/proc/PID/smaps: Add PMD migration entry parsing

Now, when reading /proc/PID/smaps, the PMD migration entry in page table
is simply ignored.  To improve the accuracy of /proc/PID/smaps, its
parsing and processing is added.

To test the patch, we run pmbench to eat 400 MB memory in background,
then run /usr/bin/migratepages and `cat /proc/PID/smaps` every second.
The issue as follows can be reproduced within 60 seconds.

Before the patch, for the fully populated 400 MB anonymous VMA, some THP
pages under migration may be lost as below.

  7f3f6a7e5000-7f3f837e5000 rw-p 00000000 00:00 0
  Size:             409600 kB
  KernelPageSize:        4 kB
  MMUPageSize:           4 kB
  Rss:              407552 kB
  Pss:              407552 kB
  Shared_Clean:          0 kB
  Shared_Dirty:          0 kB
  Private_Clean:         0 kB
  Private_Dirty:    407552 kB
  Referenced:       301056 kB
  Anonymous:        407552 kB
  LazyFree:              0 kB
  AnonHugePages:    405504 kB
  ShmemPmdMapped:        0 kB
  FilePmdMapped:        0 kB
  Shared_Hugetlb:        0 kB
  Private_Hugetlb:       0 kB
  Swap:                  0 kB
  SwapPss:               0 kB
  Locked:                0 kB
  THPeligible: 1
  VmFlags: rd wr mr mw me ac

After the patch, it will be always,

  7f3f6a7e5000-7f3f837e5000 rw-p 00000000 00:00 0
  Size:             409600 kB
  KernelPageSize:        4 kB
  MMUPageSize:           4 kB
  Rss:              409600 kB
  Pss:              409600 kB
  Shared_Clean:          0 kB
  Shared_Dirty:          0 kB
  Private_Clean:         0 kB
  Private_Dirty:    409600 kB
  Referenced:       294912 kB
  Anonymous:        409600 kB
  LazyFree:              0 kB
  AnonHugePages:    407552 kB
  ShmemPmdMapped:        0 kB
  FilePmdMapped:        0 kB
  Shared_Hugetlb:        0 kB
  Private_Hugetlb:       0 kB
  Swap:                  0 kB
  SwapPss:               0 kB
  Locked:                0 kB
  THPeligible: 1
  VmFlags: rd wr mr mw me ac

Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Link: http://lkml.kernel.org/r/20200403123059.1846960-1-ying.huang@intel.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: ptdump: expand type of 'val' in note_page()

The page table entry is passed in the 'val' argument to note_page(),
however this was previously an "unsigned long" which is fine on 64-bit
platforms.  But for 32 bit x86 it is not always big enough to contain a
page table entry which may be 64 bits.

Change the type to u64 to ensure that it is always big enough.

[akpm@linux-foundation.org: fix riscv]
Reported-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20200521152308.33096-3-steven.price@arm.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

x86: mm: ptdump: calculate effective permissions correctly

Patch series "Fix W+X debug feature on x86"

Jan alerted me[1] that the W+X detection debug feature was broken in x86
by my change[2] to switch x86 to use the generic ptdump infrastructure.

Fundamentally the approach of trying to move the calculation of
effective permissions into note_page() was broken because note_page() is
only called for 'leaf' entries and the effective permissions are passed
down via the internal nodes of the page tree.  The solution I've taken
here is to create a new (optional) callback which is called for all
nodes of the page tree and therefore can calculate the effective
permissions.

Secondly on some configurations (32 bit with PAE) "unsigned long" is not
large enough to store the table entries.  The fix here is simple - let's
just use a u64.

[1] https://lore.kernel.org/lkml/d573dc7e-e742-84de-473d-f971142fa319@suse.com/
[2] 2ae27137b2db ("x86: mm: convert dump_pagetables to use walk_page_range")

This patch (of 2):

By switching the x86 page table dump code to use the generic code the
effective permissions are no longer calculated correctly because the
note_page() function is only called for *leaf* entries.  To calculate
the actual effective permissions it is necessary to observe the full
hierarchy of the page tree.

Introduce a new callback for ptdump which is called for every entry and
can therefore update the prot_levels array correctly.  note_page() can
then simply access the appropriate element in the array.

[steven.price@arm.com: make the assignment conditional on val != 0]
Link: http://lkml.kernel.org/r/430c8ab4-e7cd-6933-dde6-087fac6db872@arm.com
Fixes: 2ae27137b2db ("x86: mm: convert dump_pagetables to use walk_page_range")
Reported-by: Jan Beulich <jbeulich@suse.com>
Signed-off-by: Steven Price <steven.price@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Qian Cai <cai@lca.pw>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20200521152308.33096-1-steven.price@arm.com
Link: http://lkml.kernel.org/r/20200521152308.33096-2-steven.price@arm.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: fix memcg_kmem_bypass() for remote memcg charging

While trying to use remote memcg charging in an out-of-tree kernel
module I found it's not working, because the current thread is a
workqueue thread.

As we will probably encounter this issue in the future as the users of
memalloc_use_memcg() grow, and it's nothing wrong for this usage, it's
better we fix it now.

Signed-off-by: Zefan Li <lizefan@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Roman Gushchin <guro@fb.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Link: http://lkml.kernel.org/r/1d202a12-26fe-0012-ea14-f025ddcd044a@huawei.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcg: automatically penalize tasks with high swap use

Add a memory.swap.high knob, which can be used to protect the system
from SWAP exhaustion.  The mechanism used for penalizing is similar to
memory.high penalty (sleep on return to user space).

That is not to say that the knob itself is equivalent to memory.high.
The objective is more to protect the system from potentially buggy tasks
consuming a lot of swap and impacting other tasks, or even bringing the
whole system to stand still with complete SWAP exhaustion.  Hopefully
without the need to find per-task hard limits.

Slowing misbehaving tasks down gradually allows user space oom killers
or other protection mechanisms to react.  oomd and earlyoom already do
killing based on swap exhaustion, and memory.swap.high protection will
help implement such userspace oom policies more reliably.

We can use one counter for number of pages allocated under pressure to
save struct task space and avoid two separate hierarchy walks on the hot
path.  The exact overage is calculated on return to user space, anyway.

Take the new high limit into account when determining if swap is "full".
Borrowing the explanation from Johannes:

  The idea behind "swap full" is that as long as the workload has plenty
  of swap space available and it's not changing its memory contents, it
  makes sense to generously hold on to copies of data in the swap device,
  even after the swapin.  A later reclaim cycle can drop the page without
  any IO.  Trading disk space for IO.

  But the only two ways to reclaim a swap slot is when they're faulted
  in and the references go away, or by scanning the virtual address space
  like swapoff does - which is very expensive (one could argue it's too
  expensive even for swapoff, it's often more practical to just reboot).

  So at some point in the fill level, we have to start freeing up swap
  slots on fault/swapin.  Otherwise we could eventually run out of swap
  slots while they're filled with copies of data that is also in RAM.

  We don't want to OOM a workload because its available swap space is
  filled with redundant cache.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Chris Down <chris@chrisdown.name>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200527195846.102707-5-kuba@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcg: move cgroup high memory limit setting into struct page_counter

High memory limit is currently recorded directly in struct mem_cgroup.
We are about to add a high limit for swap, move the field to struct
page_counter and add some helpers.

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Chris Down <chris@chrisdown.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/20200527195846.102707-4-kuba@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcg: move penalty delay clamping out of calculate_high_delay()

We will want to call calculate_high_delay() twice - once for memory and
once for swap, and we should apply the clamp value to sum of the
penalties.  Clamping has to be applied outside of calculate_high_delay().

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Chris Down <chris@chrisdown.name>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/20200527195846.102707-3-kuba@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/memcg: prepare for swap over-high accounting and penalty calculation

Patch series "memcg: Slow down swap allocation as the available space
gets depleted", v6.

Tejun describes the problem as follows:

When swap runs out, there's an abrupt change in system behavior - the
anonymous memory suddenly becomes unmanageable which readily breaks any
sort of memory isolation and can bring down the whole system.  To avoid
that, oomd [1] monitors free swap space and triggers kills when it drops
below the specific threshold (e.g.  15%).

While this works, it's far from ideal:

 - Depending on IO performance and total swap size, a given
   headroom might not be enough or too much.

 - oomd has to monitor swap depletion in addition to the usual
   pressure metrics and it currently doesn't consider memory.swap.max.

Solve this by adapting parts of the approach that memory.high uses -
slow down allocation as the resource gets depleted turning the depletion
behavior from abrupt cliff one to gradual degradation observable through
memory pressure metric.

[1] https://github.com/facebookincubator/oomd

This patch (of 4):

Slice the memory overage calculation logic a little bit so we can reuse
it to apply a similar penalty to the swap.  The logic which accesses the
memory-specific fields (use and high values) has to be taken out of
calculate_high_delay().

Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Chris Down <chris@chrisdown.name>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Link: http://lkml.kernel.org/r/20200527195846.102707-1-kuba@kernel.org
Link: http://lkml.kernel.org/r/20200527195846.102707-2-kuba@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: expose root cgroup's memory.stat

One way to measure the efficiency of memory reclaim is to look at the
ratio (pgscan+pfrefill)/pgsteal.  However at the moment these stats are
not updated consistently at the system level and the ratio of these are
not very meaningful.  The pgsteal and pgscan are updated for only global
reclaim while pgrefill gets updated for global as well as cgroup
reclaim.

Please note that this difference is only for system level vmstats.  The
cgroup stats returned by memory.stat are actually consistent.  The
cgroup's pgsteal contains number of reclaimed pages for global as well
as cgroup reclaim.  So, one way to get the system level stats is to get
these stats from root's memory.stat, so, expose memory.stat for the root
cgroup.

From Johannes Weiner:
There are subtle differences between /proc/vmstat and
memory.stat, and cgroup-aware code that wants to watch the full
hierarchy currently has to know about these intricacies and
translate semantics back and forth.

Generally having the fully recursive memory.stat at the root
level could help a broader range of usecases.

Why not fix the stats by including both the global and cgroup reclaim
activity instead of exposing root cgroup's memory.stat? The reason is
the benefit of having metrics exposing the activity that happens purely
due to machine capacity rather than localized activity that happens due
to the limits throughout the cgroup tree.  Additionally there are
userspace tools like sysstat(sar) which reads these stats to inform
about the system level reclaim activity.  So, we should not break such
use-cases.

