zsmalloc-zs_object_copy-replace-email-link-to-doc-checkpatch-fixes

ERROR: code indent should use tabs where possible
#26: FILE: mm/zsmalloc.c:1562:
+         * kunmap_atomic(d_addr). For more details see$

ERROR: code indent should use tabs where possible
#27: FILE: mm/zsmalloc.c:1563:
+         * Documentation/mm/highmem$

total: 2 errors, 0 warnings, 10 lines checked

NOTE: For some of the reported defects, checkpatch may be able to
      mechanically convert to the typical style using --fix or --fix-inplace.

NOTE: Whitespace errors detected.
      You may wish to use scripts/cleanpatch or scripts/cleanfile

./patches/zsmalloc-zs_object_copy-replace-email-link-to-doc.patch has style problems, please review.

NOTE: If any of the errors are false positives, please report
      them to the maintainer, see CHECKPATCH in MAINTAINERS.

Please run checkpatch prior to sending patches

Cc: Alexey Romanov <avromanov@sberdevices.ru>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Cc: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

zsmalloc: zs_object_copy: replace email link to doc

Emails are not documentation.

Link: https://lkml.kernel.org/r/20220815144825.39001-1-avromanov@sberdevices.ru
Signed-off-by: Alexey Romanov <avromanov@sberdevices.ru>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm-oom_kill-add-trace-logs-in-process_mrelease-system-call-fix

reduce 80-col overflows

Cc: Charan Teja Kalla <quic_charante@quicinc.com>
Cc: Pratyush Brahma <quic_pbrahma@quicinc.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: oom_kill: add trace logs in process_mrelease() system call

The process_mrelease() system call is used to release the memory of a
dying process from the context of the caller, which is similar to and uses
the functions of the oom reaper logic.  There exist trace logs for a
process when reaped by the oom reaper.  Just extend the same to when done
by the process_mrelease() system call.

This patch provides information on how much memory is freed from a process
which is being reaped.  Adding trace events in the process_mrelease() path
when process is being reaped would enable more holistic debug as it
happens in oom_reap_task_mm() currently.

This extends the debug functionality for the events as described in [1] to
the process_mrelease() system call.  Now the coverage of trace events is
complete.

[1] https://lore.kernel.org/all/20170530185231.GA13412@castle/T/#u

Link: https://lkml.kernel.org/r/20220816060017.17996-1-pbrahma@qti.qualcomm.com
Signed-off-by: Charan Teja Kalla <quic_charante@quicinc.com>
Signed-off-by: Pratyush Brahma <quic_pbrahma@quicinc.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: migrate: do not retry 10 times for the subpages of fail-to-migrate THP

If THP is failed to migrate due to -ENOSYS or -ENOMEM case, the THP will
be split, and the subpages of fail-to-migrate THP will be tried to migrate
again, so we should not account the retry counter in the second loop,
since we already accounted 'nr_thp_failed' in the first loop.

Moreover we also do not need retry 10 times for -EAGAIN case for the
subpages of fail-to-migrate THP in the second loop, since we already
regarded the THP as migration failure, and save some migration time (for
the worst case, will try 512 * 10 times) according to previous discussion
[1].

[1] https://lore.kernel.org/linux-mm/87r13a7n04.fsf@yhuang6-desk2.ccr.corp.intel.com/

Link: https://lkml.kernel.org/r/20220817081408.513338-9-ying.huang@intel.com
Tested-by: "Huang, Ying" <ying.huang@intel.com>
Signed-off-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Oscar Salvador <osalvador@suse.de>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

migrate_pages(): fix failure counting for retry

After 10 retries, we will give up and the remaining pages will be counted
as failure in nr_failed and nr_thp_failed.  We should count the failure in
nr_failed_pages too.  This is done in this patch.

Link: https://lkml.kernel.org/r/20220817081408.513338-8-ying.huang@intel.com
Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

migrate_pages(): fix failure counting for THP splitting

If THP is failed to be migrated, it may be split and retry.  But after
splitting, the head page will be left in "from" list, although THP
migration failure has been counted already.  If the head page is failed to
be migrated too, the failure will be counted twice incorrectly.  So this
is fixed in this patch via moving the head page of THP after splitting to
"thp_split_pages" too.

Link: https://lkml.kernel.org/r/20220817081408.513338-7-ying.huang@intel.com
Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

migrate_pages(): fix failure counting for THP on -ENOSYS

If THP or hugetlbfs page migration isn't supported, unmap_and_move() or
unmap_and_move_huge_page() will return -ENOSYS.  For THP, splitting will
be tried, but if splitting doesn't succeed, the THP will be left in "from"
list wrongly.  If some other pages are retried, the THP migration failure
will counted again.  This is fixed via moving the failure THP from "from"
to "ret_pages".

Another issue of the original code is that the unsupported failure
processing isn't consistent between THP and hugetlbfs page.  Make them
consistent in this patch to make the code easier to be understood too.

Link: https://lkml.kernel.org/r/20220817081408.513338-6-ying.huang@intel.com
Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

migrate_pages(): fix failure counting for THP subpages retrying

If THP is failed to be migrated for -ENOSYS and -ENOMEM, the THP will be
split into thp_split_pages, and after other pages are migrated, pages in
thp_split_pages will be migrated with no_subpage_counting == true, because
its failure have been counted already.  If some pages in thp_split_pages
are retried during migration, we should not count their failure if
no_subpage_counting == true too.  This is done this patch to fix the
failure counting for THP subpages retrying.

Link: https://lkml.kernel.org/r/20220817081408.513338-5-ying.huang@intel.com
Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

migrate_pages(): fix THP failure counting for -ENOMEM

In unmap_and_move(), if the new THP cannot be allocated, -ENOMEM will be
returned, and migrate_pages() will try to split the THP unless "reason" is
MR_NUMA_MISPLACED (that is, nosplit == true).  But when nosplit == true,
the THP migration failure will not be counted.

This is incorrect, so in this patch, the THP migration failure will be
counted for -ENOMEM regardless of nosplit is true or false.  The nr_failed
counting isn't fixed because it's not used.  Added some comments for it
per Baolin's suggestion.

Link: https://lkml.kernel.org/r/20220817081408.513338-4-ying.huang@intel.com
Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

migrate_pages(): remove unnecessary list_safe_reset_next()

Before commit b5bade978e9b ("mm: migrate: fix the return value of
migrate_pages()"), the tail pages of THP will be put in the "from"
list directly.  So one of the loop cursors (page2) needs to be reset,
as is done in try_split_thp() via list_safe_reset_next().  But after
the commit, the tail pages of THP will be put in a dedicated
list (thp_split_pages).  That is, the "from" list will not be changed
during splitting.  So, it's unnecessary to call list_safe_reset_next()
anymore.

This is a code cleanup, no functionality changes are expected.

Link: https://lkml.kernel.org/r/20220817081408.513338-3-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

migrate: fix syscall move_pages() return value for failure

Patch series "migrate_pages(): fix several bugs in error path", v3.

During review the code of migrate_pages() and build a test program for
it.  Several bugs in error path are identified and fixed in this
series.

Most patches are tested via

- Apply error-inject.patch in Linux kernel
- Compile test-migrate.c (with -lnuma)
- Test with test-migrate.sh

error-inject.patch, test-migrate.c, and test-migrate.sh are as below.
It turns out that error injection is an important tool to fix bugs in
error path.

This patch (of 8):

The return value of move_pages() syscall is incorrect when counting
the remaining pages to be migrated.  For example, for the following
test program,

"
 #define _GNU_SOURCE

 #include <stdbool.h>
 #include <stdio.h>
 #include <string.h>
 #include <stdlib.h>
 #include <errno.h>

 #include <fcntl.h>
 #include <sys/uio.h>
 #include <sys/mman.h>
 #include <sys/types.h>
 #include <unistd.h>
 #include <numaif.h>
 #include <numa.h>

 #ifndef MADV_FREE
 #define MADV_FREE 8 /* free pages only if memory pressure */
 #endif

 #define ONE_MB (1024 * 1024)
 #define MAP_SIZE (16 * ONE_MB)
 #define THP_SIZE (2 * ONE_MB)
 #define THP_MASK (THP_SIZE - 1)

