324857-6-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: convert PageHugeTemporary() to HPageTemporary flag

Use new hugetlb specific HPageTemporary flag to replace the
PageHugeTemporary() interfaces.  PageHugeTemporary does contain a
PageHuge() check.  However, this interface is only used within hugetlb
code where we know we are dealing with a hugetlb page.  Therefore, the
check can be eliminated.

Link: https://lkml.kernel.org/r/20210122195231.324857-5-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: convert page_huge_active() HPageMigratable flag

Use the new hugetlb page specific flag HPageMigratable to replace the
page_huge_active interfaces.  By it's name, page_huge_active implied that
a huge page was on the active list.  However, that is not really what code
checking the flag wanted to know.  It really wanted to determine if the
huge page could be migrated.  This happens when the page is actually added
to the page cache and/or task page table.  This is the reasoning behind
the name change.

The VM_BUG_ON_PAGE() calls in the *_huge_active() interfaces are not
really necessary as we KNOW the page is a hugetlb page.  Therefore, they
are removed.

The routine page_huge_active checked for PageHeadHuge before testing the
active bit.  This is unnecessary in the case where we hold a reference or
lock and know it is a hugetlb head page.  page_huge_active is also called
without holding a reference or lock (scan_movable_pages), and can race
with code freeing the page.  The extra check in page_huge_active shortened
the race window, but did not prevent the race.  Offline code calling
scan_movable_pages already deals with these races, so removing the check
is acceptable.  Add comment to racy code.

[songmuchun@bytedance.com: remove set_page_huge_active() declaration from include/linux/hugetlb.h]
Link: https://lkml.kernel.org/r/CAMZfGtUda+KoAZscU0718TN61cSFwp4zy=y2oZ=+6Z2TAZZwng@mail.gmail.com
Link: https://lkml.kernel.org/r/20210122195231.324857-3-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlb: use page.private for hugetlb specific page flags

Patch series "create hugetlb flags to consolidate state", v3.

While discussing a series of hugetlb fixes in [1], it became evident that
the hugetlb specific page state information is stored in a somewhat
haphazard manner.  Code dealing with state information would be easier to
read, understand and maintain if this information was stored in a
consistent manner.

This series uses page.private of the hugetlb head page for storing a set
of hugetlb specific page flags.  Routines are priovided for test, set and
clear of the flags.

[1] https://lore.kernel.org/r/20210106084739.63318-1-songmuchun@bytedance.com

This patch (of 4):

As hugetlbfs evolved, state information about hugetlb pages was added.
One 'convenient' way of doing this was to use available fields in tail
pages.  Over time, it has become difficult to know the meaning or contents
of fields simply by looking at a small bit of code.  Sometimes, the naming
is just confusing.  For example: The PagePrivate flag indicates a huge
page reservation was consumed and needs to be restored if an error is
encountered and the page is freed before it is instantiated.  The
page.private field contains the pointer to a subpool if the page is
associated with one.

In an effort to make the code more readable, use page.private to contain
hugetlb specific page flags.  These flags will have test, set and clear
functions similar to those used for 'normal' page flags.  More
importantly, an enum of flag values will be created with names that
actually reflect their purpose.

In this patch,
- Create infrastructure for hugetlb specific page flag functions
- Move subpool pointer to page[1].private to make way for flags
  Create routines with meaningful names to modify subpool field
- Use new HPageRestoreReserve flag instead of PagePrivate

Conversion of other state information will happen in subsequent patches.

Link: https://lkml.kernel.org/r/20210122195231.324857-1-mike.kravetz@oracle.com
Link: https://lkml.kernel.org/r/20210122195231.324857-2-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>