Suggested-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Yafang Shao <laoar.shao@gmail.com>
Acked-by: Chris Down <chris@chrisdown.name>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Roman Gushchin <guro@fb.com>
Cc: Michal Hocko <mhocko@kernel.org>
Link: http://lkml.kernel.org/r/20200508170630.94406-1-shakeelb@google.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: memcontrol: simplify value comparison between count and limit

When the variables count and limit have the same value(count == limit),
the result of min(margin, limit - count) statement should be 0 and the
variable margin is set to 0.  So in this case, the min() statement is
not necessary and we can directly set the variable margin to 0.

Signed-off-by: Kaixu Xia <kaixuxia@tencent.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Link: http://lkml.kernel.org/r/1587479661-27237-1-git-send-email-kaixuxia@tencent.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, memcg: add workingset_restore in memory.stat

There's a new workingset counter introduced in commit 1899ad18c607 ("mm:
workingset: tell cache transitions from workingset thrashing").  With
the help of this counter we can know the workingset is transitioning or
thrashing.  To leverage the benifit of this counter to memcg, we should
introduce it into memory.stat.  Then we could know the workingset of the
workload inside a memcg better.

Bellow is the verification of this new counter in memory.stat.  Read a
file into the memory and then read it again to make these pages be
active.  The size of this file is 1G.  (memory.max is greater than file
size) The counters in memory.stat will be

inactive_file 0
active_file 1073639424

workingset_refault 0
workingset_activate 0
workingset_restore 0
workingset_nodereclaim 0

Trigger the memcg reclaim by setting a lower value to memory.high, and
then some pages will be demoted into inactive list, and then some pages
in the inactive list will be evicted into the storage.

inactive_file 498094080
active_file 310063104

workingset_refault 0
workingset_activate 0
workingset_restore 0
workingset_nodereclaim 0

Then recover the memory.high and read the file into memory again.  As a
result of it, the transitioning will occur.  Bellow is the result of
this transitioning,

inactive_file 498094080
active_file 575397888

workingset_refault 64746
workingset_activate 64746
workingset_restore 64746
workingset_nodereclaim 0

Signed-off-by: Yafang Shao <laoar.shao@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Chris Down <chris@chrisdown.name>
Cc: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Shakeel Butt <shakeelb@google.com>
Link: http://lkml.kernel.org/r/20200504153522.11553-1-laoar.shao@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/swap.h: delete meaningless __add_to_swap_cache() declaration

Since commit 8d93b41c09d1 ("mm: Convert add_to_swap_cache to XArray"),
__add_to_swap_cache and add_to_swap_cache are combined into one
function.  There is no __add_to_swap_cache() anymore.

Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Link: http://lkml.kernel.org/r/1590810326-2493-1-git-send-email-linmiaohe@huawei.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: swapfile: fix /proc/swaps heading and Size/Used/Priority alignment

Fix the heading and Size/Used/Priority field alignments in /proc/swaps.
If the Size and/or Used value is >= 10000000 (8 bytes), then the
alignment by using tab characters is broken.

This patch maintains the use of tabs for alignment.  If spaces are
preferred, we can just use a Field Width specifier for the bytes and
inuse fields.  That way those fields don't have to be a multiple of 8
bytes in width.  E.g., with a field width of 12, both Size and Used
would always fit on the first line of an 80-column wide terminal (only
Priority would be on the second line).

There are actually 2 problems: heading alignment and field width.  On an
xterm, if Used is 7 bytes in length, the tab does nothing, and the
display is like this, with no space/tab between the Used and Priority
fields.  (ugh)

Filename Type Size Used Priority
/dev/sda8                               partition 16779260 2023012-1

To be clear, if one does 'cat /proc/swaps >/tmp/proc.swaps', it does look
different, like so:

Filename Type Size Used Priority
/dev/sda8                               partition 16779260 2086988 -1

Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Link: http://lkml.kernel.org/r/c0ffb41a-81ac-ddfa-d452-a9229ecc0387@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

swap: reduce lock contention on swap cache from swap slots allocation

In some swap scalability test, it is found that there are heavy lock
contention on swap cache even if we have split one swap cache radix tree
per swap device to one swap cache radix tree every 64 MB trunk in commit
4b3ef9daa4fc ("mm/swap: split swap cache into 64MB trunks").

The reason is as follow.  After the swap device becomes fragmented so
that there's no free swap cluster, the swap device will be scanned
linearly to find the free swap slots.  swap_info_struct->cluster_next is
the next scanning base that is shared by all CPUs.  So nearby free swap
slots will be allocated for different CPUs.  The probability for
multiple CPUs to operate on the same 64 MB trunk is high.  This causes
the lock contention on the swap cache.

To solve the issue, in this patch, for SSD swap device, a percpu version
next scanning base (cluster_next_cpu) is added.  Every CPU will use its
own per-cpu next scanning base.  And after finishing scanning a 64MB
trunk, the per-cpu scanning base will be changed to the beginning of
another randomly selected 64MB trunk.  In this way, the probability for
multiple CPUs to operate on the same 64 MB trunk is reduced greatly.
Thus the lock contention is reduced too.  For HDD, because sequential
access is more important for IO performance, the original shared next
scanning base is used.

To test the patch, we have run 16-process pmbench memory benchmark on a
2-socket server machine with 48 cores.  One ram disk is configured as the
swap device per socket.  The pmbench working-set size is much larger than
the available memory so that swapping is triggered.  The memory read/write
ratio is 80/20 and the accessing pattern is random.  In the original
implementation, the lock contention on the swap cache is heavy.  The perf
profiling data of the lock contention code path is as following,

 _raw_spin_lock_irq.add_to_swap_cache.add_to_swap.shrink_page_list:      7.91
 _raw_spin_lock_irqsave.__remove_mapping.shrink_page_list:               7.11
 _raw_spin_lock.swapcache_free_entries.free_swap_slot.__swap_entry_free: 2.51
 _raw_spin_lock_irqsave.swap_cgroup_record.mem_cgroup_uncharge_swap:     1.66
 _raw_spin_lock_irq.shrink_inactive_list.shrink_lruvec.shrink_node:      1.29
 _raw_spin_lock.free_pcppages_bulk.drain_pages_zone.drain_pages:         1.03
 _raw_spin_lock_irq.shrink_active_list.shrink_lruvec.shrink_node:        0.93

After applying this patch, it becomes,

 _raw_spin_lock.swapcache_free_entries.free_swap_slot.__swap_entry_free: 3.58
 _raw_spin_lock_irq.shrink_inactive_list.shrink_lruvec.shrink_node:      2.3
 _raw_spin_lock_irqsave.swap_cgroup_record.mem_cgroup_uncharge_swap:     2.26
 _raw_spin_lock_irq.shrink_active_list.shrink_lruvec.shrink_node:        1.8
 _raw_spin_lock.free_pcppages_bulk.drain_pages_zone.drain_pages:         1.19

The lock contention on the swap cache is almost eliminated.

And the pmbench score increases 18.5%.  The swapin throughput increases
18.7% from 2.96 GB/s to 3.51 GB/s.  While the swapout throughput increases
18.5% from 2.99 GB/s to 3.54 GB/s.

We need really fast disk to show the benefit.  I have tried this on 2
Intel P3600 NVMe disks.  The performance improvement is only about 1%.
The improvement should be better on the faster disks, such as Intel Optane
disk.

[ying.huang@intel.com: fix cluster_next_cpu allocation and freeing, per Daniel]
Link: http://lkml.kernel.org/r/20200525002648.336325-1-ying.huang@intel.com
[ying.huang@intel.com: v4]
Link: http://lkml.kernel.org/r/20200529010840.928819-1-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200520031502.175659-1-ying.huang@intel.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: use prandom_u32_max()

To improve the code readability and take advantage of the common
implementation.

Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200512081013.520201-1-ying.huang@intel.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: __swap_entry_free() always free 1 entry

__swap_entry_free() always frees 1 entry.  Let's remove the usage.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/r/20200501015259.32237-2-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: classify SWAP_MAP_XXX to make it more readable

swap_info_struct->swap_map[] encodes some flag and count. And to
do some condition check, it also introduces some special values.

Currently those macros are defined with some magic order, which makes
audience hard to understand the exact meaning.

This patch split those macros into three categories:

    flag
    special value for first swap_map
    special value for continued swap_map

May this help audiences a little.

[akpm@linux-foundation.org: tweak capitalization in comments]
Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/r/20200501015259.32237-1-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

swap: try to scan more free slots even when fragmented

Now, the scalability of swap code will drop much when the swap device
becomes fragmented, because the swap slots allocation batching stops
working.  To solve the problem, in this patch, we will try to scan a
little more swap slots with restricted effort to batch the swap slots
allocation even if the swap device is fragmented.  Test shows that the
benchmark score can increase up to 37.1% with the patch.  Details are as
follows.

The swap code has a per-cpu cache of swap slots.  These batch swap space
allocations to improve swap subsystem scaling.  In the following code
path,

  add_to_swap()
    get_swap_page()
      refill_swap_slots_cache()
        get_swap_pages()
  scan_swap_map_slots()

scan_swap_map_slots() and get_swap_pages() can return multiple swap
slots for each call.  These slots will be cached in the per-CPU swap
slots cache, so that several following swap slot requests will be
fulfilled there to avoid the lock contention in the lower level swap
space allocation/freeing code path.

But this only works when there are free swap clusters.  If a swap device
becomes so fragmented that there's no free swap clusters,
scan_swap_map_slots() and get_swap_pages() will return only one swap
slot for each call in the above code path.  Effectively, this falls back
to the situation before the swap slots cache was introduced, the heavy
lock contention on the swap related locks kills the scalability.

Why does it work in this way? Because the swap device could be large,
and the free swap slot scanning could be quite time consuming, to avoid
taking too much time to scanning free swap slots, the conservative
method was used.

In fact, this can be improved via scanning a little more free slots with
strictly restricted effort.  Which is implemented in this patch.  In
scan_swap_map_slots(), after the first free swap slot is gotten, we will
try to scan a little more, but only if we haven't scanned too many slots
(< LATENCY_LIMIT).  That is, the added scanning latency is strictly
restricted.