 #define ERR_EXIT_ON(cond, msg) \
 do { \
 int __cond_in_macro = (cond); \
 if (__cond_in_macro) \
 error_exit(__cond_in_macro, (msg)); \
 } while (0)

 void error_msg(int ret, int nr, int *status, const char *msg)
 {
 int i;

 fprintf(stderr, "Error: %s, ret : %d, error: %s\n",
 msg, ret, strerror(errno));

 if (!nr)
 return;
 fprintf(stderr, "status: ");
 for (i = 0; i < nr; i++)
 fprintf(stderr, "%d ", status[i]);
 fprintf(stderr, "\n");
 }

 void error_exit(int ret, const char *msg)
 {
 error_msg(ret, 0, NULL, msg);
 exit(1);
 }

 int page_size;

 bool do_vmsplice;
 bool do_thp;

 static int pipe_fds[2];
 void *addr;
 char *pn;
 char *pn1;
 void *pages[2];
 int status[2];

 void prepare()
 {
 int ret;
 struct iovec iov;

 if (addr) {
 munmap(addr, MAP_SIZE);
 close(pipe_fds[0]);
 close(pipe_fds[1]);
 }

 ret = pipe(pipe_fds);
 ERR_EXIT_ON(ret, "pipe");

 addr = mmap(NULL, MAP_SIZE, PROT_READ | PROT_WRITE,
     MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
 ERR_EXIT_ON(addr == MAP_FAILED, "mmap");
 if (do_thp) {
 ret = madvise(addr, MAP_SIZE, MADV_HUGEPAGE);
 ERR_EXIT_ON(ret, "advise hugepage");
 }

 pn = (char *)(((unsigned long)addr + THP_SIZE) & ~THP_MASK);
 pn1 = pn + THP_SIZE;
 pages[0] = pn;
 pages[1] = pn1;
 *pn = 1;

 if (do_vmsplice) {
 iov.iov_base = pn;
 iov.iov_len = page_size;
 ret = vmsplice(pipe_fds[1], &iov, 1, 0);
 ERR_EXIT_ON(ret < 0, "vmsplice");
 }

 status[0] = status[1] = 1024;
 }

 void test_migrate()
 {
 int ret;
 int nodes[2] = { 1, 1 };
 pid_t pid = getpid();

 prepare();
 ret = move_pages(pid, 1, pages, nodes, status, MPOL_MF_MOVE_ALL);
 error_msg(ret, 1, status, "move 1 page");

 prepare();
 ret = move_pages(pid, 2, pages, nodes, status, MPOL_MF_MOVE_ALL);
 error_msg(ret, 2, status, "move 2 pages, page 1 not mapped");

 prepare();
 *pn1 = 1;
 ret = move_pages(pid, 2, pages, nodes, status, MPOL_MF_MOVE_ALL);
 error_msg(ret, 2, status, "move 2 pages");

 prepare();
 *pn1 = 1;
 nodes[1] = 0;
 ret = move_pages(pid, 2, pages, nodes, status, MPOL_MF_MOVE_ALL);
 error_msg(ret, 2, status, "move 2 pages, page 1 to node 0");
 }

 int main(int argc, char *argv[])
 {
 numa_run_on_node(0);
 page_size = getpagesize();

 test_migrate();

 fprintf(stderr, "\nMake page 0 cannot be migrated:\n");
 do_vmsplice = true;
 test_migrate();

 fprintf(stderr, "\nTest THP:\n");
 do_thp = true;
 do_vmsplice = false;
 test_migrate();

 fprintf(stderr, "\nTHP: make page 0 cannot be migrated:\n");
 do_vmsplice = true;
 test_migrate();

 return 0;
 }
"

The output of the current kernel is,

"
Error: move 1 page, ret : 0, error: Success
status: 1
Error: move 2 pages, page 1 not mapped, ret : 0, error: Success
status: 1 -14
Error: move 2 pages, ret : 0, error: Success
status: 1 1
Error: move 2 pages, page 1 to node 0, ret : 0, error: Success
status: 1 0

Make page 0 cannot be migrated:
Error: move 1 page, ret : 0, error: Success
status: 1024
Error: move 2 pages, page 1 not mapped, ret : 1, error: Success
status: 1024 -14
Error: move 2 pages, ret : 0, error: Success
status: 1024 1024
Error: move 2 pages, page 1 to node 0, ret : 1, error: Success
status: 1024 1024
"

While the expected output is,

"
Error: move 1 page, ret : 0, error: Success
status: 1
Error: move 2 pages, page 1 not mapped, ret : 0, error: Success
status: 1 -14
Error: move 2 pages, ret : 0, error: Success
status: 1 1
Error: move 2 pages, page 1 to node 0, ret : 0, error: Success
status: 1 0

Make page 0 cannot be migrated:
Error: move 1 page, ret : 1, error: Success
status: 1024
Error: move 2 pages, page 1 not mapped, ret : 1, error: Success
status: 1024 -14
Error: move 2 pages, ret : 1, error: Success
status: 1024 1024
Error: move 2 pages, page 1 to node 0, ret : 2, error: Success
status: 1024 1024
"

Fix this via correcting the remaining pages counting.  With the fix,
the output for the test program as above is expected.

Link: https://lkml.kernel.org/r/20220817081408.513338-1-ying.huang@intel.com
Link: https://lkml.kernel.org/r/20220817081408.513338-2-ying.huang@intel.com
Fixes: 5984fabb6e82 ("mm: move_pages: report the number of non-attempted pages")
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Cc: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Yang Shi <shy828301@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/hmm-tests: add test for dirty bits

We were not correctly copying PTE dirty bits to pages during
migrate_vma_setup() calls.  This could potentially lead to data loss, so
add a test for this.

Link: https://lkml.kernel.org/r/23069a5c6e07d16d4c4f0951ff003591ffc4f656.1660635033.git-series.apopple@nvidia.com
Signed-off-by: Alistair Popple <apopple@nvidia.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Alex Sierra <alex.sierra@amd.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Felix Kuehling <felix.kuehling@amd.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Karol Herbst <kherbst@redhat.com>
Cc: Logan Gunthorpe <logang@deltatee.com>
Cc: Lyude Paul <lyude@redhat.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Paul Mackerras <paulus@ozlabs.org>
Cc: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: remove unused local variable dst_entry in copy_hugetlb_page_range()

Commit a0e4f7b82610 ("mm/hugetlb: make detecting shared pte more
reliable") modifies copy_hugetlb_page_range() such that
huge_ptep_get(dst_pte) and the local variable dst_entry is not used
explicitly in this function.

Remove this unused local variable dst_entry in copy_hugetlb_page_range().
No functional change.

Link: https://lkml.kernel.org/r/20220822082525.26071-1-lukas.bulwahn@gmail.com
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: make detecting shared pte more reliable

If the pagetables are shared, we shouldn't copy or take references.  Since
src could have unshared and dst shares with another vma, huge_pte_none()
is thus used to determine whether dst_pte is shared.  But this check isn't
reliable.  A shared pte could have pte none in pagetable in fact.  The
page count of ptep page should be checked here in order to reliably
determine whether pte is shared.

Link: https://lkml.kernel.org/r/20220816130553.31406-7-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: fix sysfs group leak in hugetlb_unregister_node()

The sysfs group per_node_hstate_attr_group and hstate_demote_attr_group
when h->demote_order != 0 are created in hugetlb_register_node().  But
these sysfs groups are not removed when unregister the node, thus sysfs
group is leaked.  Using sysfs_remove_group() to fix this issue.

Link: https://lkml.kernel.org/r/20220816130553.31406-6-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Fengwei Yin <fengwei.yin@intel.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: hugetlb_vmemmap: add missing smp_wmb() before set_pte_at()

The memory barrier smp_wmb() is needed to make sure that preceding stores
to the page contents become visible before the below set_pte_at() write.

Link: https://lkml.kernel.org/r/20220816130553.31406-5-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Yin Fengwei <fengwei.yin@intel.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm-hugetlb-fix-missing-call-to-restore_reserve_on_error-v2

remove err == -EEXIST check and retry logic

Link: https://lkml.kernel.org/r/20220823030209.57434-4-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: fix missing call to restore_reserve_on_error()

When huge_add_to_page_cache() fails, the page is freed directly without
calling restore_reserve_on_error() to restore reserve for newly allocated
pages not in page cache.  Fix this by calling restore_reserve_on_error()
when huge_add_to_page_cache fails.

Link: https://lkml.kernel.org/r/20220816130553.31406-4-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: fix WARN_ON(!kobj) in sysfs_create_group()

If sysfs_create_group() fails with hstate_attr_group, hstate_kobjs[hi]
will be set to NULL.  Then it will be passed to sysfs_create_group() if
h->demote_order != 0 thus triggering WARN_ON(!kobj) check.  Fix this by
making sure hstate_kobjs[hi] != NULL when calling sysfs_create_group.

Link: https://lkml.kernel.org/r/20220816130553.31406-3-linmiaohe@huawei.com
Fixes: 79dfc695525f ("hugetlb: add demote hugetlb page sysfs interfaces")
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: fix incorrect update of max_huge_pages

Patch series "A few fixup patches for hugetlb".

This series contains a few fixup patches to fix incorrect update of
max_huge_pages, fix WARN_ON(!kobj) in sysfs_create_group() and so on.
More details can be found in the respective changelogs.

This patch (of 6):

There should be pages_per_huge_page(h) /
pages_per_huge_page(target_hstate) pages incremented for
target_hstate->max_huge_pages when page is demoted.  Update max_huge_pages
accordingly for consistency.

Link: https://lkml.kernel.org/r/20220816130553.31406-1-linmiaohe@huawei.com
Link: https://lkml.kernel.org/r/20220816130553.31406-2-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/compaction: fix set skip in fast_find_migrateblock

When we successfully find a pageblock in fast_find_migrateblock(), the
block will be set skip-flag through set_pageblock_skip().  However, when
entering isolate_migratepages_block(), the whole pageblock will be skipped
due to the branch 'if (!valid_page && IS_ALIGNED(low_pfn,
pageblock_nr_pages))'.  Eventually we will goto isolate_abort and isolate
nothing.  That cause fast_find_migrateblock useless.

