mm: kfence: pass a pointer to virt_to_page()

Functions that work on a pointer to virtual memory such as virt_to_pfn()
and users of that function such as virt_to_page() are supposed to pass a
pointer to virtual memory, ideally a (void *) or other pointer.  However
since many architectures implement virt_to_pfn() as a macro, this function
becomes polymorphic and accepts both a (unsigned long) and a (void *).

If we instead implement a proper virt_to_pfn(void *addr) function the
following happens (occurred on arch/arm):

mm/kfence/core.c:558:30: warning: passing argument 1
  of 'virt_to_pfn' makes pointer from integer without a
  cast [-Wint-conversion]

In one case we can refer to __kfence_pool directly (and that is a proper
(char *) pointer) and in the other call site we use an explicit cast.

Link: https://lkml.kernel.org/r/20220630084124.691207-4-linus.walleij@linaro.org
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Marco Elver <elver@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/highmem: pass a pointer to virt_to_page()

Functions that work on a pointer to virtual memory such as virt_to_pfn()
and users of that function such as virt_to_page() are supposed to pass a
pointer to virtual memory, ideally a (void *) or other pointer.  However
since many architectures implement virt_to_pfn() as a macro, this function
becomes polymorphic and accepts both a (unsigned long) and a (void *).

If we instead implement a proper virt_to_pfn(void *addr) function the
following happens (occurred on arch/arm):

mm/highmem.c:153:29: warning: passing argument 1 of
  'virt_to_pfn' makes pointer from integer without a
  cast [-Wint-conversion]

We already have a proper void * pointer in the scope of this function
named "vaddr" so pass that instead.

Link: https://lkml.kernel.org/r/20220630084124.691207-3-linus.walleij@linaro.org
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

lib/test_free_pages.c: pass a pointer to virt_to_page()

In a recent change to the Arm architecture with the end goal of removing
highmem we need to convert virt_to_phys() and virt_to_pfn() to static
inline functions.

This will make them strongly typed.

However since virt_to_* is always implemented as macros they have become
polymorphic and accept both (void *) and e.g.  unsigned long as arguments.

Other functions such as virt_to_page() simply wrap virt_to_pfn() and get
affected indirectly.

To be able to proceed, patch mm to use (void *) as argument to affected
functions in all instances.

This patch (of 5):

A pointer into virtual memory is represented by a (void *) not an u32, so
the compiler warns:

lib/test_free_pages.c:20:50: warning: passing argument 1
  of 'virt_to_pfn' makes pointer from integer without a
  cast [-Wint-conversion]

Fix this with an explicit cast.

Link: https://lkml.kernel.org/r/20220630084124.691207-1-linus.walleij@linaro.org
Link: https://lkml.kernel.org/r/20220630084124.691207-2-linus.walleij@linaro.org
Signed-off-by: Linus Walleij <linus.walleij@linaro.org>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Marco Elver <elver@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/memcontrol.c: replace cgroup_memory_nokmem with mem_cgroup_kmem_disabled()

mem_cgroup_kmem_disabled() checks whether the kmem accounting is off.
Therefore, replace cgroup_memory_nokmem with mem_cgroup_kmem_disabled(),
which is the same work in percpu.c and slab_common.c.

Link: https://lkml.kernel.org/r/20220625061844.226764-1-xiangyang3@huawei.com
Signed-off-by: Xiang Yang <xiangyang3@huawei.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Roman Gushchin <roman.gushchin@linux.dev>
Acked-by: Souptick Joarder (HPE) <jrdr.linux@gmail.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: replace local_lock with normal spinlock

struct per_cpu_pages is no longer strictly local as PCP lists can be
drained remotely using a lock for protection.  While the use of local_lock
works, it goes against the intent of local_lock which is for "pure CPU
local concurrency control mechanisms and not suited for inter-CPU
concurrency control" (Documentation/locking/locktypes.rst)

local_lock protects against migration between when the percpu pointer is
accessed and the pcp->lock acquired.  The lock acquisition is a preemption
point so in the worst case, a task could migrate to another NUMA node and
accidentally allocate remote memory.  The main requirement is to pin the
task to a CPU that is suitable for PREEMPT_RT and !PREEMPT_RT.

Replace local_lock with helpers that pin a task to a CPU, lookup the
per-cpu structure and acquire the embedded lock.  It's similar to
local_lock without breaking the intent behind the API.  It is not a
complete API as only the parts needed for PCP-alloc are implemented but in
theory, the generic helpers could be promoted to a general API if there
was demand for an embedded lock within a per-cpu struct with a guarantee
that the per-cpu structure locked matches the running CPU and cannot use
get_cpu_var due to RT concerns.  PCP requires these semantics to avoid
accidentally allocating remote memory.

[mgorman@techsingularity.net: use pcp_spin_trylock_irqsave instead of pcpu_spin_trylock_irqsave]
Link: https://lkml.kernel.org/r/20220627084645.GA27531@techsingularity.net
Link: https://lkml.kernel.org/r/20220624125423.6126-8-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Tested-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Yu Zhao <yuzhao@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: remotely drain per-cpu lists

Some setups, notably NOHZ_FULL CPUs, are too busy to handle the per-cpu
drain work queued by __drain_all_pages().  So introduce a new mechanism to
remotely drain the per-cpu lists.  It is made possible by remotely locking
'struct per_cpu_pages' new per-cpu spinlocks.  A benefit of this new
scheme is that drain operations are now migration safe.

There was no observed performance degradation vs.  the previous scheme.
Both netperf and hackbench were run in parallel to triggering the
__drain_all_pages(NULL, true) code path around ~100 times per second.  The
new scheme performs a bit better (~5%), although the important point here
is there are no performance regressions vs.  the previous mechanism.
Per-cpu lists draining happens only in slow paths.

Minchan Kim tested an earlier version and reported;

My workload is not NOHZ CPUs but run apps under heavy memory
pressure so they goes to direct reclaim and be stuck on
drain_all_pages until work on workqueue run.

unit: nanosecond
max(dur)        avg(dur)                count(dur)
166713013       487511.77786438033      1283

From traces, system encountered the drain_all_pages 1283 times and
worst case was 166ms and avg was 487us.

The other problem was alloc_contig_range in CMA. The PCP draining
takes several hundred millisecond sometimes though there is no
memory pressure or a few of pages to be migrated out but CPU were
fully booked.

Your patch perfectly removed those wasted time.

Link: https://lkml.kernel.org/r/20220624125423.6126-7-mgorman@techsingularity.net
Signed-off-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Yu Zhao <yuzhao@google.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: protect PCP lists with a spinlock

Currently the PCP lists are protected by using local_lock_irqsave to
prevent migration and IRQ reentrancy but this is inconvenient.  Remote
draining of the lists is impossible and a workqueue is required and every
task allocation/free must disable then enable interrupts which is
expensive.

As preparation for dealing with both of those problems, protect the
lists with a spinlock.  The IRQ-unsafe version of the lock is used
because IRQs are already disabled by local_lock_irqsave.  spin_trylock
is used in combination with local_lock_irqsave() but later will be
replaced with a spin_trylock_irqsave when the local_lock is removed.

The per_cpu_pages still fits within the same number of cache lines after
this patch relative to before the series.

struct per_cpu_pages {
        spinlock_t                 lock;                 /*     0     4 */
        int                        count;                /*     4     4 */
        int                        high;                 /*     8     4 */
        int                        batch;                /*    12     4 */
        short int                  free_factor;          /*    16     2 */
        short int                  expire;               /*    18     2 */

        /* XXX 4 bytes hole, try to pack */

        struct list_head           lists[13];            /*    24   208 */

        /* size: 256, cachelines: 4, members: 7 */
        /* sum members: 228, holes: 1, sum holes: 4 */
        /* padding: 24 */
} __attribute__((__aligned__(64)));

There is overhead in the fast path due to acquiring the spinlock even
though the spinlock is per-cpu and uncontended in the common case.  Page
Fault Test (PFT) running on a 1-socket reported the following results on a
1 socket machine.

                                     5.19.0-rc3               5.19.0-rc3
                                        vanilla      mm-pcpspinirq-v5r16
Hmean     faults/sec-1   869275.7381 (   0.00%)   874597.5167 *   0.61%*
Hmean     faults/sec-3  2370266.6681 (   0.00%)  2379802.0362 *   0.40%*
Hmean     faults/sec-5  2701099.7019 (   0.00%)  2664889.7003 *  -1.34%*
Hmean     faults/sec-7  3517170.9157 (   0.00%)  3491122.8242 *  -0.74%*
Hmean     faults/sec-8  3965729.6187 (   0.00%)  3939727.0243 *  -0.66%*

There is a small hit in the number of faults per second but given that the
results are more stable, it's borderline noise.