To test the patch, we have run 16-process pmbench memory benchmark on a
2-socket server machine with 48 cores.  Multiple ram disks are
configured as the swap devices.  The pmbench working-set size is much
larger than the available memory so that swapping is triggered.  The
memory read/write ratio is 80/20 and the accessing pattern is random, so
the swap space becomes highly fragmented during the test.  In the
original implementation, the lock contention on swap related locks is
very heavy.  The perf profiling data of the lock contention code path is
as following,

 _raw_spin_lock.get_swap_pages.get_swap_page.add_to_swap:             21.03
 _raw_spin_lock_irq.shrink_inactive_list.shrink_lruvec.shrink_node:    1.92
 _raw_spin_lock_irq.shrink_active_list.shrink_lruvec.shrink_node:      1.72
 _raw_spin_lock.free_pcppages_bulk.drain_pages_zone.drain_pages:       0.69

While after applying this patch, it becomes,

 _raw_spin_lock_irq.shrink_inactive_list.shrink_lruvec.shrink_node:    4.89
 _raw_spin_lock_irq.shrink_active_list.shrink_lruvec.shrink_node:      3.85
 _raw_spin_lock.free_pcppages_bulk.drain_pages_zone.drain_pages:       1.1
 _raw_spin_lock_irqsave.pagevec_lru_move_fn.__lru_cache_add.do_swap_page: 0.88

That is, the lock contention on the swap locks is eliminated.

And the pmbench score increases 37.1%.  The swapin throughput increases
45.7% from 2.02 GB/s to 2.94 GB/s.  While the swapout throughput increases
45.3% from 2.04 GB/s to 2.97 GB/s.

Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200427030023.264780-1-ying.huang@intel.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: omit a duplicate code by compare tmp and max first

There are two duplicate code to handle the case when there is no available
swap entry.  To avoid this, we can compare tmp and max first and let the
second guard do its job.

No functional change is expected.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200421213824.8099-3-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: tmp is always smaller than max

If tmp is bigger or equal to max, we would jump to new_cluster.

Return true directly.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200421213824.8099-2-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: found_free could be represented by (tmp < max)

This is not necessary to use the variable found_free to record the status.
Just check tmp and max is enough.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200421213824.8099-1-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: remove the extra check in scan_swap_map_slots()

scan_swap_map_slots() is only called by scan_swap_map() and
get_swap_pages().  Both ensure nr would not exceed SWAP_BATCH.

Just remove it.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200325220309.9803-2-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: simplify the calculation of n_goal

Use min3() to simplify the comparison and make it more self-explaining.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200325220309.9803-1-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: remove the unnecessary goto for SSD case

Now we can see there is redundant goto for SSD case.  In these two places,
we can just let the code walk through to the correct tag instead of
explicitly jump to it.

Let's remove them for better readability.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/20200328060520.31449-4-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: explicitly show ssd/non-ssd is handled mutually exclusive

The code shows if this is ssd, it will jump to specific tag and skip the
following code for non-ssd.

Let's use "else if" to explicitly show the mutually exclusion for
ssd/non-ssd to reduce ambiguity.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/20200328060520.31449-3-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile.c: offset is only used when there is more slots

scan_swap_map_slots() is used to iterate swap_map[] array for an
available swap entry.  While after several optimizations, e.g.  for ssd
case, the logic of this function is a little not easy to catch.

This patchset tries to clean up the logic a little:

  * shows the ssd/non-ssd case is handled mutually exclusively
  * remove some unnecessary goto for ssd case

This patch (of 3):

When si->cluster_nr is zero, function would reach done and return.  The
increased offset would not be used any more.  This means we can move the
offset increment into the if clause.

This brings a further code cleanup possibility.

Signed-off-by: Wei Yang <richard.weiyang@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Link: http://lkml.kernel.org/r/20200328060520.31449-1-richard.weiyang@gmail.com
Link: http://lkml.kernel.org/r/20200328060520.31449-2-richard.weiyang@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: swap: properly update readahead statistics in unuse_pte_range()

In unuse_pte_range() we blindly swap-in pages without checking if the
swap entry is already present in the swap cache.

By doing this, the hit/miss ratio used by the swap readahead heuristic
is not properly updated and this leads to non-optimal performance during
swapoff.

Tracing the distribution of the readahead size returned by the swap
readahead heuristic during swapoff shows that a small readahead size is
used most of the time as if we had only misses (this happens both with
cluster and vma readahead), for example:

r::swapin_nr_pages(unsigned long offset):unsigned long:$retval
        COUNT      EVENT
        36948      $retval = 8
        44151      $retval = 4
        49290      $retval = 1
        527771     $retval = 2

Checking if the swap entry is present in the swap cache, instead, allows
to properly update the readahead statistics and the heuristic behaves in a
better way during swapoff, selecting a bigger readahead size:

r::swapin_nr_pages(unsigned long offset):unsigned long:$retval
        COUNT      EVENT
        1618       $retval = 1
        4960       $retval = 2
        41315      $retval = 4
        103521     $retval = 8

In terms of swapoff performance the result is the following:

Testing environment
===================

 - Host:
   CPU: 1.8GHz Intel Core i7-8565U (quad-core, 8MB cache)
   HDD: PC401 NVMe SK hynix 512GB
   MEM: 16GB

 - Guest (kvm):
   8GB of RAM
   virtio block driver
   16GB swap file on ext4 (/swapfile)

Test case
=========
 - allocate 85% of memory
 - `systemctl hibernate` to force all the pages to be swapped-out to the
   swap file
 - resume the system
 - measure the time that swapoff takes to complete:
   # /usr/bin/time swapoff /swapfile

Result (swapoff time)
======
                  5.6 vanilla   5.6 w/ this patch
                  -----------   -----------------
cluster-readahead      22.09s              12.19s
    vma-readahead      18.20s              15.33s

Conclusion
==========

The specific use case this patch is addressing is to improve swapoff
performance in cloud environments when a VM has been hibernated, resumed
and all the memory needs to be forced back to RAM by disabling swap.

This change allows to better exploits the advantages of the readahead
heuristic during swapoff and this improvement allows to to speed up the
resume process of such VMs.

[andrea.righi@canonical.com: update changelog]
Link: http://lkml.kernel.org/r/20200418084705.GA147642@xps-13
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Anchal Agarwal <anchalag@amazon.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Vineeth Remanan Pillai <vpillai@digitalocean.com>
Cc: Kelley Nielsen <kelleynnn@gmail.com>
Link: http://lkml.kernel.org/r/20200416180132.GB3352@xps-13
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swap_state: fix a data race in swapin_nr_pages

"prev_offset" is a static variable in swapin_nr_pages() that can be
accessed concurrently with only mmap_sem held in read mode as noticed by
KCSAN,

 BUG: KCSAN: data-race in swap_cluster_readahead / swap_cluster_readahead

 write to 0xffffffff92763830 of 8 bytes by task 14795 on cpu 17:
  swap_cluster_readahead+0x2a6/0x5e0
  swapin_readahead+0x92/0x8dc
  do_swap_page+0x49b/0xf20
  __handle_mm_fault+0xcfb/0xd70
  handle_mm_fault+0xfc/0x2f0
  do_page_fault+0x263/0x715
  page_fault+0x34/0x40

 1 lock held by (dnf)/14795:
  #0: ffff897bd2e98858 (&mm->mmap_sem#2){++++}-{3:3}, at: do_page_fault+0x143/0x715
  do_user_addr_fault at arch/x86/mm/fault.c:1405
  (inlined by) do_page_fault at arch/x86/mm/fault.c:1535
 irq event stamp: 83493
 count_memcg_event_mm+0x1a6/0x270
 count_memcg_event_mm+0x119/0x270
 __do_softirq+0x365/0x589
 irq_exit+0xa2/0xc0

 read to 0xffffffff92763830 of 8 bytes by task 1 on cpu 22:
  swap_cluster_readahead+0xfd/0x5e0
  swapin_readahead+0x92/0x8dc
  do_swap_page+0x49b/0xf20
  __handle_mm_fault+0xcfb/0xd70
  handle_mm_fault+0xfc/0x2f0
  do_page_fault+0x263/0x715
  page_fault+0x34/0x40

 1 lock held by systemd/1:
  #0: ffff897c38f14858 (&mm->mmap_sem#2){++++}-{3:3}, at: do_page_fault+0x143/0x715
 irq event stamp: 43530289
 count_memcg_event_mm+0x1a6/0x270
 count_memcg_event_mm+0x119/0x270
 __do_softirq+0x365/0x589
 irq_exit+0xa2/0xc0

Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Marco Elver <elver@google.com>
Cc: Hugh Dickins <hughd@google.com>
Link: http://lkml.kernel.org/r/20200402213748.2237-1-cai@lca.pw
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/swapfile: use list_{prev,next}_entry() instead of open-coding

Use list_{prev,next}_entry() instead of list_entry() for better
code readability.

Signed-off-by: chenqiwu <chenqiwu@xiaomi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: David Hildenbrand <david@redhat.com>
Cc: Wei Yang <richard.weiyang@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Pankaj Gupta <pankaj.gupta.linux@gmail.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Baoquan He <bhe@redhat.com>
Link: http://lkml.kernel.org/r/1586599916-15456-2-git-send-email-qiwuchen55@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/gup.c: further document vma_permits_fault()

Describe the caller's responsibilities when passing
FAULT_FLAG_ALLOW_RETRY.

Link: http://lkml.kernel.org/r/1586915606.5647.5.camel@mtkswgap22
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ivtv: convert get_user_pages() --> pin_user_pages()

This code was using get_user_pages*(), in a "Case 2" scenario
(DMA/RDMA), using the categorization from [1].  That means that it's
time to convert the get_user_pages*() + put_page() calls to
pin_user_pages*() + unpin_user_pages() calls.

There is some helpful background in [2]: basically, this is a small part
of fixing a long-standing disconnect between pinning pages, and file
systems' use of those pages.

[1] Documentation/core-api/pin_user_pages.rst

[2] "Explicit pinning of user-space pages":
    https://lwn.net/Articles/807108/

Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Walls <awalls@md.metrocast.net>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Link: http://lkml.kernel.org/r/20200518012157.1178336-3-jhubbard@nvidia.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/gup: introduce pin_user_pages_unlocked

Introduce pin_user_pages_unlocked(), which is nearly identical to the
get_user_pages_unlocked() that it wraps, except that it sets FOLL_PIN
and rejects FOLL_GET.

Signed-off-by: John Hubbard <jhubbard@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Andy Walls <awalls@md.metrocast.net>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Link: http://lkml.kernel.org/r/20200518012157.1178336-2-jhubbard@nvidia.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/gup.c: update the documentation

This patch is an attempt to update the documentation.

 - Add/ remove extra * based on type of function static/global.

 - Add description for functions and their input arguments.

[akpm@linux-foundation.org: s@/*@/**@]
Signed-off-by: Souptick Joarder <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/r/1588013630-4497-1-git-send-email-jrdr.linux@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/writeback: discard NR_UNSTABLE_NFS, use NR_WRITEBACK instead

After an NFS page has been written it is considered "unstable" until a
COMMIT request succeeds.  If the COMMIT fails, the page will be
re-written.