In this patch, when we find a suitable pageblock in fast_find_
migrateblock, we do noting but let isolate_migratepages_block to set skip
flag to the pageblock after scan it.  Normally, we would isolate some
pages from the fast-find block.

I use mmtest/thpscale-madvhugepage test it. Here is the result:
                            baseline               patch
Amean     fault-both-1      1331.66 (   0.00%)     1261.04 *   5.30%*
Amean     fault-both-3      1383.95 (   0.00%)     1191.69 *  13.89%*
Amean     fault-both-5      1568.13 (   0.00%)     1445.20 *   7.84%*
Amean     fault-both-7      1819.62 (   0.00%)     1555.13 *  14.54%*
Amean     fault-both-12     1106.96 (   0.00%)     1149.43 *  -3.84%*
Amean     fault-both-18     2196.93 (   0.00%)     1875.77 *  14.62%*
Amean     fault-both-24     2642.69 (   0.00%)     2671.21 *  -1.08%*
Amean     fault-both-30     2901.89 (   0.00%)     2857.32 *   1.54%*
Amean     fault-both-32     3747.00 (   0.00%)     3479.23 *   7.15%*

Link: https://lkml.kernel.org/r/20220713062009.597255-1-zhouchuyi@bytedance.com
Fixes: 70b44595eafe9 ("mm, compaction: use free lists to quickly locate a migration source")
Signed-off-by: zhouchuyi <zhouchuyi@bytedance.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory tiering: adjust hot threshold automatically

The promotion hot threshold is workload and system configuration
dependent.  So in this patch, a method to adjust the hot threshold
automatically is implemented.  The basic idea is to control the number of
the candidate promotion pages to match the promotion rate limit.  If the
hint page fault latency of a page is less than the hot threshold, we will
try to promote the page, and the page is called the candidate promotion
page.

If the number of the candidate promotion pages in the statistics interval
is much more than the promotion rate limit, the hot threshold will be
decreased to reduce the number of the candidate promotion pages.
Otherwise, the hot threshold will be increased to increase the number of
the candidate promotion pages.

To make the above method works, in each statistics interval, the total
number of the pages to check (on which the hint page faults occur) and the
hot/cold distribution need to be stable.  Because the page tables are
scanned linearly in NUMA balancing, but the hot/cold distribution isn't
uniform along the address usually, the statistics interval should be
larger than the NUMA balancing scan period.  So in the patch, the max scan
period is used as statistics interval and it works well in our tests.

Link: https://lkml.kernel.org/r/20220713083954.34196-4-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Tested-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: osalvador <osalvador@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wei Xu <weixugc@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zhong Jiang <zhongjiang-ali@linux.alibaba.com>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory tiering: rate limit NUMA migration throughput

In NUMA balancing memory tiering mode, if there are hot pages in slow
memory node and cold pages in fast memory node, we need to promote/demote
hot/cold pages between the fast and cold memory nodes.

A choice is to promote/demote as fast as possible.  But the CPU cycles and
memory bandwidth consumed by the high promoting/demoting throughput will
hurt the latency of some workload because of accessing inflating and slow
memory bandwidth contention.

A way to resolve this issue is to restrict the max promoting/demoting
throughput.  It will take longer to finish the promoting/demoting.  But
the workload latency will be better.  This is implemented in this patch as
the page promotion rate limit mechanism.

The number of the candidate pages to be promoted to the fast memory node
via NUMA balancing is counted, if the count exceeds the limit specified by
the users, the NUMA balancing promotion will be stopped until the next
second.

A new sysctl knob kernel.numa_balancing_promote_rate_limit_MBps is added
for the users to specify the limit.

Link: https://lkml.kernel.org/r/20220713083954.34196-3-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Tested-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: osalvador <osalvador@suse.de>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Rik van Riel <riel@surriel.com>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Wei Xu <weixugc@google.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zhong Jiang <zhongjiang-ali@linux.alibaba.com>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

memory tiering: hot page selection with hint page fault latency

Patch series "memory tiering: hot page selection", v4.

To optimize page placement in a memory tiering system with NUMA balancing,
the hot pages in the slow memory nodes need to be identified.
Essentially, the original NUMA balancing implementation selects the mostly
recently accessed (MRU) pages to promote.  But this isn't a perfect
algorithm to identify the hot pages.  Because the pages with quite low
access frequency may be accessed eventually given the NUMA balancing page
table scanning period could be quite long (e.g.  60 seconds).  So in this
patchset, we implement a new hot page identification algorithm based on
the latency between NUMA balancing page table scanning and hint page
fault.  Which is a kind of mostly frequently accessed (MFU) algorithm.

In NUMA balancing memory tiering mode, if there are hot pages in slow
memory node and cold pages in fast memory node, we need to promote/demote
hot/cold pages between the fast and cold memory nodes.

A choice is to promote/demote as fast as possible.  But the CPU cycles and
memory bandwidth consumed by the high promoting/demoting throughput will
hurt the latency of some workload because of accessing inflating and slow
memory bandwidth contention.

A way to resolve this issue is to restrict the max promoting/demoting
throughput.  It will take longer to finish the promoting/demoting.  But
the workload latency will be better.  This is implemented in this patchset
as the page promotion rate limit mechanism.

The promotion hot threshold is workload and system configuration
dependent.  So in this patchset, a method to adjust the hot threshold
automatically is implemented.  The basic idea is to control the number of
the candidate promotion pages to match the promotion rate limit.

We used the pmbench memory accessing benchmark tested the patchset on a
2-socket server system with DRAM and PMEM installed.  The test results are
as follows,

pmbench score promote rate
 (accesses/s) MB/s
------------- ------------
base   146887704.1        725.6
hot selection     165695601.2        544.0
rate limit   162814569.8        165.2
auto adjustment   170495294.0                  136.9

From the results above,

With hot page selection patch [1/3], the pmbench score increases about
12.8%, and promote rate (overhead) decreases about 25.0%, compared with
base kernel.

With rate limit patch [2/3], pmbench score decreases about 1.7%, and
promote rate decreases about 69.6%, compared with hot page selection
patch.

With threshold auto adjustment patch [3/3], pmbench score increases about
4.7%, and promote rate decrease about 17.1%, compared with rate limit
patch.

Baolin helped to test the patchset with MySQL on a machine which contains
1 DRAM node (30G) and 1 PMEM node (126G).

sysbench /usr/share/sysbench/oltp_read_write.lua \
......
--tables=200 \
--table-size=1000000 \
--report-interval=10 \
--threads=16 \
--time=120

The tps can be improved about 5%.

This patch (of 3):

To optimize page placement in a memory tiering system with NUMA balancing,
the hot pages in the slow memory node need to be identified.  Essentially,
the original NUMA balancing implementation selects the mostly recently
accessed (MRU) pages to promote.  But this isn't a perfect algorithm to
identify the hot pages.  Because the pages with quite low access frequency
may be accessed eventually given the NUMA balancing page table scanning
period could be quite long (e.g.  60 seconds).  The most frequently
accessed (MFU) algorithm is better.

So, in this patch we implemented a better hot page selection algorithm.
Which is based on NUMA balancing page table scanning and hint page fault
as follows,

- When the page tables of the processes are scanned to change PTE/PMD
  to be PROT_NONE, the current time is recorded in struct page as scan
  time.

- When the page is accessed, hint page fault will occur.  The scan
  time is gotten from the struct page.  And The hint page fault
  latency is defined as

    hint page fault time - scan time

The shorter the hint page fault latency of a page is, the higher the
probability of their access frequency to be higher.  So the hint page
fault latency is a better estimation of the page hot/cold.

It's hard to find some extra space in struct page to hold the scan time.
Fortunately, we can reuse some bits used by the original NUMA balancing.

NUMA balancing uses some bits in struct page to store the page accessing
CPU and PID (referring to page_cpupid_xchg_last()).  Which is used by the
multi-stage node selection algorithm to avoid to migrate pages shared
accessed by the NUMA nodes back and forth.  But for pages in the slow
memory node, even if they are shared accessed by multiple NUMA nodes, as
long as the pages are hot, they need to be promoted to the fast memory
node.  So the accessing CPU and PID information are unnecessary for the
slow memory pages.  We can reuse these bits in struct page to record the
scan time.  For the fast memory pages, these bits are used as before.

For the hot threshold, the default value is 1 second, which works well in
our performance test.  All pages with hint page fault latency < hot
threshold will be considered hot.

It's hard for users to determine the hot threshold.  So we don't provide a
kernel ABI to set it, just provide a debugfs interface for advanced users
to experiment.  We will continue to work on a hot threshold automatic
adjustment mechanism.

The downside of the above method is that the response time to the workload
hot spot changing may be much longer.  For example,

- A previous cold memory area becomes hot

- The hint page fault will be triggered.  But the hint page fault
  latency isn't shorter than the hot threshold.  So the pages will
  not be promoted.