[akpm@linux-foundation.org: add missing local_unlock_irqrestore() on contention path]
Link: https://lkml.kernel.org/r/20220624125423.6126-6-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Yu Zhao <yuzhao@google.com>
Reviewed-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Tested-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: remove mistaken page == NULL check in rmqueue

If a page allocation fails, the ZONE_BOOSTER_WATERMARK should be tested,
cleared and kswapd woken whether the allocation attempt was via the PCP or
directly via the buddy list.

Remove the page == NULL so the ZONE_BOOSTED_WATERMARK bit is checked
unconditionally.  As it is unlikely that ZONE_BOOSTED_WATERMARK is set,
mark the branch accordingly.

Link: https://lkml.kernel.org/r/20220624125423.6126-5-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Yu Zhao <yuzhao@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: split out buddy removal code from rmqueue into separate helper

This is a preparation page to allow the buddy removal code to be reused in
a later patch.

No functional change.

Link: https://lkml.kernel.org/r/20220624125423.6126-4-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Minchan Kim <minchan@kernel.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Yu Zhao <yuzhao@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: use only one PCP list for THP-sized allocations

The per_cpu_pages is cache-aligned on a standard x86-64 distribution
configuration but a later patch will add a new field which would push the
structure into the next cache line.  Use only one list to store THP-sized
pages on the per-cpu list.  This assumes that the vast majority of
THP-sized allocations are GFP_MOVABLE but even if it was another type, it
would not contribute to serious fragmentation that potentially causes a
later THP allocation failure.  Align per_cpu_pages on the cacheline
boundary to ensure there is no false cache sharing.

After this patch, the structure sizing is;

struct per_cpu_pages {
        int                        count;                /*     0     4 */
        int                        high;                 /*     4     4 */
        int                        batch;                /*     8     4 */
        short int                  free_factor;          /*    12     2 */
        short int                  expire;               /*    14     2 */
        struct list_head           lists[13];            /*    16   208 */

        /* size: 256, cachelines: 4, members: 6 */
        /* padding: 32 */
} __attribute__((__aligned__(64)));

Link: https://lkml.kernel.org/r/20220624125423.6126-3-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Minchan Kim <minchan@kernel.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Yu Zhao <yuzhao@google.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: add page->buddy_list and page->pcp_list

Patch series "Drain remote per-cpu directly", v5.

Some setups, notably NOHZ_FULL CPUs, may be running realtime or
latency-sensitive applications that cannot tolerate interference due to
per-cpu drain work queued by __drain_all_pages().  Introduce a new
mechanism to remotely drain the per-cpu lists.  It is made possible by
remotely locking 'struct per_cpu_pages' new per-cpu spinlocks.  This has
two advantages, the time to drain is more predictable and other unrelated
tasks are not interrupted.

This series has the same intent as Nicolas' series "mm/page_alloc: Remote
per-cpu lists drain support" -- avoid interference of a high priority task
due to a workqueue item draining per-cpu page lists.  While many workloads
can tolerate a brief interruption, it may cause a real-time task running
on a NOHZ_FULL CPU to miss a deadline and at minimum, the draining is
non-deterministic.

Currently an IRQ-safe local_lock protects the page allocator per-cpu
lists.  The local_lock on its own prevents migration and the IRQ disabling
protects from corruption due to an interrupt arriving while a page
allocation is in progress.

This series adjusts the locking.  A spinlock is added to struct
per_cpu_pages to protect the list contents while local_lock_irq is
ultimately replaced by just the spinlock in the final patch.  This allows
a remote CPU to safely.  Follow-on work should allow the spin_lock_irqsave
to be converted to spin_lock to avoid IRQs being disabled/enabled in most
cases.  The follow-on patch will be one kernel release later as it is
relatively high risk and it'll make bisections more clear if there are any
problems.

Patch 1 is a cosmetic patch to clarify when page->lru is storing buddy pages
and when it is storing per-cpu pages.

Patch 2 shrinks per_cpu_pages to make room for a spin lock. Strictly speaking
this is not necessary but it avoids per_cpu_pages consuming another
cache line.

Patch 3 is a preparation patch to avoid code duplication.

Patch 4 is a minor correction.

Patch 5 uses a spin_lock to protect the per_cpu_pages contents while still
relying on local_lock to prevent migration, stabilise the pcp
lookup and prevent IRQ reentrancy.

Patch 6 remote drains per-cpu pages directly instead of using a workqueue.

Patch 7 uses a normal spinlock instead of local_lock for remote draining

This patch (of 7):

The page allocator uses page->lru for storing pages on either buddy or PCP
lists.  Create page->buddy_list and page->pcp_list as a union with
page->lru.  This is simply to clarify what type of list a page is on in
the page allocator.

No functional change intended.

[minchan@kernel.org: fix page lru fields in macros]
Link: https://lkml.kernel.org/r/20220624125423.6126-2-mgorman@techsingularity.net
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Tested-by: Minchan Kim <minchan@kernel.org>
Acked-by: Minchan Kim <minchan@kernel.org>
Reviewed-by: Nicolas Saenz Julienne <nsaenzju@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Tested-by: Yu Zhao <yuzhao@google.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb: lazy page table copies in fork()

Lazy page table copying at fork time was introduced with commit
d992895ba2b2 ("[PATCH] Lazy page table copies in fork()").  At the time,
hugetlb was very new and did not support page faulting.  As a result, it
was excluded.  When full page fault support was added for hugetlb, the
exclusion was not removed.

Simply remove the check that prevents lazy copying of hugetlb page tables
at fork.  Of course, like other mappings this only applies to shared
mappings.

Lazy page table copying at fork will be less advantageous for hugetlb
mappings because:
- There are fewer page table entries with hugetlb
- hugetlb pmds can be shared instead of copied

In any case, completely eliminating the copy at fork time should speed
things up.

Link: https://lkml.kernel.org/r/20220621235620.291305-5-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@linux.ibm.com>
Cc: James Houghton <jthoughton@google.com>
Cc: kernel test robot <lkp@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Naoya Horiguchi <naoya.horiguchi@linux.dev>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rolf Eike Beer <eike-kernel@sf-tec.de>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb: do not update address in huge_pmd_unshare

As an optimization for loops sequentially processing hugetlb address
ranges, huge_pmd_unshare would update a passed address if it unshared a
pmd.  Updating a loop control variable outside the loop like this is
generally a bad idea.  These loops are now using hugetlb_mask_last_page to
optimize scanning when non-present ptes are discovered.  The same can be
done when huge_pmd_unshare returns 1 indicating a pmd was unshared.

Remove address update from huge_pmd_unshare.  Change the passed argument
type and update all callers.  In loops sequentially processing addresses
use hugetlb_mask_last_page to update address if pmd is unshared.

[sfr@canb.auug.org.au: fix an unused variable warning/error]
Link: https://lkml.kernel.org/r/20220622171117.70850960@canb.auug.org.au
Link: https://lkml.kernel.org/r/20220621235620.291305-4-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Acked-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@linux.ibm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: kernel test robot <lkp@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Naoya Horiguchi <naoya.horiguchi@linux.dev>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rolf Eike Beer <eike-kernel@sf-tec.de>
Cc: Will Deacon <will@kernel.org>
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

arm64/hugetlb: implement arm64 specific hugetlb_mask_last_page

The HugeTLB address ranges are linearly scanned during fork, unmap and
remap operations, and the linear scan can skip to the end of range mapped
by the page table page if hitting a non-present entry, which can help to
speed linear scanning of the HugeTLB address ranges.

So hugetlb_mask_last_page() is introduced to help to update the address in
the loop of HugeTLB linear scanning with getting the last huge page mapped
by the associated page table page[1], when a non-present entry is
encountered.

Considering ARM64 specific cont-pte/pmd size HugeTLB, this patch
implemented an ARM64 specific hugetlb_mask_last_page() to help this case.

[1] https://lore.kernel.org/linux-mm/20220527225849.284839-1-mike.kravetz@oracle.com/

[baolin.wang@linux.alibaba.com: fix build]
Link: https://lkml.kernel.org/r/a14e7b39-6a8a-4609-b4a1-84ac574f5c96@linux.alibaba.com
Link: https://lkml.kernel.org/r/20220621235620.291305-3-mike.kravetz@oracle.com
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Acked-by: Muchun Song <songmuchun@bytedance.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christian Borntraeger <borntraeger@linux.ibm.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: James Houghton <jthoughton@google.com>
Cc: kernel test robot <lkp@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Naoya Horiguchi <naoya.horiguchi@linux.dev>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rolf Eike Beer <eike-kernel@sf-tec.de>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb: skip to end of PT page mapping when pte not present

Patch series "hugetlb: speed up linear address scanning", v2.