These "unstable" pages are currently accounted as "reclaimable", either
in WB_RECLAIMABLE, or in NR_UNSTABLE_NFS which is included in a
'reclaimable' count.  This might have made sense when sending the COMMIT
required a separate action by the VFS/MM (e.g.  releasepage() used to
send a COMMIT).  However now that all writes generated by ->writepages()
will automatically be followed by a COMMIT (since commit 919e3bd9a875
("NFS: Ensure we commit after writeback is complete")) it makes more
sense to treat them as writeback pages.

So this patch removes NR_UNSTABLE_NFS and accounts unstable pages in
NR_WRITEBACK and WB_WRITEBACK.

A particular effect of this change is that when
wb_check_background_flush() calls wb_over_bg_threshold(), the latter
will report 'true' a lot less often as the 'unstable' pages are no
longer considered 'dirty' (as there is nothing that writeback can do
about them anyway).

Currently wb_check_background_flush() will trigger writeback to NFS even
when there are relatively few dirty pages (if there are lots of unstable
pages), this can result in small writes going to the server (10s of
Kilobytes rather than a Megabyte) which hurts throughput.  With this
patch, there are fewer writes which are each larger on average.

Where the NR_UNSTABLE_NFS count was included in statistics
virtual-files, the entry is retained, but the value is hard-coded as
zero.  static trace points and warning printks which mentioned this
counter no longer report it.

[akpm@linux-foundation.org: re-layout comment]
[akpm@linux-foundation.org: fix printk warning]
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Acked-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Acked-by: Michal Hocko <mhocko@suse.com> [mm]
Cc: Christoph Hellwig <hch@lst.de>
Cc: Chuck Lever <chuck.lever@oracle.com>
Link: http://lkml.kernel.org/r/87d06j7gqa.fsf@notabene.neil.brown.name
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/writeback: replace PF_LESS_THROTTLE with PF_LOCAL_THROTTLE

PF_LESS_THROTTLE exists for loop-back nfsd (and a similar need in the
loop block driver and callers of prctl(PR_SET_IO_FLUSHER)), where a
daemon needs to write to one bdi (the final bdi) in order to free up
writes queued to another bdi (the client bdi).

The daemon sets PF_LESS_THROTTLE and gets a larger allowance of dirty
pages, so that it can still dirty pages after other processses have been
throttled.  The purpose of this is to avoid deadlock that happen when
the PF_LESS_THROTTLE process must write for any dirty pages to be freed,
but it is being thottled and cannot write.

This approach was designed when all threads were blocked equally,
independently on which device they were writing to, or how fast it was.
Since that time the writeback algorithm has changed substantially with
different threads getting different allowances based on non-trivial
heuristics.  This means the simple "add 25%" heuristic is no longer
reliable.

The important issue is not that the daemon needs a *larger* dirty page
allowance, but that it needs a *private* dirty page allowance, so that
dirty pages for the "client" bdi that it is helping to clear (the bdi
for an NFS filesystem or loop block device etc) do not affect the
throttling of the daemon writing to the "final" bdi.

This patch changes the heuristic so that the task is not throttled when
the bdi it is writing to has a dirty page count below below (or equal
to) the free-run threshold for that bdi.  This ensures it will always be
able to have some pages in flight, and so will not deadlock.

In a steady-state, it is expected that PF_LOCAL_THROTTLE tasks might
still be throttled by global threshold, but that is acceptable as it is
only the deadlock state that is interesting for this flag.

This approach of "only throttle when target bdi is busy" is consistent
with the other use of PF_LESS_THROTTLE in current_may_throttle(), were
it causes attention to be focussed only on the target bdi.

So this patch
 - renames PF_LESS_THROTTLE to PF_LOCAL_THROTTLE,
 - removes the 25% bonus that that flag gives, and
 - If PF_LOCAL_THROTTLE is set, don't delay at all unless the
   global and the local free-run thresholds are exceeded.

Note that previously realtime threads were treated the same as
PF_LESS_THROTTLE threads.  This patch does *not* change the behvaiour
for real-time threads, so it is now different from the behaviour of nfsd
and loop tasks.  I don't know what is wanted for realtime.

[akpm@linux-foundation.org: coding style fixes]
Signed-off-by: NeilBrown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Acked-by: Chuck Lever <chuck.lever@oracle.com> [nfsd]
Cc: Christoph Hellwig <hch@lst.de>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Trond Myklebust <trond.myklebust@hammerspace.com>
Link: http://lkml.kernel.org/r/87ftbf7gs3.fsf@notabene.neil.brown.name
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/page-writeback.c: remove unused variable

Commit 64081362e8ff ("mm/page-writeback.c: fix range_cyclic writeback
vs writepages deadlock") left unused variable, remove it.

Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: David Hildenbrand <david@redhat.com>
Link: http://lkml.kernel.org/r/20200528033740.17269-1-yuchao0@huawei.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/filemap.c: remove misleading comment

We no longer return 0 here and the comment doesn't tell us anything that
we don't already know (SIGBUS is a pretty good indicator that things
didn't work out).

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Link: http://lkml.kernel.org/r/20200529123243.20640-1-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm_types.h: change set_page_private to inline function

Change it to inline function to make callers use the proper argument.  And
no need for it to be macro per Andrew's comment [1].

[1] https://lore.kernel.org/lkml/20200518221235.1fa32c38e5766113f78e3f0d@linux-foundation.org/

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/r/20200525203149.18802-1-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/migrate.c: call detach_page_private to cleanup code

We can cleanup code a little by call detach_page_private here.

[akpm@linux-foundation.org: use attach_page_private(), per Dave]
  http://lkml.kernel.org/r/20200521225220.GV2005@dread.disaster.area
[akpm@linux-foundation.org: clear PagePrivate]
Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200519214049.15179-1-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

buffer_head.h: remove attach_page_buffers

All the callers have replaced attach_page_buffers with the new function
attach_page_private, so remove it.

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andreas Dilger <adilger@dilger.ca>
Link: http://lkml.kernel.org/r/20200517214718.468-10-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

orangefs: use attach/detach_page_private

Since the new pair function is introduced, we can call them to clean the
code in orangefs.

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Mike Marshall <hubcap@omnibond.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Martin Brandenburg <martin@omnibond.com>
Link: http://lkml.kernel.org/r/20200517214718.468-9-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ntfs: replace attach_page_buffers with attach_page_private

Call the new function since attach_page_buffers will be removed.

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Anton Altaparmakov <anton@tuxera.com>
Link: http://lkml.kernel.org/r/20200517214718.468-8-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

iomap: use attach/detach_page_private

Since the new pair function is introduced, we can call them to clean the
code in iomap.

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Link: http://lkml.kernel.org/r/20200517214718.468-7-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

f2fs: use attach/detach_page_private

Since the new pair function is introduced, we can call them to clean the
code in f2fs.h.

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Chao Yu <yuchao0@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Link: http://lkml.kernel.org/r/20200517214718.468-6-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/buffer.c: use attach/detach_page_private

Since the new pair function is introduced, we can call them to clean the
code in buffer.c.

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Link: http://lkml.kernel.org/r/20200517214718.468-5-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

btrfs: use attach/detach_page_private

Since the new pair function is introduced, we can call them to clean the
code in btrfs.

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: David Sterba <dsterba@suse.com>
Cc: Chris Mason <clm@fb.com>
Cc: Josef Bacik <josef@toxicpanda.com>
Link: http://lkml.kernel.org/r/20200517214718.468-4-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

md: remove __clear_page_buffers and use attach/detach_page_private

After introduction attach/detach_page_private in pagemap.h, we can remove
the duplicated code and call the new functions.

Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Song Liu <song@kernel.org>
Link: http://lkml.kernel.org/r/20200517214718.468-3-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/pagemap.h: introduce attach/detach_page_private

Patch series "Introduce attach/detach_page_private to cleanup code".

This patch (of 10):

The logic in attach_page_buffers and __clear_page_buffers are quite
paired, but

1. they are located in different files.

2. attach_page_buffers is implemented in buffer_head.h, so it could be
   used by other files. But __clear_page_buffers is static function in
   buffer.c and other potential users can't call the function, md-bitmap
   even copied the function.

So, introduce the new attach/detach_page_private to replace them.  With
the new pair of function, we will remove the usage of attach_page_buffers
and __clear_page_buffers in next patches.  Thanks for suggestions about
the function name from Alexander Viro, Andreas Grünbacher, Christoph
Hellwig and Matthew Wilcox.

Suggested-by: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Guoqing Jiang <guoqing.jiang@cloud.ionos.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: "Darrick J. Wong" <darrick.wong@oracle.com>
Cc: William Kucharski <william.kucharski@oracle.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Andreas Gruenbacher <agruenba@redhat.com>
Cc: Yang Shi <yang.shi@linux.alibaba.com>
Cc: Yafang Shao <laoar.shao@gmail.com>
Cc: Song Liu <song@kernel.org>
Cc: Chris Mason <clm@fb.com>
Cc: Josef Bacik <josef@toxicpanda.com>
Cc: David Sterba <dsterba@suse.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Chao Yu <chao@kernel.org>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Anton Altaparmakov <anton@tuxera.com>
Cc: Mike Marshall <hubcap@omnibond.com>
Cc: Martin Brandenburg <martin@omnibond.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: Roman Gushchin <guro@fb.com>
Cc: Andreas Dilger <adilger@dilger.ca>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Dave Chinner <david@fromorbit.com>
Link: http://lkml.kernel.org/r/20200517214718.468-1-guoqing.jiang@cloud.ionos.com
Link: http://lkml.kernel.org/r/20200517214718.468-2-guoqing.jiang@cloud.ionos.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

iomap: convert from readpages to readahead

Use the new readahead operation in iomap.  Convert XFS and ZoneFS to use
it.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <darrick.wong@oracle.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-26-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fuse: convert from readpages to readahead

Implement the new readahead operation in fuse by using __readahead_batch()
to fill the array of pages in fuse_args_pages directly.  This lets us
inline fuse_readpages_fill() into fuse_readahead().