- When the memory area is scanned again, maybe after a scan period,
  the hint page fault latency measured will be shorter than the hot
  threshold and the pages will be promoted.

To mitigate this, if there are enough free space in the fast memory node,
the hot threshold will not be used, all pages will be promoted upon the
hint page fault for fast response.

Thanks Zhong Jiang reported and tested the fix for a bug when disabling
memory tiering mode dynamically.

Link: https://lkml.kernel.org/r/20220713083954.34196-1-ying.huang@intel.com
Link: https://lkml.kernel.org/r/20220713083954.34196-2-ying.huang@intel.com
Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Tested-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Rik van Riel <riel@surriel.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Wei Xu <weixugc@google.com>
Cc: osalvador <osalvador@suse.de>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Zhong Jiang <zhongjiang-ali@linux.alibaba.com>
Cc: Oscar Salvador <osalvador@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/util.c: add warning if __vm_enough_memory fails

If a process has not enough memory to allocate a new virtual mapping, we
may meet verious kinds of error, eg, fork cannot allocate memory, SIGBUS
error in shmem, but it is difficult to confirm them, let's add some debug
information to easily to check this scenario if __vm_enough_memory fails.

Link: https://lkml.kernel.org/r/20220726145428.8030-1-wangkefeng.wang@huawei.com
Reported-by: Yongqiang Liu <liuyongqiang13@huawei.com>
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Acked-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: add more BUILD_BUG_ONs to gfp_migratetype()

gfp_migratetype() also expects GFP_RECLAIMABLE and
GFP_MOVABLE|GFP_RECLAIMABLE to be shiftable into MIGRATE_* enum values, so
add some more BUILD_BUG_ONs to reflect this assumption.

Link: https://linux-review.googlesource.com/id/Iae64e2182f75c3aca776a486b71a72571d66d83e
Link: https://lkml.kernel.org/r/20220726230241.3770532-1-pcc@google.com
Signed-off-by: Peter Collingbourne <pcc@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm-gupc-simplify-and-fix-check_and_migrate_movable_pages-return-codes-fix-fix

fix code layout, per John

Cc: Alistair Popple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Shigeru Yoshida <syoshida@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/gup.c: Fix return value for __gup_longterm_locked()

__get_user_pages_locked() may return the number of pages less than
nr_pages.  So __gup_longterm_locked() have to return the number of pages
__get_user_pages_locked() returns if it succeeded, not nr_pages requested.

Link: https://lkml.kernel.org/r/20220821183547.950370-1-syoshida@redhat.com
Fixes: 61c63c2076d9 (mm/gup.c: simplify and fix check_and_migrate_movable_pages() return codes)
Signed-off-by: Shigeru Yoshida <syoshida@redhat.com>
Reviewed-by: John Hubbard <jhubbard@nvidia.com>
Reported-by: syzbot+616ff0452fec30f4dcfd@syzkaller.appspotmail.com
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/gup.c: simplify and fix check_and_migrate_movable_pages() return codes

When pinning pages with FOLL_LONGTERM check_and_migrate_movable_pages() is
called to migrate pages out of zones which should not contain any longterm
pinned pages.

When migration succeeds all pages will have been unpinned so pinning needs
to be retried.  This is indicated by returning zero.  When all pages are
in the correct zone the number of pinned pages is returned.

However migration can also fail, in which case pages are unpinned and
-ENOMEM is returned.  However if the failure was due to not being unable
to isolate a page zero is returned.  This leads to indefinite looping in
__gup_longterm_locked().

Fix this by simplifying the return codes such that zero indicates all
pages were successfully pinned in the correct zone while errors indicate
either pages were migrated and pinning should be retried or that migration
has failed and therefore the pinning operation should fail.

Link: https://lkml.kernel.org/r/20220729024645.764366-1-apopple@nvidia.com
Signed-off-by: Alistair Popple <apopple@nvidia.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Jason Gunthorpe <jgg@nvidia.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb_cgroup: use helper for_each_hstate and hstate_index

Use helper for_each_hstate and hstate_index to iterate the hstate and get
the hstate index. Minor readability improvement.

Link: https://lkml.kernel.org/r/20220729080106.12752-6-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb_cgroup: use helper macro NUMA_NO_NODE

It's better to use NUMA_NO_NODE instead of magic number -1. Minor
readability improvement.

Link: https://lkml.kernel.org/r/20220729080106.12752-5-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb_cgroup: remove unneeded return value

The return value of set_hugetlb_cgroup and set_hugetlb_cgroup_rsvd are
always ignored. Remove them to clean up the code.

Link: https://lkml.kernel.org/r/20220729080106.12752-4-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mina Almasry <almasrymina@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb_cgroup: hugetlbfs: use helper macro SZ_1{K,M,G}

Use helper macro SZ_1K, SZ_1M and SZ_1G to do the size conversion. Minor
readability improvement.

Link: https://lkml.kernel.org/r/20220729080106.12752-3-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

hugetlb_cgroup: remove unneeded nr_pages > 0 check

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

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

This patch (of 5):

When code reaches here, nr_pages must be > 0. Remove unneeded nr_pages > 0
check to simplify the code.

Link: https://lkml.kernel.org/r/20220729080106.12752-1-linmiaohe@huawei.com
Link: https://lkml.kernel.org/r/20220729080106.12752-2-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: Mina Almasry <almasrymina@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

Kselftests: remove support of libhugetlbfs from kselftests

libhugetlbfs, the user side utitlity to work with hugepages, does not have
any active support.  There are only 2 selftests which are part of in
vm/hmm_test.c that depends on libhugetlbfs.

This patch modifies the tests so that they will not require libhugetlb
library.

Link: https://lkml.kernel.org/r/20220801070231.13831-1-tsahu@linux.ibm.com
Signed-off-by: Tarun Sahu <tsahu@linux.ibm.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com>
Cc: Jerome Glisse <jglisse@redhat.com>
Cc: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Vaibhav Jain <vaibhav@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

kfence: add sysfs interface to disable kfence for selected slabs.

By default kfence allocation can happen for any slab object, whose size is
up to PAGE_SIZE, as long as that allocation is the first allocation after
expiration of kfence sample interval.  But in certain debugging scenarios
we may be interested in debugging corruptions involving some specific slub
objects like dentry or ext4_* etc.  In such cases limiting kfence for
allocations involving only specific slub objects will increase the
probablity of catching the issue since kfence pool will not be consumed by
other slab objects.

This patch introduces a sysfs interface
'/sys/kernel/slab/<name>/skip_kfence' to disable kfence for specific
slabs.  Having the interface work in this way does not impact
current/default behavior of kfence and allows us to use kfence for
specific slabs (when needed) as well.  The decision to skip/use kfence is
taken depending on whether kmem_cache.flags has (newly introduced)
SLAB_SKIP_KFENCE flag set or not.

Link: https://lkml.kernel.org/r/20220814195353.2540848-1-imran.f.khan@oracle.com
Signed-off-by: Imran Khan <imran.f.khan@oracle.com>
Reviewed-by: Vlastimil Babka <vbabka@suse.cz>
Reviewed-by: Marco Elver <elver@google.com>
Reviewed-by: Hyeonggon Yoo <42.hyeyoo@gmail.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: migration: fix the FOLL_GET failure on following huge page

Not all huge page APIs support FOLL_GET option, so move_pages() syscall
will fail to get the page node information for some huge pages.

Like x86 on linux 5.19 with 1GB huge page API follow_huge_pud(), it will
return NULL page for FOLL_GET when calling move_pages() syscall with the
NULL 'nodes' parameter, the 'status' parameter has '-2' error in array.

Note: follow_huge_pud() now supports FOLL_GET in linux 6.0.
Link: https://lore.kernel.org/all/20220714042420.1847125-3-naoya.horiguchi@linux.dev
But these huge page APIs don't support FOLL_GET:
  1. follow_huge_pud() in arch/s390/mm/hugetlbpage.c
  2. follow_huge_addr() in arch/ia64/mm/hugetlbpage.c
     It will cause WARN_ON_ONCE for FOLL_GET.
  3. follow_huge_pgd() in mm/hugetlb.c

This is an temporary solution to mitigate the side effect of the race
condition fix by calling follow_page() with FOLL_GET set for huge pages.

After supporting follow huge page by FOLL_GET is done, this fix can be
reverted safely.

Link: https://lkml.kernel.org/r/20220823135841.934465-2-haiyue.wang@intel.com
Link: https://lkml.kernel.org/r/20220812084921.409142-1-haiyue.wang@intel.com
Fixes: 4cd614841c06 ("mm: migration: fix possible do_pages_stat_array racing with memory offline")
Signed-off-by: Haiyue Wang <haiyue.wang@intel.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Reviewed-by: Baolin Wang <baolin.wang@linux.alibaba.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/vmscan: make the annotations of refaults code at the right place

After patch "mm/workingset: prepare the workingset detection
infrastructure for anon LRU", we can handle the refaults of anonymous
pages too.  So the annotations of refaults should cover both of anonymous
pages and file pages.