At unmap, fork and remap time hugetlb address ranges are linearly scanned.
We can optimize these scans if the ranges are sparsely populated.

Also, enable page table "Lazy copy" for hugetlb at fork.

NOTE: Architectures not defining CONFIG_ARCH_WANT_GENERAL_HUGETLB need to
add an arch specific version hugetlb_mask_last_page() to take advantage of
sparse address scanning improvements.  Baolin Wang added the routine for
arm64.  Other architectures which could be optimized are: ia64, mips,
parisc, powerpc, s390, sh and sparc.

This patch (of 4):

HugeTLB address ranges are linearly scanned during fork, unmap and remap
operations.  If a non-present entry is encountered, the code currently
continues to the next huge page aligned address.  However, a non-present
entry implies that the page table page for that entry is not present.
Therefore, the linear scan can skip to the end of range mapped by the page
table page.  This can speed operations on large sparsely populated hugetlb
mappings.

Create a new routine hugetlb_mask_last_page() that will return an address
mask.  When the mask is ORed with an address, the result will be the
address of the last huge page mapped by the associated page table page.
Use this mask to update addresses in routines which linearly scan hugetlb
address ranges when a non-present pte is encountered.

hugetlb_mask_last_page is related to the implementation of huge_pte_offset
as hugetlb_mask_last_page is called when huge_pte_offset returns NULL.
This patch only provides a complete hugetlb_mask_last_page implementation
when CONFIG_ARCH_WANT_GENERAL_HUGETLB is defined.  Architectures which
provide their own versions of huge_pte_offset can also provide their own
version of hugetlb_mask_last_page.

Link: https://lkml.kernel.org/r/20220621235620.291305-1-mike.kravetz@oracle.com
Link: https://lkml.kernel.org/r/20220621235620.291305-2-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Tested-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Acked-by: Muchun Song <songmuchun@bytedance.com>
Reported-by: kernel test robot <lkp@intel.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Naoya Horiguchi <naoya.horiguchi@linux.dev>
Cc: James Houghton <jthoughton@google.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Paul Walmsley <paul.walmsley@sifive.com>
Cc: Christian Borntraeger <borntraeger@linux.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Rolf Eike Beer <eike-kernel@sf-tec.de>
Cc: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

kasan: separate double free case from invalid free

Currently, KASAN describes all invalid-free/double-free bugs as
"double-free or invalid-free".  This is ambiguous.

KASAN should report "double-free" when a double-free is a more likely
cause (the address points to the start of an object) and report
"invalid-free" otherwise [1].

[1] https://bugzilla.kernel.org/show_bug.cgi?id=212193

Link: https://lkml.kernel.org/r/20220615062219.22618-1-Kuan-Ying.Lee@mediatek.com
Signed-off-by: Kuan-Ying Lee <Kuan-Ying.Lee@mediatek.com>
Reviewed-by: Dmitry Vyukov <dvyukov@google.com>
Reviewed-by: Andrey Konovalov <andreyknvl@gmail.com>
Cc: Andrey Ryabinin <ryabinin.a.a@gmail.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Cc: Matthias Brugger <matthias.bgg@gmail.com>
Cc: Chinwen Chang <chinwen.chang@mediatek.com>
Cc: Yee Lee <yee.lee@mediatek.com>
Cc: Andrew Yang <andrew.yang@mediatek.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

doc: proc: fix the description to THPeligible

The THPeligible bit shows 1 if and only if the VMA is eligible for
allocating THP and the THP is also PMD mappable.  Some misaligned file
VMAs may be eligible for allocating THP but the THP can't be mapped by
PMD.  Make this more explicitly to avoid ambiguity.

Link: https://lkml.kernel.org/r/20220616174840.1202070-8-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: khugepaged: reorg some khugepaged helpers

The khugepaged_{enabled|always|req_madv} are not khugepaged only anymore,
move them to huge_mm.h and rename to hugepage_flags_xxx, and remove
khugepaged_req_madv due to no users.

Also move khugepaged_defrag to khugepaged.c since its only caller is in
that file, it doesn't have to be in a header file.

Link: https://lkml.kernel.org/r/20220616174840.1202070-7-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: thp: kill __transhuge_page_enabled()

The page fault path checks THP eligibility with __transhuge_page_enabled()
which does the similar thing as hugepage_vma_check(), so use
hugepage_vma_check() instead.

However page fault allows DAX and !anon_vma cases, so added a new flag,
in_pf, to hugepage_vma_check() to make page fault work correctly.

The in_pf flag is also used to skip shmem and file THP for page fault
since shmem handles THP in its own shmem_fault() and file THP allocation
on fault is not supported yet.

Also remove hugepage_vma_enabled() since hugepage_vma_check() is the only
caller now, it is not necessary to have a helper function.

Link: https://lkml.kernel.org/r/20220616174840.1202070-6-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: thp: kill transparent_hugepage_active()

The transparent_hugepage_active() was introduced to show THP eligibility
bit in smaps in proc, smaps is the only user.  But it actually does the
similar check as hugepage_vma_check() which is used by khugepaged.  We
definitely don't have to maintain two similar checks, so kill
transparent_hugepage_active().

This patch also fixed the wrong behavior for VM_NO_KHUGEPAGED vmas.

Also move hugepage_vma_check() to huge_memory.c and huge_mm.h since it
is not only for khugepaged anymore.

[akpm@linux-foundation.org: check vma->vm_mm, per Zach]
[akpm@linux-foundation.org: add comment to vdso check]
Link: https://lkml.kernel.org/r/20220616174840.1202070-5-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: khugepaged: better comments for anon vma check in hugepage_vma_revalidate

The hugepage_vma_revalidate() needs to check if the vma is still anonymous
vma or not since the address may be unmapped then remapped to file before
khugepaged reaquired the mmap_lock.

The old comment is not quite helpful, elaborate this with better comment.

Link: https://lkml.kernel.org/r/20220616174840.1202070-4-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: thp: consolidate vma size check to transhuge_vma_suitable

There are couple of places that check whether the vma size is ok for THP
or whether address fits, they are open coded and duplicate, use
transhuge_vma_suitable() to do the job by passing in (vma->end -
HPAGE_PMD_SIZE).

Move vma size check into hugepage_vma_check().  This will make
khugepaged_enter() is as same as khugepaged_enter_vma().  There is just
one caller for khugepaged_enter(), replace it to khugepaged_enter_vma()
and remove khugepaged_enter().

Link: https://lkml.kernel.org/r/20220616174840.1202070-3-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: khugepaged: check THP flag in hugepage_vma_check()

Patch series "Cleanup transhuge_xxx helpers", v5.

This series is the follow-up of the discussion about cleaning up
transhuge_xxx helpers at
https://lore.kernel.org/linux-mm/627a71f8-e879-69a5-ceb3-fc8d29d2f7f1@suse.cz/.

THP has a bunch of helpers that do VMA sanity check for different paths,
they do the similar checks for the most callsites and have a lot duplicate
codes.  And it is confusing what helpers should be used at what
conditions.

This series reorganized and cleaned up the code so that we could
consolidate all the checks into hugepage_vma_check().

The transhuge_vma_enabled(), transparent_hugepage_active() and
__transparent_hugepage_enabled() are killed by this series.

This patch (of 7):

Currently the THP flag check in hugepage_vma_check() will fallthrough if
the flag is NEVER and VM_HUGEPAGE is set.  This is not a problem for now
since all the callers have the flag checked before or can't be invoked if
the flag is NEVER.

However, the following patch will call hugepage_vma_check() in more
places, for example, page fault, so this flag must be checked in
hugepge_vma_check().

Link: https://lkml.kernel.org/r/20220616174840.1202070-1-shy828301@gmail.com
Link: https://lkml.kernel.org/r/20220616174840.1202070-2-shy828301@gmail.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

xfs: add dax dedupe support

Introduce xfs_mmaplock_two_inodes_and_break_dax_layout() for dax files who
are going to be deduped.  After that, call compare range function only
when files are both DAX or not.

Link: https://lkml.kernel.org/r/20220603053738.1218681-15-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

xfs: support CoW in fsdax mode

In fsdax mode, WRITE and ZERO on a shared extent need CoW performed.
After that, new allocated extents needs to be remapped to the file.  So,
add a CoW identification in ->iomap_begin(), and implement ->iomap_end()
to do the remapping work.