[willy@infradead.org: build fix]
Link: http://lkml.kernel.org/r/20200415025938.GB5820@bombadil.infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Acked-by: Miklos Szeredi <mszeredi@redhat.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-25-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

f2fs: pass the inode to f2fs_mpage_readpages

This function now only uses the mapping argument to look up the inode, and
both callers already have the inode, so just pass the inode instead of the
mapping.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Acked-by: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-24-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

f2fs: convert from readpages to readahead

Use the new readahead operation in f2fs

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Acked-by: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-23-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ext4: pass the inode to ext4_mpage_readpages

This function now only uses the mapping argument to look up the inode, and
both callers already have the inode, so just pass the inode instead of the
mapping.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-22-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ext4: convert from readpages to readahead

Use the new readahead operation in ext4

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-21-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

erofs: convert compressed files from readpages to readahead

Use the new readahead operation in erofs.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Acked-by: Gao Xiang <gaoxiang25@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-20-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

erofs: convert uncompressed files from readpages to readahead

Use the new readahead operation in erofs

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Acked-by: Gao Xiang <gaoxiang25@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-19-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

btrfs: convert from readpages to readahead

Implement the new readahead method in btrfs using the new
readahead_page_batch() function.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-18-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs: convert mpage_readpages to mpage_readahead

Implement the new readahead aop and convert all callers (block_dev,
exfat, ext2, fat, gfs2, hpfs, isofs, jfs, nilfs2, ocfs2, omfs, qnx6,
reiserfs & udf).

The callers are all trivial except for GFS2 & OCFS2.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Junxiao Bi <junxiao.bi@oracle.com> # ocfs2
Reviewed-by: Joseph Qi <joseph.qi@linux.alibaba.com> # ocfs2
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-17-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: use memalloc_nofs_save in readahead path

Ensure that memory allocations in the readahead path do not attempt to
reclaim file-backed pages, which could lead to a deadlock.  It is
possible, though unlikely this is the root cause of a problem observed
by Cong Wang.

Reported-by: Cong Wang <xiyou.wangcong@gmail.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-16-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: document why we don't set PageReadahead

If the page is already in cache, we don't set PageReadahead on it.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-15-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add page_cache_readahead_unbounded

ext4 and f2fs have duplicated the guts of the readahead code so they can
read past i_size.  Instead, separate out the guts of the readahead code
so they can call it directly.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-14-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move end_index check out of readahead loop

By reducing nr_to_read, we can eliminate this check from inside the loop.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-13-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add readahead address space operation

This replaces ->readpages with a saner interface:
 - Return void instead of an ignored error code.
 - Page cache is already populated with locked pages when ->readahead
   is called.
 - New arguments can be passed to the implementation without changing
   all the filesystems that use a common helper function like
   mpage_readahead().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-12-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: put readahead pages in cache earlier

When populating the page cache for readahead, mappings that use
->readpages must populate the page cache themselves as the pages are
passed on a linked list which would normally be used for the page
cache's LRU.  For mappings that use ->readpage or the upcoming
->readahead method, we can put the pages into the page cache as soon as
they're allocated, which solves a race between readahead and direct IO.
It also lets us remove the gfp argument from read_pages().

Use the new readahead_page() API to implement the repeated calls to
->readpage(), just like most filesystems will.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-11-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove 'page_offset' from readahead loop

Replace the page_offset variable with 'index + i'.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-10-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: rename readahead loop variable to 'i'

Change the type of page_idx to unsigned long, and rename it -- it's just
a loop counter, not a page index.

Suggested-by: John Hubbard <jhubbard@nvidia.com>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-9-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: rename various 'offset' parameters to 'index'

The word 'offset' is used ambiguously to mean 'byte offset within a
page', 'byte offset from the start of the file' and 'page offset from
the start of the file'.

Use 'index' to mean 'page offset from the start of the file' throughout
the readahead code.

[ We should probably rename the 'pgoff_t' type to 'pgidx_t' too - Linus ]

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-8-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: use readahead_control to pass arguments

In this patch, only between __do_page_cache_readahead() and
read_pages(), but it will be extended in upcoming patches.  The
read_pages() function becomes aops centric, as this makes the most sense
by the end of the patchset.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-7-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add new readahead_control API

Filesystems which implement the upcoming ->readahead method will get
their pages by calling readahead_page() or readahead_page_batch().
These functions support large pages, even though none of the filesystems
to be converted do yet.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-6-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move readahead nr_pages check into read_pages

Simplify the callers by moving the check for nr_pages and the BUG_ON
into read_pages().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-5-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: ignore return value of ->readpages

We used to assign the return value to a variable, which we then ignored.
Remove the pretence of caring.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-4-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: return void from various readahead functions

ondemand_readahead has two callers, neither of which use the return
value.  That means that both ra_submit and __do_page_cache_readahead()
can return void, and we don't need to worry that a present page in the
readahead window causes us to return a smaller nr_pages than we ought to
have.

Similarly, no caller uses the return value from
force_page_cache_readahead().

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-3-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move readahead prototypes from mm.h

Patch series "Change readahead API", v11.

This series adds a readahead address_space operation to replace the
readpages operation.  The key difference is that pages are added to the
page cache as they are allocated (and then looked up by the filesystem)
instead of passing them on a list to the readpages operation and having
the filesystem add them to the page cache.  It's a net reduction in code
for each implementation, more efficient than walking a list, and solves
the direct-write vs buffered-read problem reported by yu kuai at
http://lkml.kernel.org/r/20200116063601.39201-1-yukuai3@huawei.com

The only unconverted filesystems are those which use fscache.  Their
conversion is pending Dave Howells' rewrite which will make the
conversion substantially easier.  This should be completed by the end of
the year.

I want to thank the reviewers/testers; Dave Chinner, John Hubbard, Eric
Biggers, Johannes Thumshirn, Dave Sterba, Zi Yan, Christoph Hellwig and
Miklos Szeredi have done a marvellous job of providing constructive
criticism.

These patches pass an xfstests run on ext4, xfs & btrfs with no
regressions that I can tell (some of the tests seem a little flaky
before and remain flaky afterwards).

This patch (of 25):

The readahead code is part of the page cache so should be found in the
pagemap.h file.  force_page_cache_readahead is only used within mm, so
move it to mm/internal.h instead.  Remove the parameter names where they
add no value, and rename the ones which were actively misleading.

Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: William Kucharski <william.kucharski@oracle.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: Cong Wang <xiyou.wangcong@gmail.com>
Cc: Darrick J. Wong <darrick.wong@oracle.com>
Cc: Dave Chinner <dchinner@redhat.com>
Cc: Eric Biggers <ebiggers@google.com>
Cc: Gao Xiang <gaoxiang25@huawei.com>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Joseph Qi <joseph.qi@linux.alibaba.com>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Miklos Szeredi <mszeredi@redhat.com>
Link: http://lkml.kernel.org/r/20200414150233.24495-1-willy@infradead.org
Link: http://lkml.kernel.org/r/20200414150233.24495-2-willy@infradead.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm, dump_page(): do not crash with invalid mapping pointer

We have seen a following problem on a RPi4 with 1G RAM:

    BUG: Bad page state in process systemd-hwdb  pfn:35601
    page:ffff7e0000d58040 refcount:15 mapcount:131221 mapping:efd8fe765bc80080 index:0x1 compound_mapcount: -32767
    Unable to handle kernel paging request at virtual address efd8fe765bc80080
    Mem abort info:
      ESR = 0x96000004
      Exception class = DABT (current EL), IL = 32 bits
      SET = 0, FnV = 0
      EA = 0, S1PTW = 0
    Data abort info:
      ISV = 0, ISS = 0x00000004
      CM = 0, WnR = 0
    [efd8fe765bc80080] address between user and kernel address ranges
    Internal error: Oops: 96000004 [#1] SMP
    Modules linked in: btrfs libcrc32c xor xor_neon zlib_deflate raid6_pq mmc_block xhci_pci xhci_hcd usbcore sdhci_iproc sdhci_pltfm sdhci mmc_core clk_raspberrypi gpio_raspberrypi_exp pcie_brcmstb bcm2835_dma gpio_regulator phy_generic fixed sg scsi_mod efivarfs
    Supported: No, Unreleased kernel
    CPU: 3 PID: 408 Comm: systemd-hwdb Not tainted 5.3.18-8-default #1 SLE15-SP2 (unreleased)
    Hardware name: raspberrypi rpi/rpi, BIOS 2020.01 02/21/2020
    pstate: 40000085 (nZcv daIf -PAN -UAO)
    pc : __dump_page+0x268/0x368
    lr : __dump_page+0xc4/0x368
    sp : ffff000012563860
    x29: ffff000012563860 x28: ffff80003ddc4300
    x27: 0000000000000010 x26: 000000000000003f
    x25: ffff7e0000d58040 x24: 000000000000000f
    x23: efd8fe765bc80080 x22: 0000000000020095
    x21: efd8fe765bc80080 x20: ffff000010ede8b0
    x19: ffff7e0000d58040 x18: ffffffffffffffff
    x17: 0000000000000001 x16: 0000000000000007
    x15: ffff000011689708 x14: 3030386362353637
    x13: 6566386466653a67 x12: 6e697070616d2031
    x11: 32323133313a746e x10: 756f6370616d2035
    x9 : ffff00001168a840 x8 : ffff00001077a670
    x7 : 000000000000013d x6 : ffff0000118a43b5
    x5 : 0000000000000001 x4 : ffff80003dd9e2c8
    x3 : ffff80003dd9e2c8 x2 : 911c8d7c2f483500
    x1 : dead000000000100 x0 : efd8fe765bc80080
    Call trace:
     __dump_page+0x268/0x368
     bad_page+0xd4/0x168
     check_new_page_bad+0x80/0xb8
     rmqueue_bulk.constprop.26+0x4d8/0x788
     get_page_from_freelist+0x4d4/0x1228
     __alloc_pages_nodemask+0x134/0xe48
     alloc_pages_vma+0x198/0x1c0
     do_anonymous_page+0x1a4/0x4d8
     __handle_mm_fault+0x4e8/0x560
     handle_mm_fault+0x104/0x1e0
     do_page_fault+0x1e8/0x4c0
     do_translation_fault+0xb0/0xc0
     do_mem_abort+0x50/0xb0
     el0_da+0x24/0x28
    Code: f9401025 8b8018a0 9a851005 17ffffca (f94002a0)

Besides the underlying issue with page->mapping containing a bogus value
for some reason, we can see that __dump_page() crashed by trying to read
the pointer at mapping->host, turning a recoverable warning into full
Oops.

It can be expected that when page is reported as bad state for some
reason, the pointers there should not be trusted blindly.

So this patch treats all data in __dump_page() that depends on
page->mapping as lava, using probe_kernel_read_strict().  Ideally this
would include the dentry->d_parent recursively, but that would mean
changing printk handler for %pd.  Chances of reaching the dentry
printing part with an initially bogus mapping pointer should be rather
low, though.

Also prefix printing mapping->a_ops with a description of what is being
printed.  In case the value is bogus, %ps will print raw value instead
of the symbol name and then it's not obvious at all that it's printing
a_ops.