Link: https://lkml.kernel.org/r/20220813080757.59131-1-yang.yang29@zte.com.cn
Fixes: 170b04b7ae4963 ("mm/workingset: prepare the workingset detection infrastructure for anon LRU")
Signed-off-by: Yang Yang <yang.yang29@zte.com.cn>
Signed-off-by: CGEL ZTE <cgel.zte@gmail.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/core: simplify the parameter passing for region split operation

The parameter 'struct damon_ctx *ctx' is unnecessary in damon region split
operation, so we can remove it.

Link: https://lkml.kernel.org/r/1660403943-29124-1-git-send-email-kaixuxia@tencent.com
Signed-off-by: Kaixu Xia <kaixuxia@tencent.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

tools/vm/page_owner_sort: fix -f option

The -f option is to filter out the information of blocks whose memory has
not been released, I noticed some blocks should not be filtered out.

Commit 9cc7e96aa846 ("mm/page_owner: record timestamp and pid") records
the allocation timestamp (ts_nsec) of all pages.

Commit 866b48526217 ("mm/page_owner: record the timestamp of all pages
during free") records the free timestamp (free_ts_nsec) of all pages.
When the page is allocated for the first time, the initial value of
free_ts_nsec is 0, and the corresponding time will be obtained when the
page is released.  But during reallocation the free_ts_nsec will not reset
to 0 again.  In particular, when page migration occurs, these two
timestamps will be the same.

Now page_owner_sort removes all text blocks whose free_ts_nsec is not 0
when using -f option.  However, this way can only select pages allocated
for the first time.  If a freed page is reallocated, free_ts_nsec will be
less than ts_nsec; if page migration occurs, the two timestamps will be
equal.  These cases should be considered as pages are not released.

So I fix the function is_need() to keep text blocks that meet the above
two conditions when using -f option.

Link: https://lkml.kernel.org/r/20220812155515.30846-1-caoyixuan2019@email.szu.edu.cn
Signed-off-by: Yixuan Cao <caoyixuan2019@email.szu.edu.cn>
Cc: Chongxi Zhao <zhaochongxi2019@email.szu.edu.cn>
Cc: Jiajian Ye <yejiajian2018@email.szu.edu.cn>
Cc: Yuhong Feng <yuhongf@szu.edu.cn>
Cc: Liam Mark <lmark@codeaurora.org>
Cc: Georgi Djakov <georgi.djakov@linaro.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/util: reduce stack usage of folio_mapcount

folio_test_hugetlb() will call PageHeadHuge which is a function call,
and blocks the compiler from recognizing this redundant load.

After rearranging the code, stack usage is dropped from 32 to 24, and
the function size is smaller (tested on GCC 12):

Before:
Stack usage:
mm/util.c:845:5:folio_mapcount  32      static
Size:
0000000000000ea0 00000000000000c7 T folio_mapcount

After:
Stack usage:
mm/util.c:845:5:folio_mapcount  24      static
Size:
0000000000000ea0 00000000000000b0 T folio_mapcount

Link: https://lkml.kernel.org/r/20220801173155.92008-1-ryncsn@gmail.com
Signed-off-by: Kairui Song <kasong@tencent.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: only search higher order when fallback

It seems unnecessary to search pages with order < alloc_order in
fallback allocation.

This can currently happen with ALLOC_NOFRAGMENT and alloc_order >
pageblock_order, so add a test to prevent it.

[vbabka@suse.cz: changelog addition]
Link: https://lkml.kernel.org/r/20220803025121.47018-1-wuyun.abel@bytedance.com
Signed-off-by: Abel Wu <wuyun.abel@bytedance.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

page_alloc: remove inactive initialization

The allocation address of the table pointer variable is first performed in
the function, no initialization assignment is required, and no invalid
pointer will appear.

Link: https://lkml.kernel.org/r/20220803064118.3664-1-kunyu@nfschina.com
Signed-off-by: Li kunyu <kunyu@nfschina.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/hugetlb: add dedicated func to get 'allowed' nodemask for current process

Muchun Song found that after MPOL_PREFERRED_MANY policy was introduced in
commit b27abaccf8e8 ("mm/mempolicy: add MPOL_PREFERRED_MANY for multiple
preferred nodes"), the policy_nodemask_current()'s semantics for this new
policy has been changed, which returns 'preferred' nodes instead of
'allowed' nodes.

With the changed semantic of policy_nodemask_current, a task with
MPOL_PREFERRED_MANY policy could fail to get its reservation even though
it can fall back to other nodes (either defined by cpusets or all online
nodes) for that reservation failing mmap calles unnecessarily early.

The fix is to not consider MPOL_PREFERRED_MANY for reservations at all
because they, unlike MPOL_MBIND, do not pose any actual hard constrain.

Michal suggested the policy_nodemask_current() is only used by hugetlb,
and could be moved to hugetlb code with more explicit name to enforce the
'allowed' semantics for which only MPOL_BIND policy matters.

apply_policy_zone() is made extern to be called in hugetlb code and its
return value is changed to bool.

[1]. https://lore.kernel.org/lkml/20220801084207.39086-1-songmuchun@bytedance.com/t/

Link: https://lkml.kernel.org/r/20220805005903.95563-1-feng.tang@intel.com
Fixes: b27abaccf8e8 ("mm/mempolicy: add MPOL_PREFERRED_MANY for multiple preferred nodes")
Signed-off-by: Feng Tang <feng.tang@intel.com>
Reported-by: Muchun Song <songmuchun@bytedance.com>
Suggested-by: Michal Hocko <mhocko@suse.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Cc: Ben Widawsky <bwidawsk@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

filemap: make the accounting of thrashing more consistent

Once upon a time, we only support accounting thrashing of page cache.
Then Joonsoo introduced workingset detection for anonymous pages and we
gained the ability to account thrashing of them[1].

So let delayacct account both the thrashing of page cache and anonymous
pages, this could make the codes more consistent and simpler.

[1] commit aae466b0052e ("mm/swap: implement workingset detection for anonymous LRU")

Link: https://lkml.kernel.org/r/20220805033838.1714674-1-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: Balbir Singh <bsingharora@gmail.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Yang Yang <yang.yang29@zte.com.cn>
Cc: David Hildenbrand <david@redhat.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/mempolicy: fix lock contention on mems_allowed

The mems_allowed field can be modified by other tasks, so it isn't safe to
access it with alloc_lock unlocked even in the current process context.

Say there are two tasks: A from cpusetA is performing set_mempolicy(2),
and B is changing cpusetA's cpuset.mems:

  A (set_mempolicy) B (echo xx > cpuset.mems)
  -------------------------------------------------------
  pol = mpol_new();
update_tasks_nodemask(cpusetA) {
  foreach t in cpusetA {
    cpuset_change_task_nodemask(t) {
  mpol_set_nodemask(pol) {
      task_lock(t); // t could be A
    new = f(A->mems_allowed);
      update t->mems_allowed;
    pol.create(pol, new);
      task_unlock(t);
  }
    }
  }
}
  task_lock(A);
  A->mempolicy = pol;
  task_unlock(A);

In this case A's pol->nodes is computed by old mems_allowed, and could
be inconsistent with A's new mems_allowed.

While it is different when replacing vmas' policy: the pol->nodes is
gone wild only when current_cpuset_is_being_rebound():

  A (mbind) B (echo xx > cpuset.mems)
  -------------------------------------------------------
  pol = mpol_new();
  mmap_write_lock(A->mm);
cpuset_being_rebound = cpusetA;
update_tasks_nodemask(cpusetA) {
  foreach t in cpusetA {
    cpuset_change_task_nodemask(t) {
  mpol_set_nodemask(pol) {
      task_lock(t); // t could be A
    mask = f(A->mems_allowed);
      update t->mems_allowed;
    pol.create(pol, mask);
      task_unlock(t);
  }
    }
  foreach v in A->mm {
    if (cpuset_being_rebound == cpusetA)
      pol.rebind(pol, cpuset.mems);
    v->vma_policy = pol;
  }
  mmap_write_unlock(A->mm);
    mmap_write_lock(t->mm);
    mpol_rebind_mm(t->mm);
    mmap_write_unlock(t->mm);
  }
}
cpuset_being_rebound = NULL;

In this case, the cpuset.mems, which has already done updating, is finally
used for calculating pol->nodes, rather than A->mems_allowed.  So it is OK
to call mpol_set_nodemask() with alloc_lock unlocked when doing mbind(2).

Link: https://lkml.kernel.org/r/20220811124157.74888-1-wuyun.abel@bytedance.com
Fixes: 78b132e9bae9 ("mm/mempolicy: remove or narrow the lock on current")
Signed-off-by: Abel Wu <wuyun.abel@bytedance.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Wei Yang <richard.weiyang@gmail.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/cma_debug: show complete cma name in debugfs directories

Currently only 12 characters of the cma name is being used as the debug
directories where as the cma name can be of length CMA_MAX_NAME(=64)
characters.  One side problem with this is having 2 cma's with first
common 12 characters would end up in trying to create directories with
same name and fails with -EEXIST thus can limit cma debug functionality.