[akpm@linux-foundation.org: make xfs_dax_fault() static]
Link: https://lkml.kernel.org/r/20220603053738.1218681-14-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fsdax: dedup file range to use a compare function

With dax we cannot deal with readpage() etc.  So, we create a dax
comparison function which is similar with vfs_dedupe_file_range_compare().
And introduce dax_remap_file_range_prep() for filesystem use.

Link: https://lkml.kernel.org/r/20220603053738.1218681-13-ruansy.fnst@fujitsu.com
Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fsdax: add dax_iomap_cow_copy() for dax zero

Punch hole on a reflinked file needs dax_iomap_cow_copy() too.  Otherwise,
data in not aligned area will be not correct.  So, add the CoW operation
for not aligned case in dax_memzero().

Link: https://lkml.kernel.org/r/20220603053738.1218681-12-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Ritesh Harjani <riteshh@linux.ibm.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fsdax: replace mmap entry in case of CoW

Replace the existing entry to the newly allocated one in case of CoW.
Also, we mark the entry as PAGECACHE_TAG_TOWRITE so writeback marks this
entry as writeprotected.  This helps us snapshots so new write pagefaults
after snapshots trigger a CoW.

Link: https://lkml.kernel.org/r/20220603053738.1218681-11-ruansy.fnst@fujitsu.com
Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ritesh Harjani <riteshh@linux.ibm.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fsdax: introduce dax_iomap_cow_copy()

In the case where the iomap is a write operation and iomap is not equal to
srcmap after iomap_begin, we consider it is a CoW operation.

In this case, the destination (iomap->addr) points to a newly allocated
extent.  It is needed to copy the data from srcmap to the extent.  In
theory, it is better to copy the head and tail ranges which is outside of
the non-aligned area instead of copying the whole aligned range.  But in
dax page fault, it will always be an aligned range.  So copy the whole
range in this case.

Link: https://lkml.kernel.org/r/20220603053738.1218681-10-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fsdax: output address in dax_iomap_pfn() and rename it

Add address output in dax_iomap_pfn() in order to perform a memcpy() in
CoW case.  Since this function both output address and pfn, rename it to
dax_iomap_direct_access().

[ruansy.fnst@fujitsu.com: initialize `rc', per Dan]
Link: https://lore.kernel.org/linux-fsdevel/Yp8FUZnO64Qvyx5G@kili/
Link: https://lkml.kernel.org/r/20220607143837.161174-1-ruansy.fnst@fujitsu.com
Link: https://lkml.kernel.org/r/20220603053738.1218681-9-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Ritesh Harjani <riteshh@linux.ibm.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fsdax: set a CoW flag when associate reflink mappings

Introduce a PAGE_MAPPING_DAX_COW flag to support association with CoW file
mappings.  In this case, since the dax-rmap has already took the
responsibility to look up for shared files by given dax page, the
page->mapping is no longer to used for rmap but for marking that this dax
page is shared.  And to make sure disassociation works fine, we use
page->index as refcount, and clear page->mapping to the initial state when
page->index is decreased to 0.

With the help of this new flag, it is able to distinguish normal case and
CoW case, and keep the warning in normal case.

Link: https://lkml.kernel.org/r/20220603053738.1218681-8-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

xfs: implement ->notify_failure() for XFS

Introduce xfs_notify_failure.c to handle failure related works, such as
implement ->notify_failure(), register/unregister dax holder in xfs, and
so on.

If the rmap feature of XFS enabled, we can query it to find files and
metadata which are associated with the corrupt data.  For now all we do is
kill processes with that file mapped into their address spaces, but future
patches could actually do something about corrupt metadata.

After that, the memory failure needs to notify the processes who are using
those files.

Link: https://lkml.kernel.org/r/20220603053738.1218681-7-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: introduce mf_dax_kill_procs() for fsdax case

This new function is a variant of mf_generic_kill_procs that accepts a
file, offset pair instead of a struct to support multiple files sharing a
DAX mapping.  It is intended to be called by the file systems as part of
the memory_failure handler after the file system performed a reverse
mapping from the storage address to the file and file offset.

Link: https://lkml.kernel.org/r/20220603053738.1218681-6-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

fsdax: introduce dax_lock_mapping_entry()

The current dax_lock_page() locks dax entry by obtaining mapping and index
in page.  To support 1-to-N RMAP in NVDIMM, we need a new function to lock
a specific dax entry corresponding to this file's mapping,index.  And
output the page corresponding to the specific dax entry for caller use.

Link: https://lkml.kernel.org/r/20220603053738.1218681-5-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

pagemap,pmem: introduce ->memory_failure()

When memory-failure occurs, we call this function which is implemented by
each kind of devices.  For the fsdax case, pmem device driver implements
it.  Pmem device driver will find out the filesystem in which the
corrupted page located in.

With dax_holder notify support, we are able to notify the memory failure
from pmem driver to upper layers.  If there is something not support in
the notify routine, memory_failure will fall back to the generic hanlder.

Link: https://lkml.kernel.org/r/20220603053738.1218681-4-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: factor helpers for memory_failure_dev_pagemap

memory_failure_dev_pagemap code is a bit complex before introduce RMAP
feature for fsdax.  So it is needed to factor some helper functions to
simplify these code.

[akpm@linux-foundation.org: fix CONFIG_HUGETLB_PAGE=n build]
[zhengbin13@huawei.com: fix redefinition of mf_generic_kill_procs]
Link: https://lkml.kernel.org/r/20220628112143.1170473-1-zhengbin13@huawei.com
Link: https://lkml.kernel.org/r/20220603053738.1218681-3-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Signed-off-by: Zheng Bin <zhengbin13@huawei.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dan Williams <dan.j.williams@intel.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.wiliams@intel.com>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

dax: introduce holder for dax_device

Patch series "v14 fsdax-rmap + v11 fsdax-reflink", v2.

The patchset fsdax-rmap is aimed to support shared pages tracking for
fsdax.

It moves owner tracking from dax_assocaite_entry() to pmem device driver,
by introducing an interface ->memory_failure() for struct pagemap.  This
interface is called by memory_failure() in mm, and implemented by pmem
device.

Then call holder operations to find the filesystem which the corrupted
data located in, and call filesystem handler to track files or metadata
associated with this page.

Finally we are able to try to fix the corrupted data in filesystem and do
other necessary processing, such as killing processes who are using the
files affected.

The call trace is like this:
memory_failure()
|* fsdax case
|------------
|pgmap->ops->memory_failure()      => pmem_pgmap_memory_failure()
| dax_holder_notify_failure()      =>
|  dax_device->holder_ops->notify_failure() =>
|                                     - xfs_dax_notify_failure()
|  |* xfs_dax_notify_failure()
|  |--------------------------
|  |   xfs_rmap_query_range()
|  |    xfs_dax_failure_fn()
|  |    * corrupted on metadata
|  |       try to recover data, call xfs_force_shutdown()
|  |    * corrupted on file data
|  |       try to recover data, call mf_dax_kill_procs()
|* normal case
|-------------
|mf_generic_kill_procs()

The patchset fsdax-reflink attempts to add CoW support for fsdax, and
takes XFS, which has both reflink and fsdax features, as an example.

One of the key mechanisms needed to be implemented in fsdax is CoW.  Copy
the data from srcmap before we actually write data to the destination
iomap.  And we just copy range in which data won't be changed.

Another mechanism is range comparison.  In page cache case, readpage() is
used to load data on disk to page cache in order to be able to compare
data.  In fsdax case, readpage() does not work.  So, we need another
compare data with direct access support.

With the two mechanisms implemented in fsdax, we are able to make reflink
and fsdax work together in XFS.

This patch (of 14):

To easily track filesystem from a pmem device, we introduce a holder for
dax_device structure, and also its operation.  This holder is used to
remember who is using this dax_device:

 - When it is the backend of a filesystem, the holder will be the
   instance of this filesystem.
 - When this pmem device is one of the targets in a mapped device, the
   holder will be this mapped device.  In this case, the mapped device
   has its own dax_device and it will follow the first rule.  So that we
   can finally track to the filesystem we needed.

The holder and holder_ops will be set when filesystem is being mounted,
or an target device is being activated.

Link: https://lkml.kernel.org/r/20220603053738.1218681-1-ruansy.fnst@fujitsu.com
Link: https://lkml.kernel.org/r/20220603053738.1218681-2-ruansy.fnst@fujitsu.com
Signed-off-by: Shiyang Ruan <ruansy.fnst@fujitsu.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Dan Williams <dan.j.wiliams@intel.com>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jane Chu <jane.chu@oracle.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Dan Williams <dan.j.williams@intel.com>
Cc: Goldwyn Rodrigues <rgoldwyn@suse.com>
Cc: Ritesh Harjani <riteshh@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

tools: add selftests to hmm for COW in device memory

The objective is to test device migration mechanism in pages marked as
COW, for private and coherent device type.  In case of writing to COW
private page(s), a page fault will migrate pages back to system memory
first.  Then, these pages will be duplicated.  In case of COW device
coherent type, pages are duplicated directly from device memory.