Reported-by: Petr Tesarik <ptesarik@suse.cz>
Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: John Hubbard <jhubbard@nvidia.com>
Link: http://lkml.kernel.org/r/20200331165454.12263-1-vbabka@suse.cz
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Documentation/vm/slub.rst: s/Toggle/Enable/

"toggle" means to change a boolean thing's state.  This operation
doesn't do that - it sets it to "true".

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Rafael Aquini <aquini@redhat.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Pekka Enberg <penberg@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/slub: fix stack overruns with SLUB_STATS

There is no need to copy SLUB_STATS items from root memcg cache to new
memcg cache copies.  Doing so could result in stack overruns because the
store function only accepts 0 to clear the stat and returns an error for
everything else while the show method would print out the whole stat.

Then, the mismatch of the lengths returns from show and store methods
happens in memcg_propagate_slab_attrs():

else if (root_cache->max_attr_size < ARRAY_SIZE(mbuf))
buf = mbuf;

max_attr_size is only 2 from slab_attr_store(), then, it uses mbuf[64]
in show_stat() later where a bounch of sprintf() would overrun the stack
variable.  Fix it by always allocating a page of buffer to be used in
show_stat() if SLUB_STATS=y which should only be used for debug purpose.

  # echo 1 > /sys/kernel/slab/fs_cache/shrink
  BUG: KASAN: stack-out-of-bounds in number+0x421/0x6e0
  Write of size 1 at addr ffffc900256cfde0 by task kworker/76:0/53251

  Hardware name: HPE ProLiant DL385 Gen10/ProLiant DL385 Gen10, BIOS A40 07/10/2019
  Workqueue: memcg_kmem_cache memcg_kmem_cache_create_func
  Call Trace:
    number+0x421/0x6e0
    vsnprintf+0x451/0x8e0
    sprintf+0x9e/0xd0
    show_stat+0x124/0x1d0
    alloc_slowpath_show+0x13/0x20
    __kmem_cache_create+0x47a/0x6b0

  addr ffffc900256cfde0 is located in stack of task kworker/76:0/53251 at offset 0 in frame:
   process_one_work+0x0/0xb90

  this frame has 1 object:
   [32, 72) 'lockdep_map'

  Memory state around the buggy address:
   ffffc900256cfc80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
   ffffc900256cfd00: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  >ffffc900256cfd80: 00 00 00 00 00 00 00 00 00 00 00 00 f1 f1 f1 f1
                                                         ^
   ffffc900256cfe00: 00 00 00 00 00 f2 f2 f2 00 00 00 00 00 00 00 00
   ffffc900256cfe80: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
  ==================================================================
  Kernel panic - not syncing: stack-protector: Kernel stack is corrupted in: __kmem_cache_create+0x6ac/0x6b0
  Workqueue: memcg_kmem_cache memcg_kmem_cache_create_func
  Call Trace:
    __kmem_cache_create+0x6ac/0x6b0

Fixes: 107dab5c92d5 ("slub: slub-specific propagation changes")
Signed-off-by: Qian Cai <cai@lca.pw>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Glauber Costa <glauber@scylladb.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>
Link: http://lkml.kernel.org/r/20200429222356.4322-1-cai@lca.pw
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

slub: remove kmalloc under list_lock from list_slab_objects() V2

list_slab_objects() is called when a slab is destroyed and there are
objects still left to list the objects in the syslog.  This is a pretty
rare event.

And there it seems we take the list_lock and call kmalloc while holding
that lock.

Perform the allocation in free_partial() before the list_lock is taken.

Fixes: bbd7d57bfe852d9788bae5fb171c7edb4021d8ac ("slub: Potential stack overflow")
Signed-off-by: Christopher Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Cc: Yu Zhao <yuzhao@google.com>
Link: http://lkml.kernel.org/r/alpine.DEB.2.21.2002031721250.1668@www.lameter.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

slub: Remove userspace notifier for cache add/remove

I came across some unnecessary uevents once again which reminded me
this.  The patch seems to be lost in the leaves of the original
discussion [1], so resending.

[1] https://lore.kernel.org/r/alpine.DEB.2.21.2001281813130.745@www.lameter.com

Kmem caches are internal kernel structures so it is strange that
userspace notifiers would be needed.  And I am not aware of any use of
these notifiers.  These notifiers may just exist because in the initial
slub release the sysfs code was copied from another subsystem.

Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Michal Koutný <mkoutny@suse.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Link: http://lkml.kernel.org/r/20200423115721.19821-1-mkoutny@suse.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/slub.c: fix corrupted freechain in deactivate_slab()

The slub_debug is able to fix the corrupted slab freelist/page.
However, alloc_debug_processing() only checks the validity of current
and next freepointer during allocation path.  As a result, once some
objects have their freepointers corrupted, deactivate_slab() may lead to
page fault.

Below is from a test kernel module when 'slub_debug=PUF,kmalloc-128
slub_nomerge'.  The test kernel corrupts the freepointer of one free
object on purpose.  Unfortunately, deactivate_slab() does not detect it
when iterating the freechain.

  BUG: unable to handle page fault for address: 00000000123456f8
  #PF: supervisor read access in kernel mode
  #PF: error_code(0x0000) - not-present page
  PGD 0 P4D 0
  Oops: 0000 [#1] SMP PTI
  ... ...
  RIP: 0010:deactivate_slab.isra.92+0xed/0x490
  ... ...
  Call Trace:
   ___slab_alloc+0x536/0x570
   __slab_alloc+0x17/0x30
   __kmalloc+0x1d9/0x200
   ext4_htree_store_dirent+0x30/0xf0
   htree_dirblock_to_tree+0xcb/0x1c0
   ext4_htree_fill_tree+0x1bc/0x2d0
   ext4_readdir+0x54f/0x920
   iterate_dir+0x88/0x190
   __x64_sys_getdents+0xa6/0x140
   do_syscall_64+0x49/0x170
   entry_SYSCALL_64_after_hwframe+0x44/0xa9

Therefore, this patch adds extra consistency check in deactivate_slab().
Once an object's freepointer is corrupted, all following objects
starting at this object are isolated.

[akpm@linux-foundation.org: fix build with CONFIG_SLAB_DEBUG=n]
Signed-off-by: Dongli Zhang <dongli.zhang@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Joe Jin <joe.jin@oracle.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>
Link: http://lkml.kernel.org/r/20200331031450.12182-1-dongli.zhang@oracle.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

usercopy: mark dma-kmalloc caches as usercopy caches

We have seen a "usercopy: Kernel memory overwrite attempt detected to
SLUB object 'dma-kmalloc-1 k' (offset 0, size 11)!" error on s390x, as
IUCV uses kmalloc() with __GFP_DMA because of memory address
restrictions.  The issue has been discussed [2] and it has been noted
that if all the kmalloc caches are marked as usercopy, there's little
reason not to mark dma-kmalloc caches too.  The 'dma' part merely means
that __GFP_DMA is used to restrict memory address range.

As Jann Horn put it [3]:
 "I think dma-kmalloc slabs should be handled the same way as normal
  kmalloc slabs. When a dma-kmalloc allocation is freshly created, it is
  just normal kernel memory - even if it might later be used for DMA -,
  and it should be perfectly fine to copy_from_user() into such
  allocations at that point, and to copy_to_user() out of them at the
  end. If you look at the places where such allocations are created, you
  can see things like kmemdup(), memcpy() and so on - all normal
  operations that shouldn't conceptually be different from usercopy in
  any relevant way."

Thus this patch marks the dma-kmalloc-* caches as usercopy.

[1] https://bugzilla.suse.com/show_bug.cgi?id=1156053
[2] https://lore.kernel.org/kernel-hardening/bfca96db-bbd0-d958-7732-76e36c667c68@suse.cz/
[3] https://lore.kernel.org/kernel-hardening/CAG48ez1a4waGk9kB0WLaSbs4muSoK0AYAVk8=XYaKj4_+6e6Hg@mail.gmail.com/

Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Christian Borntraeger <borntraeger@de.ibm.com>
Acked-by: Jiri Slaby <jslaby@suse.cz>
Cc: Jann Horn <jannh@google.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Christopher Lameter <cl@linux.com>
Cc: Julian Wiedmann <jwi@linux.ibm.com>
Cc: Ursula Braun <ubraun@linux.ibm.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: David Windsor <dave@nullcore.net>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Andy Lutomirski <luto@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Laura Abbott <labbott@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: "Martin K. Petersen" <martin.petersen@oracle.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Christoffer Dall <christoffer.dall@linaro.org>
Cc: Dave Kleikamp <dave.kleikamp@oracle.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Luis de Bethencourt <luisbg@kernel.org>
Cc: Marc Zyngier <marc.zyngier@arm.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Matthew Garrett <mjg59@google.com>
Cc: Michal Kubecek <mkubecek@suse.cz>
Link: http://lkml.kernel.org/r/7d810f6d-8085-ea2f-7805-47ba3842dc50@suse.cz
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/buffer.c: record blockdev write errors in super_block that it backs

When syncing out a block device (a'la __sync_blockdev), any error
encountered will only be recorded in the bd_inode's mapping.  When the
blockdev contains a filesystem however, we'd like to also record the
error in the super_block that's stored there.

Make mark_buffer_write_io_error also record the error in the
corresponding super_block when a writeback error occurs and the block
device contains a mounted superblock.

Since superblocks are RCU freed, hold the rcu_read_lock to ensure that
the superblock doesn't go away while we're marking it.

Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Andres Freund <andres@anarazel.de>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Link: http://lkml.kernel.org/r/20200428135155.19223-3-jlayton@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

vfs: track per-sb writeback errors and report them to syncfs

Patch series "vfs: have syncfs() return error when there are writeback
errors", v6.

Currently, syncfs does not return errors when one of the inodes fails to
be written back.  It will return errors based on the legacy AS_EIO and
AS_ENOSPC flags when syncing out the block device fails, but that's not
particularly helpful for filesystems that aren't backed by a blockdev.
It's also possible for a stray sync to lose those errors.

The basic idea in this set is to track writeback errors at the
superblock level, so that we can quickly and easily check whether
something bad happened without having to fsync each file individually.
syncfs is then changed to reliably report writeback errors after they
occur, much in the same fashion as fsync does now.

This patch (of 2):

Usually we suggest that applications call fsync when they want to ensure
that all data written to the file has made it to the backing store, but
that can be inefficient when there are a lot of open files.

Calling syncfs on the filesystem can be more efficient in some
situations, but the error reporting doesn't currently work the way most
people expect.  If a single inode on a filesystem reports a writeback
error, syncfs won't necessarily return an error.  syncfs only returns an
error if __sync_blockdev fails, and on some filesystems that's a no-op.