The 'cma-' prefix is used initially where cma areas don't have any names
and are represented by simple integer values.  Since now each cma would be
having its own name, drop 'cma-' prefix for the cma debug directories as
they are clearly evident that they are for cma debug through creating them
in /sys/kernel/debug/cma/ path.

Link: https://lkml.kernel.org/r/1660223729-22461-1-git-send-email-quic_charante@quicinc.com
Signed-off-by: Charan Teja Kalla <quic_charante@quicinc.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Pavan Kondeti <quic_pkondeti@quicinc.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: remove the end_write_func argument to __swap_writepage

The argument is always set to end_swap_bio_write, so remove the argument
and mark end_swap_bio_write static.

Link: https://lkml.kernel.org/r/20220811141741.660214-1-hch@lst.de
Signed-off-by: Christoph Hellwig <hch@lst.de>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Dan Streetman <ddstreet@ieee.org>
Cc: Vitaly Wool <vitaly.wool@konsulko.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

zsmalloc: remove unnecessary size_class NULL check

pool->size_class array elements can't be NULL, so this check
is not needed.

In the whole code, we assign pool->size_class[i] values that are
not NULL. Releasing memory for these values occurs in the
zs_destroy_pool() function, which also releases and destroys the pool.

In addition, in the zs_stats_size_show() and async_free_zspage(),
with similar iterations over the array, we don't check it for NULL
pointer.

Link: https://lkml.kernel.org/r/20220811153755.16102-3-avromanov@sberdevices.ru
Signed-off-by: Alexey Romanov <avromanov@sberdevices.ru>
Reviewed-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>

zsmalloc-zs_object_copy-add-clarifying-comment-fix

tweak comment layout

Cc: Alexey Romanov <avromanov@sberdevices.ru>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Cc: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

zsmalloc: zs_object_copy: add clarifying comment

Patch series "tidy up zsmalloc implementation"

This patchset remove some unnecessary checks and adds a clarifying
comment.  While analysing zs_object_copy() function code, I spent some
time to understand what the call kunmap_atomic(d_addr) is for.  It seems
that this point is not trivial and it is worth adding a comment.

This patch (of 2):

It's not obvious why kunmap_atomic(d_addr) call is needed.

Link: https://lkml.kernel.org/r/20220811153755.16102-1-avromanov@sberdevices.ru
Link: https://lkml.kernel.org/r/20220811153755.16102-2-avromanov@sberdevices.ru
Signed-off-by: Alexey Romanov <avromanov@sberdevices.ru>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: cache swap migration A/D bits support

Introduce a variable swap_migration_ad_supported to cache whether the arch
supports swap migration A/D bits.

Here one thing to mention is that SWP_MIG_TOTAL_BITS will internally
reference the other macro MAX_PHYSMEM_BITS, which is a function call on
x86 (constant on all the rest of archs).

It's safe to reference it in swapfile_init() because when reaching here
we're already during initcalls level 4 so we must have initialized 5-level
pgtable for x86_64 (right after early_identify_cpu() finishes).

- start_kernel
  - setup_arch
    - early_cpu_init
      - get_cpu_cap --> fetch from CPUID (including X86_FEATURE_LA57)
      - early_identify_cpu --> clear X86_FEATURE_LA57 (if early lvl5 not enabled (USE_EARLY_PGTABLE_L5))
  - arch_call_rest_init
    - rest_init
      - kernel_init
        - kernel_init_freeable
          - do_basic_setup
            - do_initcalls --> calls swapfile_init() (initcall level 4)

This should slightly speed up the migration swap entry handlings.

Link: https://lkml.kernel.org/r/20220811161331.37055-8-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: cache maximum swapfile size when init swap

We used to have swapfile_maximum_size() fetching a maximum value of
swapfile size per-arch.

As the caller of max_swapfile_size() grows, this patch introduce a
variable "swapfile_maximum_size" and cache the value of old
max_swapfile_size(), so that we don't need to calculate the value every
time.

Caching the value in swapfile_init() is safe because when reaching the
phase we should have initialized all the relevant information.  Here the
major arch to take care of is x86, which defines the max swapfile size
based on L1TF mitigation.

Here both X86_BUG_L1TF or l1tf_mitigation should have been setup properly
when reaching swapfile_init().  As a reference, the code path looks like
this for x86:

- start_kernel
  - setup_arch
    - early_cpu_init
      - early_identify_cpu --> setup X86_BUG_L1TF
  - parse_early_param
    - l1tf_cmdline --> set l1tf_mitigation
  - check_bugs
    - l1tf_select_mitigation --> set l1tf_mitigation
  - arch_call_rest_init
    - rest_init
      - kernel_init
        - kernel_init_freeable
          - do_basic_setup
            - do_initcalls --> calls swapfile_init() (initcall level 4)

The swapfile size only depends on swp pte format on non-x86 archs, so
caching it is safe too.

Since at it, rename max_swapfile_size() to arch_max_swapfile_size()
because arch can define its own function, so it's more straightforward to
have "arch_" as its prefix.  At the meantime, export swapfile_maximum_size
to replace the old usages of max_swapfile_size().

Link: https://lkml.kernel.org/r/20220811161331.37055-7-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: remember young/dirty bit for page migrations

When page migration happens, we always ignore the young/dirty bit settings
in the old pgtable, and marking the page as old in the new page table
using either pte_mkold() or pmd_mkold(), and keeping the pte clean.

That's fine from functional-wise, but that's not friendly to page reclaim
because the moving page can be actively accessed within the procedure.
Not to mention hardware setting the young bit can bring quite some
overhead on some systems, e.g.  x86_64 needs a few hundreds nanoseconds to
set the bit.  The same slowdown problem to dirty bits when the memory is
first written after page migration happened.

Actually we can easily remember the A/D bit configuration and recover the
information after the page is migrated.  To achieve it, define a new set
of bits in the migration swap offset field to cache the A/D bits for old
pte.  Then when removing/recovering the migration entry, we can recover
the A/D bits even if the page changed.

One thing to mention is that here we used max_swapfile_size() to detect
how many swp offset bits we have, and we'll only enable this feature if we
know the swp offset is big enough to store both the PFN value and the A/D
bits.  Otherwise the A/D bits are dropped like before.

Link: https://lkml.kernel.org/r/20220811161331.37055-6-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Huang Ying <ying.huang@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/thp: carry over dirty bit when thp splits on pmd

Carry over the dirty bit from pmd to pte when a huge pmd splits.  It
shouldn't be a correctness issue since when pmd_dirty() we'll have the
page marked dirty anyway, however having dirty bit carried over helps the
next initial writes of split ptes on some archs like x86.

Link: https://lkml.kernel.org/r/20220811161331.37055-5-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: Huang Ying <ying.huang@intel.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: add swp_offset_pfn() to fetch PFN from swap entry

We've got a bunch of special swap entries that stores PFN inside the swap
offset fields.  To fetch the PFN, normally the user just calls
swp_offset() assuming that'll be the PFN.

Add a helper swp_offset_pfn() to fetch the PFN instead, fetching only the
max possible length of a PFN on the host, meanwhile doing proper check
with MAX_PHYSMEM_BITS to make sure the swap offsets can actually store the
PFNs properly always using the BUILD_BUG_ON() in is_pfn_swap_entry().

One reason to do so is we never tried to sanitize whether swap offset can
really fit for storing PFN.  At the meantime, this patch also prepares us
with the future possibility to store more information inside the swp
offset field, so assuming "swp_offset(entry)" to be the PFN will not stand
any more very soon.

Replace many of the swp_offset() callers to use swp_offset_pfn() where
proper.  Note that many of the existing users are not candidates for the
replacement, e.g.:

  (1) When the swap entry is not a pfn swap entry at all, or,
  (2) when we wanna keep the whole swp_offset but only change the swp type.

For the latter, it can happen when fork() triggered on a write-migration
swap entry pte, we may want to only change the migration type from
write->read but keep the rest, so it's not "fetching PFN" but "changing
swap type only".  They're left aside so that when there're more
information within the swp offset they'll be carried over naturally in
those cases.

Since at it, dropping hwpoison_entry_to_pfn() because that's exactly what
the new swp_offset_pfn() is about.

Link: https://lkml.kernel.org/r/20220811161331.37055-4-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: "Huang, Ying" <ying.huang@intel.com>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/swap: comment all the ifdef in swapops.h

swapops.h contains quite a few layers of ifdef, some of the "else" and
"endif" doesn't get proper comment on the macro so it's hard to follow on
what are they referring to.  Add the comments.

Link: https://lkml.kernel.org/r/20220811161331.37055-3-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Suggested-by: Nadav Amit <nadav.amit@gmail.com>
Reviewed-by: Huang Ying <ying.huang@intel.com>
Reviewed-by: Alistair Popple <apopple@nvidia.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/x86: use SWP_TYPE_BITS in 3-level swap macros

Patch series "mm: Remember a/d bits for migration entries", v4.

Problem
=======

When migrating a page, right now we always mark the migrated page as old &
clean.