Link: https://lkml.kernel.org/r/20220715150521.18165-15-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

tools: add hmm gup tests for device coherent type

The intention is to test hmm device coherent type under different get user
pages paths.  Also, test gup with FOLL_LONGTERM flag set in device
coherent pages.  These pages should get migrated back to system memory.

Link: https://lkml.kernel.org/r/20220715150521.18165-14-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

tools: update test_hmm script to support SP config

Add two more parameters to set spm_addr_dev0 & spm_addr_dev1 addresses.
These two parameters configure the start SP addresses for each device in
test_hmm driver.  Consequently, this configures zone device type as
coherent.

Link: https://lkml.kernel.org/r/20220715150521.18165-13-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

tools: update hmm-test to support device coherent type

Test cases such as migrate_fault and migrate_multiple, were modified to
explicit migrate from device to sys memory without the need of page
faults, when using device coherent type.

Snapshot test case updated to read memory device type first and based on
that, get the proper returned results migrate_ping_pong test case added to
test explicit migration from device to sys memory for both private and
coherent zone types.

Helpers to migrate from device to sys memory and vicerversa were also
added.

Link: https://lkml.kernel.org/r/20220715150521.18165-12-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

lib: add support for device coherent type in test_hmm

Device Coherent type uses device memory that is coherently accesible by
the CPU.  This could be shown as SP (special purpose) memory range at the
BIOS-e820 memory enumeration.  If no SP memory is supported in system,
this could be faked by setting CONFIG_EFI_FAKE_MEMMAP.

Currently, test_hmm only supports two different SP ranges of at least
256MB size. This could be specified in the kernel parameter variable
efi_fake_mem. Ex. Two SP ranges of 1GB starting at 0x100000000 &
0x140000000 physical address. Ex.
efi_fake_mem=1G@0x100000000:0x40000,1G@0x140000000:0x40000

Private and coherent device mirror instances can be created in the same
probed.  This is done by passing the module parameters spm_addr_dev0 &
spm_addr_dev1.  In this case, it will create four instances of
device_mirror.  The first two correspond to private device type, the last
two to coherent type.  Then, they can be easily accessed from user space
through /dev/hmm_mirror<num_device>.  Usually num_device 0 and 1 are for
private, and 2 and 3 for coherent types.  If no module parameters are
passed, two instances of private type device_mirror will be created only.

Link: https://lkml.kernel.org/r/20220715150521.18165-11-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alistair Poppple <apopple@nvidia.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

lib: test_hmm add module param for zone device type

In order to configure device coherent in test_hmm, two module parameters
should be passed, which correspond to the SP start address of each device
(2) spm_addr_dev0 & spm_addr_dev1.  If no parameters are passed, private
device type is configured.

Link: https://lkml.kernel.org/r/20220715150521.18165-10-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alistair Poppple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

lib: test_hmm add ioctl to get zone device type

Add new ioctl cmd to query zone device type.  This will be used once the
test_hmm adds zone device coherent type.

Link: https://lkml.kernel.org/r/20220715150521.18165-9-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alistair Poppple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

drm/amdkfd: add SPM support for SVM

When CPU is connected throug XGMI, it has coherent access to VRAM
resource.  In this case that resource is taken from a table in the device
gmc aperture base.  This resource is used along with the device type,
which could be DEVICE_PRIVATE or DEVICE_COHERENT to create the device page
map region.

Also, MIGRATE_VMA_SELECT_DEVICE_COHERENT flag is selected for coherent
type case during migration to device.

Link: https://lkml.kernel.org/r/20220715150521.18165-8-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/gup: migrate device coherent pages when pinning instead of failing

Currently any attempts to pin a device coherent page will fail.  This is
because device coherent pages need to be managed by a device driver, and
pinning them would prevent a driver from migrating them off the device.

However this is no reason to fail pinning of these pages.  These are
coherent and accessible from the CPU so can be migrated just like pinning
ZONE_MOVABLE pages.  So instead of failing all attempts to pin them first
try migrating them out of ZONE_DEVICE.

[hch@lst.de: rebased to the split device memory checks, moved migrate_device_page to migrate_device.c]
Link: https://lkml.kernel.org/r/20220715150521.18165-7-alex.sierra@amd.com
Signed-off-by: Alistair Popple <apopple@nvidia.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: add device coherent vma selection for memory migration

This case is used to migrate pages from device memory, back to system
memory.  Device coherent type memory is cache coherent from device and CPU
point of view.

Link: https://lkml.kernel.org/r/20220715150521.18165-6-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alistair Poppple <apopple@nvidia.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: handling Non-LRU pages returned by vm_normal_pages

With DEVICE_COHERENT, we'll soon have vm_normal_pages() return
device-managed anonymous pages that are not LRU pages.  Although they
behave like normal pages for purposes of mapping in CPU page, and for COW.
They do not support LRU lists, NUMA migration or THP.

Callers to follow_page() currently don't expect ZONE_DEVICE pages,
however, with DEVICE_COHERENT we might now return ZONE_DEVICE.  Check for
ZONE_DEVICE pages in applicable users of follow_page() as well.

Link: https://lkml.kernel.org/r/20220715150521.18165-5-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com> [v2]
Reviewed-by: Alistair Popple <apopple@nvidia.com> [v6]
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: add zone device coherent type memory support

Device memory that is cache coherent from device and CPU point of view.
This is used on platforms that have an advanced system bus (like CAPI or
CXL).  Any page of a process can be migrated to such memory.  However, no
one should be allowed to pin such memory so that it can always be evicted.

[hch@lst.de: rebased ontop of the refcount changes, remove is_dev_private_or_coherent_page]
Link: https://lkml.kernel.org/r/20220715150521.18165-4-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Felix Kuehling <Felix.Kuehling@amd.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: move page zone helpers from mm.h to mmzone.h

It makes more sense to have these helpers in zone specific header
file, rather than the generic mm.h

Link: https://lkml.kernel.org/r/20220715150521.18165-3-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: David Hildenbrand <david@redhat.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: rename is_pinnable_page() to is_longterm_pinnable_page()

Patch series "Add MEMORY_DEVICE_COHERENT for coherent device memory
mapping", v9.

This patch series introduces MEMORY_DEVICE_COHERENT, a type of memory
owned by a device that can be mapped into CPU page tables like
MEMORY_DEVICE_GENERIC and can also be migrated like MEMORY_DEVICE_PRIVATE.

This patch series is mostly self-contained except for a few places where
it needs to update other subsystems to handle the new memory type.

System stability and performance are not affected according to our ongoing
testing, including xfstests.

How it works: The system BIOS advertises the GPU device memory (aka VRAM)
as SPM (special purpose memory) in the UEFI system address map.

The amdgpu driver registers the memory with devmap as
MEMORY_DEVICE_COHERENT using devm_memremap_pages.  The initial user for
this hardware page migration capability is the Frontier supercomputer
project.  This functionality is not AMD-specific.  We expect other GPU
vendors to find this functionality useful, and possibly other hardware
types in the future.

Our test nodes in the lab are similar to the Frontier configuration, with
.5 TB of system memory plus 256 GB of device memory split across 4 GPUs,
all in a single coherent address space.  Page migration is expected to
improve application efficiency significantly.  We will report empirical
results as they become available.

Coherent device type pages at gup are now migrated back to system memory
if they are being pinned long-term (FOLL_LONGTERM).  The reason is, that
long-term pinning would interfere with the device memory manager owning
the device-coherent pages (e.g.  evictions in TTM).  These series
incorporate Alistair Popple patches to do this migration from
pin_user_pages() calls.  hmm_gup_test has been added to hmm-test to test
different get user pages calls.

This series includes handling of device-managed anonymous pages returned
by vm_normal_pages.  Although they behave like normal pages for purposes
of mapping in CPU page tables and for COW, they do not support LRU lists,
NUMA migration or THP.

We also introduced a FOLL_LRU flag that adds the same behaviour to
follow_page and related APIs, to allow callers to specify that they expect
to put pages on an LRU list.

This patch (of 14):

is_pinnable_page() and folio_is_pinnable() are renamed to
is_longterm_pinnable_page() and folio_is_longterm_pinnable() respectively.
These functions are used in the FOLL_LONGTERM flag context.