It would be better if syncfs reported an error if there were any
writeback failures.  Then applications could call syncfs to see if there
are any errors on any open files, and could then call fsync on all of
the other descriptors to figure out which one failed.

This patch adds a new errseq_t to struct super_block, and has
mapping_set_error also record writeback errors there.

To report those errors, we also need to keep an errseq_t in struct file
to act as a cursor.  This patch adds a dedicated field for that purpose,
which slots nicely into 4 bytes of padding at the end of struct file on
x86_64.

An earlier version of this patch used an O_PATH file descriptor to cue
the kernel that the open file should track the superblock error and not
the inode's writeback error.

I think that API is just too weird though.  This is simpler and should
make syncfs error reporting "just work" even if someone is multiplexing
fsync and syncfs on the same fds.

Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Jan Kara <jack@suse.cz>
Cc: Andres Freund <andres@anarazel.de>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: David Howells <dhowells@redhat.com>
Link: http://lkml.kernel.org/r/20200428135155.19223-1-jlayton@kernel.org
Link: http://lkml.kernel.org/r/20200428135155.19223-2-jlayton@kernel.org
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

arch/parisc/include/asm/pgtable.h: remove unused `old_pte'

parisc's set_pte_at() macro has set-but-not-used variable:

  include/linux/pgtable.h: In function 'pte_clear_not_present_full':
  arch/parisc/include/asm/pgtable.h:96:9: warning: variable 'old_pte' set but not used [-Wunused-but-set-variable]

Reported-by: kbuild test robot <lkp@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: "James E.J. Bottomley" <James.Bottomley@HansenPartnership.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: mount shared volume without ha stack

Usually we create and use a ocfs2 shared volume on the top of ha stack.
For pcmk based ha stack, which includes DLM, corosync and pacemaker
services.

The customers complained they could not mount existent ocfs2 volume in
the single node without ha stack, e.g.  single node backup/restore
scenario.

Like this case, the customers just want to access the data from the
existent ocfs2 volume quickly, but do not want to restart or setup ha
stack.

Then, I'd like to add a mount option "nocluster", if the users use this
option to mount a ocfs2 shared volume, the whole mount will not depend
on the ha related services.  the command will mount the existent ocfs2
volume directly (like local mount), for avoiding setup the ha stack.

Signed-off-by: Gang He <ghe@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Jun Piao <piaojun@huawei.com>
Link: http://lkml.kernel.org/r/20200423053300.22661-1-ghe@suse.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ocfs2: add missing annotation for dlm_empty_lockres()

Sparse reports a warning at dlm_empty_lockres()

  warning: context imbalance in dlm_purge_lockres() - unexpected unlock

The root cause is the missing annotation at dlm_purge_lockres()

Add the missing __must_hold(&dlm->spinlock)

Signed-off-by: Jules Irenge <jbi.octave@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Fasheh <mark@fasheh.com>
Cc: Joel Becker <jlbec@evilplan.org>
Cc: Junxiao Bi <junxiao.bi@oracle.com>
Cc: Joseph Qi <jiangqi903@gmail.com>
Cc: Changwei Ge <gechangwei@live.cn>
Cc: Gang He <ghe@suse.com>
Cc: Jun Piao <piaojun@huawei.com>
Link: http://lkml.kernel.org/r/20200403160505.2832-4-jbi.octave@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

squashfs: migrate from ll_rw_block usage to BIO

ll_rw_block() function has been deprecated in favor of BIO which appears
to come with large performance improvements.

This patch decreases boot time by close to 40% when using squashfs for
the root file-system.  This is observed at least in the context of
starting an Android VM on Chrome OS using crosvm.  The patch was tested
on 4.19 as well as master.

This patch is largely based on Adrien Schildknecht's patch that was
originally sent as https://lkml.org/lkml/2017/9/22/814 though with some
significant changes and simplifications while also taking Phillip
Lougher's feedback into account, around preserving support for
FILE_CACHE in particular.

[akpm@linux-foundation.org: fix build error reported by Randy]
Link: http://lkml.kernel.org/r/319997c2-5fc8-f889-2ea3-d913308a7c1f@infradead.org
Signed-off-by: Philippe Liard <pliard@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Adrien Schildknecht <adrien+dev@schischi.me>
Cc: Phillip Lougher <phillip@squashfs.org.uk>
Cc: Guenter Roeck <groeck@chromium.org>
Cc: Daniel Rosenberg <drosen@google.com>
Link: https://chromium.googlesource.com/chromiumos/platform/crosvm
Link: http://lkml.kernel.org/r/20191106074238.186023-1-pliard@google.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge tag 'x86_cache_updates_for_5.8' of git://git./linux/kernel/git/tip/tip

Pull x86 cache resource control updates from Borislav Petkov:
 "Add support for wider Memory Bandwidth Monitoring counters by querying
  their width from CPUID.

  As a prerequsite for that, streamline and unify the CPUID detection of
  the respective resource control attributes.

  By Reinette Chatre"

* tag 'x86_cache_updates_for_5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/resctrl: Support wider MBM counters
  x86/resctrl: Support CPUID enumeration of MBM counter width
  x86/resctrl: Maintain MBM counter width per resource
  x86/resctrl: Query LLC monitoring properties once during boot
  x86/resctrl: Remove unnecessary RMID checks
  x86/cpu: Move resctrl CPUID code to resctrl/
  x86/resctrl: Rename asm/resctrl_sched.h to asm/resctrl.h

Merge tag 'x86_microcode_for_5.8' of git://git./linux/kernel/git/tip/tip

Pull x86 microcode update from Borislav Petkov:
 "A single fix for late microcode loading to handle the correct return
  value from stop_machine(), from Mihai Carabas"

* tag 'x86_microcode_for_5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/microcode: Fix return value for microcode late loading

Merge tag 'edac_updates_for_5.8' of git://git./linux/kernel/git/ras/ras

Pull EDAC updates from Borislav Petkov:

 - Fix i10nm_edac loading on some Ice Lake and Tremont/Jacobsville
   steppings due to the offset change of the bus number configuration
   register, by Qiuxu Zhuo.

 - The usual cleanups and fixes all over the place.

* tag 'edac_updates_for_5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
  EDAC/amd64: Remove redundant assignment to variable ret in hw_info_get()
  EDAC/skx: Use the mcmtr register to retrieve close_pg/bank_xor_enable
  EDAC/i10nm: Update driver to support different bus number config register offsets
  EDAC, {skx,i10nm}: Make some configurations CPU model specific
  EDAC/amd8131: Remove defined but not used bridge_str
  EDAC/thunderx: Make symbols static
  MAINTAINERS: Remove sifive_l2_cache.c from EDAC-SIFIVE pattern
  EDAC/xgene: Remove set but not used address local var
  EDAC/armada_xp: Fix some log messages

Merge tag 'printk-for-5.8' of git://git./linux/kernel/git/printk/linux

Pull printk updates from Petr Mladek:

 - Benjamin Herrenschmidt solved a problem with non-matched console
   aliases by first checking consoles defined on the command line. It is
   a more conservative approach than the previous attempts.

 - Benjamin also made sure that the console accessible via /dev/console
   always has CON_CONSDEV flag.

 - Andy Shevchenko added the %ptT modifier for printing struct time64_t.
   It extends the existing %ptR handling for struct rtc_time.

 - Bruno Meneguele fixed /dev/kmsg error value returned by unsupported
   SEEK_CUR.

 - Tetsuo Handa removed unused pr_cont_once().

... and a few small fixes.

* tag 'printk-for-5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/printk/linux:
  printk: Remove pr_cont_once()
  printk: handle blank console arguments passed in.
  kernel/printk: add kmsg SEEK_CUR handling
  printk: Fix a typo in comment "interator"->"iterator"
  usb: pulse8-cec: Switch to use %ptT
  ARM: bcm2835: Switch to use %ptT
  lib/vsprintf: Print time64_t in human readable format
  lib/vsprintf: update comment about simple_strto<foo>() functions
  printk: Correctly set CON_CONSDEV even when preferred console was not registered
  printk: Fix preferred console selection with multiple matches
  printk: Move console matching logic into a separate function
  printk: Convert a use of sprintf to snprintf in console_unlock

Merge tag 'fsverity-for-linus' of git://git./fs/fscrypt/fscrypt

Pull fsverity updates from Eric Biggers:
 "Fix kerneldoc warnings and some coding style inconsistencies.

  This mirrors the similar cleanups being done in fs/crypto/"

* tag 'fsverity-for-linus' of git://git.kernel.org/pub/scm/fs/fscrypt/fscrypt:
  fs-verity: remove unnecessary extern keywords
  fs-verity: fix all kerneldoc warnings

Merge tag 'fscrypt-for-linus' of git://git./fs/fscrypt/fscrypt

Pull fscrypt updates from Eric Biggers:

 - Add the IV_INO_LBLK_32 encryption policy flag which modifies the
   encryption to be optimized for eMMC inline encryption hardware.

 - Make the test_dummy_encryption mount option for ext4 and f2fs support
   v2 encryption policies.

 - Fix kerneldoc warnings and some coding style inconsistencies.

* tag 'fscrypt-for-linus' of git://git.kernel.org/pub/scm/fs/fscrypt/fscrypt:
  fscrypt: add support for IV_INO_LBLK_32 policies
  fscrypt: make test_dummy_encryption use v2 by default
  fscrypt: support test_dummy_encryption=v2
  fscrypt: add fscrypt_add_test_dummy_key()
  linux/parser.h: add include guards
  fscrypt: remove unnecessary extern keywords
  fscrypt: name all function parameters
  fscrypt: fix all kerneldoc warnings

Merge tag 'pstore-v5.8-rc1' of git://git./linux/kernel/git/kees/linux

Pull pstore updates from Kees Cook:
 "Fixes and new features for pstore.

  This is a pretty big set of changes (relative to past pstore pulls),
  but it has been in -next for a while. The biggest change here is the
  ability to support a block device as a pstore backend, which has been
  desired for a while. A lot of additional fixes and refactorings are
  also included, mostly in support of the new features.