However that could lead to at least two problems:

  (1) We lost the real hot/cold information while we could have persisted.
      That information shouldn't change even if the backing page is changed
      after the migration,

  (2) There can be always extra overhead on the immediate next access to
      any migrated page, because hardware MMU needs cycles to set the young
      bit again for reads, and dirty bits for write, as long as the
      hardware MMU supports these bits.

Many of the recent upstream works showed that (2) is not something trivial
and actually very measurable.  In my test case, reading 1G chunk of memory
- jumping in page size intervals - could take 99ms just because of the
extra setting on the young bit on a generic x86_64 system, comparing to
4ms if young set.

This issue is originally reported by Andrea Arcangeli.

Solution
========

To solve this problem, this patchset tries to remember the young/dirty
bits in the migration entries and carry them over when recovering the
ptes.

We have the chance to do so because in many systems the swap offset is not
really fully used.  Migration entries use swp offset to store PFN only,
while the PFN is normally not as large as swp offset and normally smaller.
It means we do have some free bits in swp offset that we can use to store
things like A/D bits, and that's how this series tried to approach this
problem.

max_swapfile_size() is used here to detect per-arch offset length in swp
entries.  We'll automatically remember the A/D bits when we find that we
have enough swp offset field to keep both the PFN and the extra bits.

Since max_swapfile_size() can be slow, the last two patches cache the
results for it and also swap_migration_ad_supported as a whole.

Known Issues / TODOs
====================

We still haven't taught madvise() to recognize the new A/D bits in
migration entries, namely MADV_COLD/MADV_FREE.  E.g.  when MADV_COLD upon
a migration entry.  It's not clear yet on whether we should clear the A
bit, or we should just drop the entry directly.

We didn't teach idle page tracking on the new migration entries, because
it'll need larger rework on the tree on rmap pgtable walk.  However it
should make it already better because before this patchset page will be
old page after migration, so the series will fix potential false negative
of idle page tracking when pages were migrated before observing.

The other thing is migration A/D bits will not start to working for
private device swap entries.  The code is there for completeness but since
private device swap entries do not yet have fields to store A/D bits, even
if we'll persistent A/D across present pte switching to migration entry,
we'll lose it again when the migration entry converted to private device
swap entry.

Tests
=====

After the patchset applied, the immediate read access test [1] of above 1G
chunk after migration can shrink from 99ms to 4ms.  The test is done by
moving 1G pages from node 0->1->0 then read it in page size jumps.  The
test is with Intel(R) Xeon(R) CPU E5-2630 v4 @ 2.20GHz.

Similar effect can also be measured when writting the memory the 1st time
after migration.

After applying the patchset, both initial immediate read/write after page
migrated will perform similarly like before migration happened.

Patch Layout
============

Patch 1-2:  Cleanups from either previous versions or on swapops.h macros.

Patch 3-4:  Prepare for the introduction of migration A/D bits

Patch 5:    The core patch to remember young/dirty bit in swap offsets.

Patch 6-7:  Cache relevant fields to make migration_entry_supports_ad() fast.

[1] https://github.com/xzpeter/clibs/blob/master/misc/swap-young.c

This patch (of 7):

Replace all the magic "5" with the macro.

Link: https://lkml.kernel.org/r/20220811161331.37055-1-peterx@redhat.com
Link: https://lkml.kernel.org/r/20220811161331.37055-2-peterx@redhat.com
Signed-off-by: Peter Xu <peterx@redhat.com>
Reviewed-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Huang Ying <ying.huang@intel.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A . Shutemov" <kirill@shutemov.name>
Cc: Alistair Popple <apopple@nvidia.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Andi Kleen <andi.kleen@intel.com>
Cc: Nadav Amit <nadav.amit@gmail.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/vmscan: define macros for refaults in struct lruvec

The magic number 0 and 1 are used in several places in vmscan.c.
Define macros for them to improve code readability.

Link: https://lkml.kernel.org/r/20220808005644.1721066-1-yang.yang29@zte.com.cn
Signed-off-by: Yang Yang <yang.yang29@zte.com.cn>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests: vm: add /dev/userfaultfd test cases to run_vmtests.sh

This new mode was recently added to the userfaultfd selftest. We want to
exercise both userfaultfd(2) as well as /dev/userfaultfd, so add both
test cases to the script.

Link: https://lkml.kernel.org/r/20220808175614.3885028-6-axelrasmussen@google.com
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Acked-by: Peter Xu <peterx@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry V. Levin <ldv@altlinux.org>
Cc: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhang Yi <yi.zhang@huawei.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

userfaultfd-update-documentation-to-describe-dev-userfaultfd-v7

improve wording in two spots in the documentation, per Mike

Link: https://lkml.kernel.org/r/20220819205201.658693-5-axelrasmussen@google.com
Acked-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry V. Levin <ldv@altlinux.org>
Cc: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Cc: Nadav Amit <namit@vmware.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhang Yi <yi.zhang@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

userfaultfd: update documentation to describe /dev/userfaultfd

Explain the different ways to create a new userfaultfd, and how access
control works for each way.

Link: https://lkml.kernel.org/r/20220808175614.3885028-5-axelrasmussen@google.com
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Acked-by: Peter Xu <peterx@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry V. Levin <ldv@altlinux.org>
Cc: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhang Yi <yi.zhang@huawei.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

userfaultfd: selftests: modify selftest to use /dev/userfaultfd

modify selftest to exit with KSFT_SKIP *only* when features are
unsupported, exiting with 1 in other error cases, per Mike

Link: https://lkml.kernel.org/r/20220819205201.658693-4-axelrasmussen@google.com
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Peter Xu <peterx@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry V. Levin <ldv@altlinux.org>
Cc: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhang Yi <yi.zhang@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

userfaultfd: selftests: modify selftest to use /dev/userfaultfd

We clearly want to ensure both userfaultfd(2) and /dev/userfaultfd keep
working into the future, so just run the test twice, using each interface.

Instead of always testing both userfaultfd(2) and /dev/userfaultfd, let
the user choose which to test.

As with other test features, change the behavior based on a new command
line flag.  Introduce the idea of "test mods", which are generic (not
specific to a test type) modifications to the behavior of the test.  This
is sort of borrowed from this RFC patch series [1], but simplified a bit.

The benefit is, in "typical" configurations this test is somewhat slow
(say, 30sec or something).  Testing both clearly doubles it, so it may not
always be desirable, as users are likely to use one or the other, but
never both, in the "real world".

[1]: https://patchwork.kernel.org/project/linux-mm/patch/20201129004548.1619714-14-namit@vmware.com/

Link: https://lkml.kernel.org/r/20220808175614.3885028-4-axelrasmussen@google.com
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Acked-by: Peter Xu <peterx@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry V. Levin <ldv@altlinux.org>
Cc: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhang Yi <yi.zhang@huawei.com>
Cc: Mike Rapoport <rppt@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

userfaultfd-add-dev-userfaultfd-for-fine-grained-access-control-v7

Handle misc_register() failure properly by propagating the error instead
of just WARN_ON-ing.

Link: https://lkml.kernel.org/r/20220819205201.658693-3-axelrasmussen@google.com
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Acked-by: Nadav Amit <namit@vmware.com>
Acked-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry V. Levin <ldv@altlinux.org>
Cc: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhang Yi <yi.zhang@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

userfaultfd: add /dev/userfaultfd for fine grained access control

Historically, it has been shown that intercepting kernel faults with
userfaultfd (thereby forcing the kernel to wait for an arbitrary amount of
time) can be exploited, or at least can make some kinds of exploits
easier.  So, in 37cd0575b8 "userfaultfd: add UFFD_USER_MODE_ONLY" we
changed things so, in order for kernel faults to be handled by
userfaultfd, either the process needs CAP_SYS_PTRACE, or this sysctl must
be configured so that any unprivileged user can do it.

In a typical implementation of a hypervisor with live migration (take
QEMU/KVM as one such example), we do indeed need to be able to handle
kernel faults.  But, both options above are less than ideal:

- Toggling the sysctl increases attack surface by allowing any
  unprivileged user to do it.

- Granting the live migration process CAP_SYS_PTRACE gives it this
  ability, but *also* the ability to "observe and control the
  execution of another process [...], and examine and change [its]
  memory and registers" (from ptrace(2)). This isn't something we need
  or want to be able to do, so granting this permission violates the
  "principle of least privilege".

This is all a long winded way to say: we want a more fine-grained way to
grant access to userfaultfd, without granting other additional permissions
at the same time.

To achieve this, add a /dev/userfaultfd misc device.  This device provides
an alternative to the userfaultfd(2) syscall for the creation of new
userfaultfds.  The idea is, any userfaultfds created this way will be able
to handle kernel faults, without the caller having any special
capabilities.  Access to this mechanism is instead restricted using e.g.
standard filesystem permissions.