Link: https://lkml.kernel.org/r/20220715150521.18165-1-alex.sierra@amd.com
Link: https://lkml.kernel.org/r/20220715150521.18165-2-alex.sierra@amd.com
Signed-off-by: Alex Sierra <alex.sierra@amd.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Felix Kuehling <Felix.Kuehling@amd.com>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/vm: add protection_keys tests to run_vmtests

Add "protected_keys" tests to "run_vmtests.sh" would help run all VM
related tests from a single shell script.

[kalpana.shetty@amd.com: Shuah Khan's review comments incorporated, added -x executable check]
Link: https://lkml.kernel.org/r/20220617202931.357-1-kalpana.shetty@amd.com
Link: https://lkml.kernel.org/r/20220610090704.296-1-kalpana.shetty@amd.com
Link: https://lkml.kernel.org/r/20220531102556.388-1-kalpana.shetty@amd.com
Signed-off-by: Kalpana Shetty <kalpana.shetty@amd.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/lru_sort: fix potential memory leak in damon_lru_sort_init()

damon_lru_sort_init() returns an error when damon_select_ops() fails
without freeing 'ctx' which allocated before.  This commit fixes the
potential memory leak by freeing 'ctx' under the situation.

Link: https://lkml.kernel.org/r/20220714170458.49727-1-sj@kernel.org
Fixes: 40e983cca927 ("mm/damon: introduce DAMON-based LRU-lists Sorting")
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/vm: only run 128TBswitch with 5-level paging

The test va_128TBswitch.c expects to be able to pass mmap an address hint
and length that cross the address 1<<47.  On x86_64, this is not possible
without 5-level page tables, so the test fails.

The test is already only run on 64-bit powerpc and x86_64 archs, but this
patch adds an additional check on x86_64 that skips the test if
PG_TABLE_LEVELS < 5.  There is precedent for checking /proc/config.gz in
selftests, e.g.  in selftests/firmware.

Running the tests produces the desired output:

sudo make -C tools/testing/selftests TARGETS=vm run_tests
---------------------------
running ./va_128TBswitch.sh
---------------------------
./va_128TBswitch.sh: PG_TABLE_LEVELS=4, must be >= 5 to run this test
[SKIP]
-------------------------------

[adam@wowsignal.io: restrict the check to x86_64]
Link: https://lkml.kernel.org/r/20220628163654.337600-1-adam@wowsignal.io
[adam@wowsignal.io: fix formatting issues, rename "die" to "fail"]
Link: https://lkml.kernel.org/r/20220701163030.415735-1-adam@wowsignal.io
Link: https://lkml.kernel.org/r/20220627163912.5581-1-adam@wowsignal.io
Signed-off-by: Adam Sindelar <adam@wowsignal.io>
Cc: Adam Sindelar <ats@fb.com>
Cc: David Vernet <void@manifault.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: try to free transhuge swapcache when possible

Transhuge swapcaches won't be freed in __collapse_huge_page_copy().  It's
because release_pte_page() is not called for these pages and thus
free_page_and_swap_cache can't grab the page lock.  These pages won't be
freed from swap cache even if we are the only user until next time
reclaim.  It shouldn't hurt indeed, but we could try to free these pages
to save more memory for system.

Link: https://lkml.kernel.org/r/20220625092816.4856-8-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: NeilBrown <neilb@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zach O'Keefe <zokeefe@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: remove unneeded return value of khugepaged_add_pte_mapped_thp()

The return value of khugepaged_add_pte_mapped_thp() is always 0 and also
ignored.  Remove it to clean up the code.

Link: https://lkml.kernel.org/r/20220625092816.4856-7-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: NeilBrown <neilb@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: use helper macro __ATTR_RW

Use helper macro __ATTR_RW to define the khugepaged attributes.  Minor
readability improvement.

Link: https://lkml.kernel.org/r/20220625092816.4856-6-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: NeilBrown <neilb@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zach O'Keefe <zokeefe@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: minor cleanup for collapse_file

nr_none is always 0 for non-shmem case because the page can be read from
the backend store.  So when nr_none !  = 0, it must be in is_shmem case.
Also only adjust the nrpages and uncharge shmem when nr_none != 0 to save
cpu cycles.

Link: https://lkml.kernel.org/r/20220625092816.4856-5-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: NeilBrown <neilb@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: trivial typo and codestyle cleanup

Fix some typos and tweak the code to meet codestyle.  No functional change
intended.

Link: https://lkml.kernel.org/r/20220625092816.4856-4-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: NeilBrown <neilb@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: stop swapping in page when VM_FAULT_RETRY occurs

When do_swap_page returns VM_FAULT_RETRY, we do not retry here and thus
swap entry will remain in pagetable.  This will result in later failure.
So stop swapping in pages in this case to save cpu cycles.  As A further
optimization, mmap_lock is released when __collapse_huge_page_swapin()
fails to avoid relocking mmap_lock.  And "swapped_in++" is moved after
error handling to make it more accurate.

Link: https://lkml.kernel.org/r/20220625092816.4856-3-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: NeilBrown <neilb@suse.de>
Cc: Peter Xu <peterx@redhat.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zach O'Keefe <zokeefe@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: remove unneeded shmem_huge_enabled() check

Patch series "A few cleanup patches for khugepaged", v2.

This series contains a few cleaup patches to remove unneeded return value,
use helper macro, fix typos and so on.  More details can be found in the
respective changelogs.

This patch (of 7):

If we reach here, khugepaged_scan_mm_slot() has already made sure that
hugepage is enabled for shmem, via its call to hugepage_vma_check().
Remove this duplicated check.

Link: https://lkml.kernel.org/r/20220625092816.4856-1-linmiaohe@huawei.com
Link: https://lkml.kernel.org/r/20220625092816.4856-2-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: David Howells <dhowells@redhat.com>
Cc: NeilBrown <neilb@suse.de>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Peter Xu <peterx@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: sparsemem: drop unexpected word 'a' in comments

there is an unexpected word 'a' in the comments that need to be dropped

Link: https://lkml.kernel.org/r/24fbdae3.c86.1819a0f31b9.Coremail.chenxuebing@jari.cn
Signed-off-by: XueBing Chen <chenxuebing@jari.cn>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: hugetlb: kill set_huge_swap_pte_at()

Commit e5251fd43007 ("mm/hugetlb: introduce set_huge_swap_pte_at()
helper") add set_huge_swap_pte_at() to handle swap entries on
architectures that support hugepages consisting of contiguous ptes.  And
currently the set_huge_swap_pte_at() is only overridden by arm64.

set_huge_swap_pte_at() provide a sz parameter to help determine the number
of entries to be updated.  But in fact, all hugetlb swap entries contain
pfn information, so we can find the corresponding folio through the pfn
recorded in the swap entry, then the folio_size() is the number of entries
that need to be updated.

And considering that users will easily cause bugs by ignoring the
difference between set_huge_swap_pte_at() and set_huge_pte_at().  Let's
handle swap entries in set_huge_pte_at() and remove the
set_huge_swap_pte_at(), then we can call set_huge_pte_at() anywhere, which
simplifies our coding.

Link: https://lkml.kernel.org/r/20220626145717.53572-1-zhengqi.arch@bytedance.com
Signed-off-by: Qi Zheng <zhengqi.arch@bytedance.com>
Acked-by: Muchun Song <songmuchun@bytedance.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: make the annotations of available memory more accurate

Not all systems use swap, so estimating available memory would help to
prevent swapping or OOM of system that not use swap.

And we need to reserve some page cache to prevent swapping or thrashing.
If somebody is accessing the pages in pagecache, and if too much would be
freed, most accesses might mean reading data from disk, i.e.  thrashing.

Link: https://lkml.kernel.org/r/20220623020833.972979-1-yang.yang29@zte.com.cn
Signed-off-by: Yang Yang <yang.yang29@zte.com.cn>
Signed-off-by: CGEL ZTE <cgel.zte@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

zram: do not lookup algorithm in backends table

Always use crypto_has_comp() so that crypto can lookup module, call
usermodhelper to load the modules, wait for usermodhelper to finish and so
on.  Otherwise crypto will do all of these steps under CPU hot-plug lock
and this looks like too much stuff to handle under the CPU hot-plug lock.
Besides this can end up in a deadlock when usermodhelper triggers a code
path that attempts to lock the CPU hot-plug lock, that zram already holds.