   - refactor pstore locking for safer module unloading (Kees Cook)

   - remove orphaned records from pstorefs when backend unloaded (Kees
     Cook)

   - refactor dump_oops parameter into max_reason (Pavel Tatashin)

   - introduce pstore/zone for common code for contiguous storage
     (WeiXiong Liao)

   - introduce pstore/blk for block device backend (WeiXiong Liao)

   - introduce mtd backend (WeiXiong Liao)"

* tag 'pstore-v5.8-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux: (35 commits)
  mtd: Support kmsg dumper based on pstore/blk
  pstore/blk: Introduce "best_effort" mode
  pstore/blk: Support non-block storage devices
  pstore/blk: Provide way to query pstore configuration
  pstore/zone: Provide way to skip "broken" zone for MTD devices
  Documentation: Add details for pstore/blk
  pstore/zone,blk: Add ftrace frontend support
  pstore/zone,blk: Add console frontend support
  pstore/zone,blk: Add support for pmsg frontend
  pstore/blk: Introduce backend for block devices
  pstore/zone: Introduce common layer to manage storage zones
  ramoops: Add "max-reason" optional field to ramoops DT node
  pstore/ram: Introduce max_reason and convert dump_oops
  pstore/platform: Pass max_reason to kmesg dump
  printk: Introduce kmsg_dump_reason_str()
  printk: honor the max_reason field in kmsg_dumper
  printk: Collapse shutdown types into a single dump reason
  pstore/ftrace: Provide ftrace log merging routine
  pstore/ram: Refactor ftrace buffer merging
  pstore/ram: Refactor DT size parsing
  ...

Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto updates from Herbert Xu:
 "API:
   - Introduce crypto_shash_tfm_digest() and use it wherever possible.
   - Fix use-after-free and race in crypto_spawn_alg.
   - Add support for parallel and batch requests to crypto_engine.

  Algorithms:
   - Update jitter RNG for SP800-90B compliance.
   - Always use jitter RNG as seed in drbg.

  Drivers:
   - Add Arm CryptoCell driver cctrng.
   - Add support for SEV-ES to the PSP driver in ccp"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (114 commits)
  crypto: hisilicon - fix driver compatibility issue with different versions of devices
  crypto: engine - do not requeue in case of fatal error
  crypto: cavium/nitrox - Fix a typo in a comment
  crypto: hisilicon/qm - change debugfs file name from qm_regs to regs
  crypto: hisilicon/qm - add DebugFS for xQC and xQE dump
  crypto: hisilicon/zip - add debugfs for Hisilicon ZIP
  crypto: hisilicon/hpre - add debugfs for Hisilicon HPRE
  crypto: hisilicon/sec2 - add debugfs for Hisilicon SEC
  crypto: hisilicon/qm - add debugfs to the QM state machine
  crypto: hisilicon/qm - add debugfs for QM
  crypto: stm32/crc32 - protect from concurrent accesses
  crypto: stm32/crc32 - don't sleep in runtime pm
  crypto: stm32/crc32 - fix multi-instance
  crypto: stm32/crc32 - fix run-time self test issue.
  crypto: stm32/crc32 - fix ext4 chksum BUG_ON()
  crypto: hisilicon/zip - Use temporary sqe when doing work
  crypto: hisilicon - add device error report through abnormal irq
  crypto: hisilicon - remove codes of directly report device errors through MSI
  crypto: hisilicon - QM memory management optimization
  crypto: hisilicon - unify initial value assignment into QM
  ...

Merge tag 'i3c/for-5.8' of git://git./linux/kernel/git/i3c/linux

Pull i3c update from Boris Brezillon:
 "Fix GETMRL's logic"

* tag 'i3c/for-5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/i3c/linux:
  i3c master: GETMRL's 3rd byte is optional even with BCR_IBI_PAYLOAD

Merge tag 'regulator-v5.8' of git://git./linux/kernel/git/broonie/regulator

Pull regulator updates from Mark Brown:
 "The big change in this release is that Matti Vaittinen has factored
  out the linear ranges support into a separate library in lib/ since it
  is also useful for at least the power subsystem (and most likely
  others too), it helps subsystems which need to map register values
  into more useful real world values do so with minimal per-driver code.

   - Factoring out of the linear ranges support into a library in lib/
     from Matti Vaittinen.

   - Trace points for bypass mode.

   - Use the consumer name in debugfs to make it easier to understand.

   - New drivers for Maxim MAX77826 and MAX8998"

* tag 'regulator-v5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator: (23 commits)
  regulator: max8998: max8998_set_current_limit() can be static
  dt-bindings: regulator: Convert anatop regulator to json-schema
  regulator: core: Add regulator bypass trace points
  regulator: extract voltage balancing code to the separate function
  regulator/mfd: max8998: Document charger regulator
  regulator: max8998: Add charger regulator
  MAINTAINERS: Add maintainer entry for linear ranges helper
  regulator: bd718x7: remove voltage change restriction from BD71847 LDOs
  lib: linear_ranges: Add missing MODULE_LICENSE()
  regulator: use linear_ranges helper
  power: supply: bd70528: rename linear_range to avoid collision
  lib/test_linear_ranges: add a test for the 'linear_ranges'
  lib: add linear ranges helpers
  regulator: db8500-prcmu: Use true,false for bool variable
  regulator: bd718x7: remove voltage change restriction from BD71847
  regulator: max77826: Remove erroneous additionalProperties
  regulator: qcom-rpmh: Fix typos in pm8150 and pm8150l
  regulator: Document bindings for max77826
  regulator: max77826: Add max77826 regulator driver
  regulator: tps80031: remove redundant assignment to variables ret and val
  ...

Merge tag 'spi-v5.8' of git://git./linux/kernel/git/broonie/spi

Pull spi updates from Mark Brown:
 "This has been a very active release for the DesignWare driver in
  particular - after a long period of inactivity we have had a lot of
  people actively working on it for unrelated reasons this cycle with
  some of that work still not landed.

  Otherwise it's been fairly quiet for the subsystem.

  Highlights include:

   - Lots of performance improvements and fixes for the DesignWare
     driver from Serge Semin, Andy Shevchenko, Wan Ahmad Zainie, Clement
     Leger, Dinh Nguyen and Jarkko Nikula.

   - Support for octal mode transfers in spidev.

   - Slave mode support for the Rockchip drivers.

   - Support for AMD controllers, Broadcom mspi and Raspberry Pi 4, and
     Intel Elkhart Lake"

* tag 'spi-v5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi: (125 commits)
  spi: spi-fsl-dspi: fix native data copy
  spi: Convert DW SPI binding to DT schema
  spi: dw: Refactor mid_spi_dma_setup() to separate DMA and IRQ config
  spi: dw: Make DMA request line assignments explicit for Intel Medfield
  spi: bcm2835: Remove shared interrupt support
  dt-bindings: snps,dw-apb-ssi: add optional reset property
  spi: dw: add reset control
  spi: bcm2835: Enable shared interrupt support
  spi: bcm2835: Implement shutdown callback
  spi: dw: Use regset32 DebugFS method to create regdump file
  spi: dw: Add DMA support to the DW SPI MMIO driver
  spi: dw: Cleanup generic DW DMA code namings
  spi: dw: Add DW SPI DMA/PCI/MMIO dependency on the DW SPI core
  spi: dw: Remove DW DMA code dependency from DW_DMAC_PCI
  spi: dw: Move Non-DMA code to the DW PCIe-SPI driver
  spi: dw: Add core suffix to the DW APB SSI core source file
  spi: dw: Fix Rx-only DMA transfers
  spi: dw: Use DMA max burst to set the request thresholds
  spi: dw: Parameterize the DMA Rx/Tx burst length
  spi: dw: Add SPI Rx-done wait method to DMA-based transfer
  ...

Merge tag 'regmap-v5.8' of git://git./linux/kernel/git/broonie/regmap

Pull regmap updates from Mark Brown:
 "This has been a very active release for the regmap API for some
  reason, a lot of it due to new devices with odd requirements that can
  sensibly be handled here.

   - Add support for buses implementing a custom reg_update_bits()
     method in case the bus has a native operation for this.

   - Support 16 bit register addresses in SMBus.

   - Allow customization of the device attached to regmap-irq.

   - Helpers for bitfield operations and per-port field initializations"

* tag 'regmap-v5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
  regmap: provide helpers for simple bit operations
  regmap: add helper for per-port regfield initialization
  regmap-i2c: add 16-bit width registers support
  regmap: Simplify implementation of the regmap_field_read_poll_timeout() macro
  regmap: Simplify implementation of the regmap_read_poll_timeout() macro
  regmap: add reg_sequence helpers
  regmap-irq: make it possible to add irq_chip do a specific device node
  regmap: Add bus reg_update_bits() support
  regmap: debugfs: check count when read regmap file

Merge tag 'hwmon-for-v5.8' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon updates from Guenter Roeck:
 "Infrastructure:
   - Add notification support

  New drivers:
   - Baikal-T1 PVT sensor driver
   - amd_energy driver to report energy counters
   - Driver for Maxim MAX16601
   - Gateworks System Controller

  Various:
   - applesmc: avoid overlong udelay()
   - dell-smm: Use one DMI match for all XPS models
   - ina2xx: Implement alert functions
   - lm70: Add support for ACPI
   - lm75: Fix coding-style warnings
   - lm90: Add max6654 support to lm90 driver
   - nct7802: Replace container_of() API
   - nct7904: Set default timeout
   - nct7904: Add watchdog function
   - pmbus: Improve initialization of 'currpage' and 'currphase'"

* tag 'hwmon-for-v5.8' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging: (24 commits)
  hwmon: Add Baikal-T1 PVT sensor driver
  hwmon: Add notification support
  dt-bindings: hwmon: Add Baikal-T1 PVT sensor binding
  hwmon: (applesmc) avoid overlong udelay()
  hwmon: (nct7904) Set default timeout
  hwmon: (amd_energy) Missing platform_driver_unregister() on error in amd_energy_init()
  MAINTAINERS: add entry for AMD energy driver
  hwmon: (amd_energy) Add documentation
  hwmon: Add amd_energy driver to report energy counters
  hwmon: (nct7802) Replace container_of() API
  hwmon: (lm90) Add max6654 support to lm90 driver
  hwmon : (nct6775) Use kobj_to_dev() API
  hwmon: (pmbus) Driver for Maxim MAX16601
  hwmon: (pmbus) Improve initialization of 'currpage' and 'currphase'
  hwmon: (adt7411) update contact email
  hwmon: (lm75) Fix all coding-style warnings on lm75 driver
  hwmon: Reduce indentation level in __hwmon_device_register()
  hwmon: (ina2xx) Implement alert functions
  hwmon: (lm70) Add support for ACPI
  hwmon: (dell-smm) Use one DMI match for all XPS models
  ...

Merge tag 'tpmdd-next-20200522' of git://git.infradead.org/users/jjs/linux-tpmdd

Pull tpm updates from Jarkko Sakkinen.

* tag 'tpmdd-next-20200522' of git://git.infradead.org/users/jjs/linux-tpmdd:
  tpm: eventlog: Replace zero-length array with flexible-array member
  tpm/tpm_ftpm_tee: Use UUID API for exporting the UUID