An example of such deadlock:

- path A. zram grabs CPU hot-plug lock, execs /sbin/modprobe from crypto
  and waits for modprobe to finish

disksize_store
 zcomp_create
  __cpuhp_state_add_instance
   __cpuhp_state_add_instance_cpuslocked
    zcomp_cpu_up_prepare
     crypto_alloc_base
      crypto_alg_mod_lookup
       call_usermodehelper_exec
        wait_for_completion_killable
         do_wait_for_common
          schedule

- path B. async work kthread that brings in scsi device. It wants to
  register CPUHP states at some point, and it needs the CPU hot-plug
  lock for that, which is owned by zram.

async_run_entry_fn
 scsi_probe_and_add_lun
  scsi_mq_alloc_queue
   blk_mq_init_queue
    blk_mq_init_allocated_queue
     blk_mq_realloc_hw_ctxs
      __cpuhp_state_add_instance
       __cpuhp_state_add_instance_cpuslocked
        mutex_lock
         schedule

- path C. modprobe sleeps, waiting for all aync works to finish.

load_module
 do_init_module
  async_synchronize_full
   async_synchronize_cookie_domain
    schedule

[senozhatsky@chromium.org: add comment]
Link: https://lkml.kernel.org/r/20220624060606.1014474-1-senozhatsky@chromium.org
Link: https://lkml.kernel.org/r/20220622023501.517125-1-senozhatsky@chromium.org
Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm, docs: fix comments that mention mem_hotplug_end()

Comments that mention mem_hotplug_end() are confusing as there is no
function called mem_hotplug_end().  Fix them by replacing all the
occurences of mem_hotplug_end() in the comments with mem_hotplug_done().

[akpm@linux-foundation.org: grammatical fixes]
Link: https://lkml.kernel.org/r/20220620071516.1286101-1-p76091292@gs.ncku.edu.tw
Signed-off-by: Yun-Ze Li <p76091292@gs.ncku.edu.tw>
Cc: Souptick Joarder <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/smaps: add Pss_Dirty

Pss is the sum of the sizes of clean and dirty private pages, and the
proportional sizes of clean and dirty shared pages:

 Private = Private_Dirty + Private_Clean
 Shared_Proportional = Shared_Dirty_Proportional + Shared_Clean_Proportional
 Pss = Private + Shared_Proportional

The Shared*Proportional fields are not present in smaps, so it is not
always possible to determine how much of the Pss is from dirty pages and
how much is from clean pages.  This information can be useful for
measuring memory usage for the purpose of optimisation, since clean pages
can usually be discarded by the kernel immediately while dirty pages
cannot.

The smaps routines in the kernel already have access to this data, so add
a Pss_Dirty to show it to userspace.  Pss_Clean is not added since it can
be calculated from Pss and Pss_Dirty.

Link: https://lkml.kernel.org/r/20220620081251.2928103-1-vincent.whitchurch@axis.com
Signed-off-by: Vincent Whitchurch <vincent.whitchurch@axis.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: rmap: simplify the hugetlb handling when unmapping or migration

According to previous discussion [1], there are so many levels of
indenting to handle the hugetlb case when unmapping or migration.  We can
combine folio_test_anon() and huge_pmd_unshare() to save one level of
indenting, by adding a local variable and moving the VM_BUG_ON() a little
forward.

No intended functional changes in this patch.

[1] https://lore.kernel.org/all/0b986dc4-5843-3e2d-c2df-5a2e9f13e6ab@oracle.com/

Link: https://lkml.kernel.org/r/28414b1b96f095e838c1e548074f8e0fc70d78cf.1655724713.git.baolin.wang@linux.alibaba.com
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/madvise: minor cleanup for swapin_walk_pmd_entry()

Passing index to pte_offset_map_lock() directly so the below calculation
can be avoided. Rename orig_pte to ptep as it's not changed. Also use
helper is_swap_pte() to improve the readability. No functional change
intended.

[akpm@linux-foundation.org: reduce scope of `ptep']
Link: https://lkml.kernel.org/r/20220618090527.37843-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: hugetlb: remove minimum_order variable

commit 641844f5616d ("mm/hugetlb: introduce minimum hugepage order") fixed
a static checker warning and introduced a global variable minimum_order to
fix the warning.  However, the local variable in
dissolve_free_huge_pages() can be initialized to
huge_page_order(&default_hstate) to fix the warning.

So remove minimum_order to simplify the code.

Link: https://lkml.kernel.org/r/20220616033846.96937-1-songmuchun@bytedance.com
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: memory_hotplug: make hugetlb_optimize_vmemmap compatible with memmap_on_memory

For now, the feature of hugetlb_free_vmemmap is not compatible with the
feature of memory_hotplug.memmap_on_memory, and hugetlb_free_vmemmap takes
precedence over memory_hotplug.memmap_on_memory.  However, someone wants
to make memory_hotplug.memmap_on_memory takes precedence over
hugetlb_free_vmemmap since memmap_on_memory makes it more likely to
succeed memory hotplug in close-to-OOM situations.  So the decision of
making hugetlb_free_vmemmap take precedence is not wise and elegant.

The proper approach is to have hugetlb_vmemmap.c do the check whether the
section which the HugeTLB pages belong to can be optimized.  If the
section's vmemmap pages are allocated from the added memory block itself,
hugetlb_free_vmemmap should refuse to optimize the vmemmap, otherwise, do
the optimization.  Then both kernel parameters are compatible.  So this
patch introduces VmemmapSelfHosted to mask any non-optimizable vmemmap
pages.  The hugetlb_vmemmap can use this flag to detect if a vmemmap page
can be optimized.

[songmuchun@bytedance.com: walk vmemmap page tables to avoid false-positive]
Link: https://lkml.kernel.org/r/20220620110616.12056-3-songmuchun@bytedance.com
Link: https://lkml.kernel.org/r/20220617135650.74901-3-songmuchun@bytedance.com
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Co-developed-by: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Oscar Salvador <osalvador@suse.de>
Acked-by: David Hildenbrand <david@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Xiongchun Duan <duanxiongchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: memory_hotplug: enumerate all supported section flags

Patch series "make hugetlb_optimize_vmemmap compatible with
memmap_on_memory", v3.

This series makes hugetlb_optimize_vmemmap compatible with
memmap_on_memory.

This patch (of 2):

We are almost running out of section flags, only one bit is available in
the worst case (powerpc with 256k pages).  However, there are still some
free bits (in ->section_mem_map) on other architectures (e.g.  x86_64 has
10 bits available, arm64 has 8 bits available with worst case of 64K
pages).  We have hard coded those numbers in code, it is inconvenient to
use those bits on other architectures except powerpc.  So transfer those
section flags to enumeration to make it easy to add new section flags in
the future.  Also, move SECTION_TAINT_ZONE_DEVICE into the scope of
CONFIG_ZONE_DEVICE to save a bit on non-zone-device case.

[songmuchun@bytedance.com: replace enum with defines per David]
Link: https://lkml.kernel.org/r/20220620110616.12056-2-songmuchun@bytedance.com
Link: https://lkml.kernel.org/r/20220617135650.74901-1-songmuchun@bytedance.com
Link: https://lkml.kernel.org/r/20220617135650.74901-2-songmuchun@bytedance.com
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Paul E. McKenney <paulmck@kernel.org>
Cc: Xiongchun Duan <duanxiongchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert __delete_from_swap_cache() to a folio

All callers now have a folio, so convert the entire function to operate
on folios.

Link: https://lkml.kernel.org/r/20220617175020.717127-23-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert delete_from_swap_cache() to take a folio

All but one caller already has a folio, so convert it to use a folio.

Link: https://lkml.kernel.org/r/20220617175020.717127-22-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: convert page_swap_flags to folio_swap_flags

The only caller already has a folio, so push the folio->page conversion
down a level.

Link: https://lkml.kernel.org/r/20220617175020.717127-21-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: convert destroy_compound_page() to destroy_large_folio()

All callers now have a folio, so push the folio->page conversion
down to this function.

[akpm@linux-foundation.org: uninline destroy_large_folio() to fix build issue]
Link: https://lkml.kernel.org/r/20220617175020.717127-20-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert __page_cache_release() to use a folio

All the callers now have a folio.  Saves several calls to compound_head,
totalling 502 bytes of text.

Link: https://lkml.kernel.org/r/20220617175020.717127-19-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert __put_compound_page() to __folio_put_large()

All the callers now have a folio, so pass it in.  This doesn't
save any text, but it does save a call to compound_head() as
folio_test_hugetlb() does not contain a call like PageHuge() does.

Link: https://lkml.kernel.org/r/20220617175020.717127-18-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert __put_single_page() to __folio_put_small()

Saves 56 bytes of text by removing a call to compound_head().

Link: https://lkml.kernel.org/r/20220617175020.717127-17-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert __put_page() to __folio_put()

Saves 11 bytes of text by removing a check of PageTail.

Link: https://lkml.kernel.org/r/20220617175020.717127-16-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert put_pages_list to use folios

Pages linked through the LRU list cannot be tail pages as ->compound_head
is in a union with one of the words of the list_head, and they cannot
be ZONE_DEVICE pages as ->pgmap is in a union with the same word.
Saves 60 bytes of text by removing a call to page_is_fake_head().

Link: https://lkml.kernel.org/r/20220617175020.717127-15-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert release_pages to use a folio internally

This function was already calling compound_head(), but now it can
cache the result of calling compound_head() and avoid calling it again.
Saves 299 bytes of text by avoiding various calls to compound_page()
and avoiding checks of PageTail.

Link: https://lkml.kernel.org/r/20220617175020.717127-14-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert try_to_free_swap to use a folio

Save a few calls to compound_head by converting the passed page to
a folio.  Reduces kernel text size by 74 bytes.

Link: https://lkml.kernel.org/r/20220617175020.717127-13-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: optimise lru_add_drain_cpu()

Do the per-cpu dereferencing of the fbatches once which saves 14 bytes
of text and several percpu relocations.

Link: https://lkml.kernel.org/r/20220617175020.717127-12-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: pull the CPU conditional out of __lru_add_drain_all()

The function is too long, so pull this complicated conditional out into
cpu_needs_drain().  This ends up shrinking the text by 14 bytes,
by allowing GCC to cache the result of calling per_cpu() instead of
relocating each lookup individually.

Link: https://lkml.kernel.org/r/20220617175020.717127-11-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: rename lru_pvecs to cpu_fbatches

No change to generated code, but this struct no longer contains any
pagevecs, and not all the folio batches it contains are lru.

Link: https://lkml.kernel.org/r/20220617175020.717127-10-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert activate_page to a folio_batch

Rename it to just 'activate', saving 696 bytes of text from removals
of compound_page() and the pagevec_lru_move_fn() infrastructure.
Inline need_activate_page_drain() into its only caller.

Link: https://lkml.kernel.org/r/20220617175020.717127-9-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert lru_lazyfree to a folio_batch

Using folios instead of pages removes several calls to compound_head(),
shrinking the kernel by 1089 bytes of text.

Link: https://lkml.kernel.org/r/20220617175020.717127-8-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert lru_deactivate to a folio_batch

Using folios instead of pages shrinks deactivate_page() and
lru_deactivate_fn() by 778 bytes between them.

Link: https://lkml.kernel.org/r/20220617175020.717127-7-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert lru_deactivate_file to a folio_batch

Use a folio throughout lru_deactivate_file_fn(), removing many hidden
calls to compound_head().  Shrinks the kernel by 864 bytes of text.

Link: https://lkml.kernel.org/r/20220617175020.717127-6-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: convert lru_add to a folio_batch

When adding folios to the LRU for the first time, the LRU flag will
already be clear, so skip the test-and-clear part of moving from one
LRU to another.

Removes 285 bytes from kernel text, mostly due to removing
__pagevec_lru_add().

Link: https://lkml.kernel.org/r/20220617175020.717127-5-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: make __pagevec_lru_add static

__pagevec_lru_add has no callers outside swap.c, so make it static,
and move it to a more logical position in the file.

Link: https://lkml.kernel.org/r/20220617175020.717127-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: add folio_batch_move_lru()

Start converting the LRU from pagevecs to folio_batches.

Combine the functionality of pagevec_add_and_need_flush() with
pagevec_lru_move_fn() in the new folio_batch_add_and_move().

Convert the lru_rotate pagevec to a folio_batch.

Adds 223 bytes total to kernel text, because we're duplicating
infrastructure.  This will be more than made up for in future patches.

Link: https://lkml.kernel.org/r/20220617175020.717127-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: add folios_put()

Patch series "Convert the swap code to be more folio-based".

There's still more to do with the swap code, but this reaps a lot of the
folio benefit.  More than 4kB of kernel text saved (with the UEK7 kernel
config).  I don't know how much that's going to translate into CPU
savings, but some of those compound_head() calls are on every page free,
so it should be noticable.  It might even be noticable just from an
I-cache consumption perspective.

This patch (of 22):

This is just a wrapper around release_pages() for now.  Place the
prototype in mm.h along with folio_put() and folio_put_refs().

Link: https://lkml.kernel.org/r/20220617175020.717127-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20220617175020.717127-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/vmscan: convert reclaim_pages() to use a folio

Remove a few hidden calls to compound_head, saving 76 bytes of text.

Link: https://lkml.kernel.org/r/20220617154248.700416-6-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/vmscan: convert shrink_active_list() to use a folio

Remove a few hidden calls to compound_head, saving 411 bytes of text.

Link: https://lkml.kernel.org/r/20220617154248.700416-5-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/vmscan: convert move_pages_to_lru() to use a folio

Remove a few hidden calls to compound_head, saving 387 bytes of text on
my test configuration.

Link: https://lkml.kernel.org/r/20220617154248.700416-4-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/vmscan: convert isolate_lru_pages() to use a folio

Remove a few hidden calls to compound_head, saving 279 bytes of text.

Link: https://lkml.kernel.org/r/20220617154248.700416-3-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/vmscan: convert reclaim_clean_pages_from_list() to folios

Patch series "nvert much of vmscan to folios"

vmscan always operates on folios since it puts the pages on the LRU list.
Switching all of these functions from pages to folios saves 1483 bytes of
text from removing all the baggage around calling compound_page() and
similar functions.

This patch (of 5):

This is a straightforward conversion which removes several hidden calls
to compound_head, saving 330 bytes of kernel text.

Link: https://lkml.kernel.org/r/20220617154248.700416-1-willy@infradead.org
Link: https://lkml.kernel.org/r/20220617154248.700416-2-willy@infradead.org
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/mprotect: try avoiding write faults for exclusive anonymous pages when changing protection

Similar to our MM_CP_DIRTY_ACCT handling for shared, writable mappings, we
can try mapping anonymous pages in a private writable mapping writable if
they are exclusive, the PTE is already dirty, and no special handling
applies.  Mapping the anonymous page writable is essentially the same
thing the write fault handler would do in this case.

Special handling is required for uffd-wp and softdirty tracking, so take
care of that properly.  Also, leave PROT_NONE handling alone for now; in
the future, we could similarly extend the logic in do_numa_page() or use
pte_mk_savedwrite() here.

While this improves mprotect(PROT_READ)+mprotect(PROT_READ|PROT_WRITE)
performance, it should also be a valuable optimization for uffd-wp, when
un-protecting.

This has been previously suggested by Peter Collingbourne in [1], relevant
in the context of the Scudo memory allocator, before we had
PageAnonExclusive.

This commit doesn't add the same handling for PMDs (i.e., anonymous THP,
anonymous hugetlb); benchmark results from Andrea indicate that there are
minor performance gains, so it's might still be valuable to streamline
that logic for all anonymous pages in the future.

As we now also set MM_CP_DIRTY_ACCT for private mappings, let's rename it
to MM_CP_TRY_CHANGE_WRITABLE, to make it clearer what's actually
happening.

Micro-benchmark courtesy of Andrea:

===
 #define _GNU_SOURCE
 #include <sys/mman.h>
 #include <stdlib.h>
 #include <string.h>
 #include <stdio.h>
 #include <unistd.h>

 #define SIZE (1024*1024*1024)

int main(int argc, char *argv[])
{
char *p;
if (posix_memalign((void **)&p, sysconf(_SC_PAGESIZE)*512, SIZE))
perror("posix_memalign"), exit(1);
if (madvise(p, SIZE, argc > 1 ? MADV_HUGEPAGE : MADV_NOHUGEPAGE))
perror("madvise");
explicit_bzero(p, SIZE);
for (int loops = 0; loops < 40; loops++) {
if (mprotect(p, SIZE, PROT_READ))
perror("mprotect"), exit(1);
if (mprotect(p, SIZE, PROT_READ|PROT_WRITE))
perror("mprotect"), exit(1);
explicit_bzero(p, SIZE);
}
}
===

Results on my Ryzen 9 3900X:

Stock 10 runs (lower is better):   AVG 6.398s, STDEV 0.043
Patched 10 runs (lower is better): AVG 3.780s, STDEV 0.026

===

[1] https://lkml.kernel.org/r/20210429214801.2583336-1-pcc@google.com

Link: https://lkml.kernel.org/r/20220614093629.76309-1-david@redhat.com
Signed-off-by: David Hildenbrand <david@redhat.com>
Suggested-by: Peter Collingbourne <pcc@google.com>
Acked-by: Peter Xu <peterx@redhat.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>