Link: https://lkml.kernel.org/r/20220808175614.3885028-3-axelrasmussen@google.com
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Acked-by: Nadav Amit <namit@vmware.com>
Acked-by: Peter Xu <peterx@redhat.com>
Acked-by: Mike Rapoport <rppt@linux.ibm.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry V. Levin <ldv@altlinux.org>
Cc: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Shuah Khan <skhan@linuxfoundation.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhang Yi <yi.zhang@huawei.com>
Cc: Mike Rapoport <rppt@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests: vm: add hugetlb_shared userfaultfd test to run_vmtests.sh

Patch series "userfaultfd: add /dev/userfaultfd for fine grained access
control", v7.

Why not ...?
============

- Why not /proc/[pid]/userfaultfd? Two main points (additional discussion [1]):

    - /proc/[pid]/* files are all owned by the user/group of the process, and
      they don't really support chmod/chown. So, without extending procfs it
      doesn't solve the problem this series is trying to solve.

    - The main argument *for* this was to support creating UFFDs for remote
      processes. But, that use case clearly calls for CAP_SYS_PTRACE, so to
      support this we could just use the UFFD syscall as-is.

- Why not use a syscall? Access to syscalls is generally controlled by
  capabilities. We don't have a capability which is used for userfaultfd access
  without also granting more / other permissions as well, and adding a new
  capability was rejected [2].

    - It's possible a LSM could be used to control access instead, but I have
      some concerns. I don't think this approach would be as easy to use,
      particularly if we were to try to solve this with something heavyweight
      like SELinux. Maybe we could pursue adding a new LSM specifically for
      this user case, but it may be too narrow of a case to justify that.

[1]: https://patchwork.kernel.org/project/linux-mm/cover/20220719195628.3415852-1-axelrasmussen@google.com/
[2]: https://lore.kernel.org/lkml/686276b9-4530-2045-6bd8-170e5943abe4@schaufler-ca.com/T/

This patch (of 5):

This not being included was just a simple oversight.  There are certain
features (like minor fault support) which are only enabled on shared
mappings, so without including hugetlb_shared we actually lose a
significant amount of test coverage.

Link: https://lkml.kernel.org/r/20220808175614.3885028-1-axelrasmussen@google.com
Link: https://lkml.kernel.org/r/20220808175614.3885028-2-axelrasmussen@google.com
Signed-off-by: Axel Rasmussen <axelrasmussen@google.com>
Reviewed-by: Shuah Khan <skhan@linuxfoundation.org>
Reviewed-by: Peter Xu <peterx@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Dave Hansen <dave.hansen@linux.intel.com>
Cc: Dmitry V. Levin <ldv@altlinux.org>
Cc: Gleb Fotengauer-Malinovskiy <glebfm@altlinux.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Nadav Amit <namit@vmware.com>
Cc: Shuah Khan <shuah@kernel.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zhang Yi <yi.zhang@huawei.com>
Cc: Mike Rapoport <rppt@linux.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/damon/dbgfs: use kmalloc for allocating only one element

Use kmalloc(...) rather than kmalloc_array(1, ...) because the number of
elements we are specifying in this case is 1, kmalloc would accomplish the
same thing and we can simplify.

Link: https://lkml.kernel.org/r/20220808220019.1680469-1-klee33@uw.edu
Signed-off-by: Kenneth Lee <klee33@uw.edu>
Reviewed-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/filemap.c: convert page_endio() to use a folio

Replace three calls to compound_head() with one.

Link: https://lkml.kernel.org/r/20220809023256.178194-1-shaoqin.huang@intel.com
Signed-off-by: Shaoqin Huang <shaoqin.huang@intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: memory-failure: cleanup try_to_split_thp_page()

Since commit 5d1fd5dc877b ("mm,hwpoison: introduce MF_MSG_UNSPLIT_THP"),
the action_result(,MF_MSG_UNSPLIT_THP,) called to show memory error event
in memory_failure(), so the pr_info() in try_to_split_thp_page() is only
needed in soft_offline_in_use_page().

Meanwhile this could also fix the unexpected prefix for "thp split failed"
due to commit 96f96763de26 ("mm: memory-failure: convert to pr_fmt()").

Link: https://lkml.kernel.org/r/20220809111813.139690-1-wangkefeng.wang@huawei.com
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Acked-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: align larger anonymous mappings on THP boundaries

Align larger anonymous memory mappings on THP boundaries by going through
thp_get_unmapped_area if THPs are enabled for the current process.

With this patch, larger anonymous mappings are now THP aligned.  When a
malloc library allocates a 2MB or larger arena, that arena can now be
mapped with THPs right from the start, which can result in better TLB hit
rates and execution time.

Link: https://lkml.kernel.org/r/20220809142457.4751229f@imladris.surriel.com
Signed-off-by: Rik van Riel <riel@surriel.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_ext: remove unused variable in offline_page_ext

Remove unused variable 'nid' in offline_page_ext().  This is not used
since the page_ext code inception.

Link: https://lkml.kernel.org/r/1659330397-11817-1-git-send-email-quic_charante@quicinc.com
Signed-off-by: Charan Teja Kalla <quic_charante@quicinc.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Anshuman Khandual <anshuman.khandual@arm.com>
Cc: Roman Gushchin <roman.gushchin@linux.dev>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Muchun Song <songmuchun@bytedance.com>
Cc: Pavan Kondeti <quic_pkondeti@quicinc.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: prevent page_frag_alloc() from corrupting the memory

A number of drivers call page_frag_alloc() with a fragment's size >
PAGE_SIZE.

In low memory conditions, __page_frag_cache_refill() may fail the order
3 cache allocation and fall back to order 0; In this case, the cache
will be smaller than the fragment, causing memory corruptions.

Prevent this from happening by checking if the newly allocated cache is
large enough for the fragment; if not, the allocation will fail and
page_frag_alloc() will return NULL.

Link: https://lkml.kernel.org/r/20220715125013.247085-1-mlombard@redhat.com
Signed-off-by: Maurizio Lombardi <mlombard@redhat.com>
Reviewed-by: Alexander Duyck <alexanderduyck@fb.com>
Cc: Chen Lin <chen45464546@163.com>
Cc: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/vm: add selftest to verify multi THP collapse

Add support to allocate and verify collapse of multiple hugepage-sized
regions into multiple THPs.

Add "nr" argument to check_huge() that instructs check_huge() to check for
exactly "nr_hpages" THPs.  This has the added benefit of now being able to
check for exactly 0 THPs, and so callsites that previously checked the
negation of exactly 1 THP are now more correct.

->collapse struct collapse_context hook has been expanded with a
"nr_hpages" argument to collapse "nr_hpages" hugepages.  The
collapse_full() test has been repurposed to collapse 4 THPs at once.  It
is expected more tests will want to test multi THP collapse (e.g.
file/shmem).

This is of particular benefit to madvise collapse context given that it
may do many THP collapses during a single syscall.

Link: https://lkml.kernel.org/r/20220706235936.2197195-19-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/vm: add selftest to verify recollapse of THPs

Add selftest specific to madvise collapse context that tests MADV_COLLAPSE
is "successful" if a hugepage-aligned/sized region is already pmd-mapped.

This test also verifies that MADV_COLLAPSE can collapse memory into THPs
even in "madvise" THP mode and the memory isn't marked VM_HUGEPAGE.

Link: https://lkml.kernel.org/r/20220706235936.2197195-18-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/vm: add MADV_COLLAPSE collapse context to selftests

Add madvise collapse context to hugepage collapse selftests.  This context
is tested with /sys/kernel/mm/transparent_hugepage/enabled set to "never"
in order to avoid unwanted interaction with khugepaged during testing.

Also, refactor updates to sysfs THP settings using a stack so that the THP
settings from nested callers can be restored.

Link: https://lkml.kernel.org/r/20220706235936.2197195-17-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/vm: dedup hugepage allocation logic

The code

p = alloc_mapping();
printf("Allocate huge page...");
madvise(p, hpage_pmd_size, MADV_HUGEPAGE);
fill_memory(p, 0, hpage_pmd_size);
if (check_huge(p))
success("OK");
else
fail("Fail");

Is repeated many times in different tests.  Add a helper, alloc_hpage()
to handle this.

Link: https://lkml.kernel.org/r/20220706235936.2197195-16-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

selftests/vm: modularize collapse selftests

Modularize the collapse action of khugepaged collapse selftests by
introducing a struct collapse_context which specifies how to collapse a
given memory range and the expected semantics of the collapse.  This can
be reused later to test other collapse contexts.

Additionally, all tests have logic that checks if a collapse occurred via
reading /proc/self/smaps, and report if this is different than expected.
Move this logic into the per-context ->collapse() hook instead of
repeating it in every test.

Link: https://lkml.kernel.org/r/20220706235936.2197195-15-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/madvise: remove CAP_SYS_ADMIN requirement for process_madvise(MADV_COLLAPSE)

process_madvise(MADV_COLLAPSE) currently requires CAP_SYS_ADMIN when not
acting on the caller's own mm.  This is maximally restrictive, and
perpetuates existing issues with CAP_SYS_ADMIN.  Remove this requirement.

When acting on an external process' memory, the biggest concerns for
process_madvise(MADV_COLLAPSE) are (1) being able to influence process
performance by moving memory, possibly between nodes, that is mapped into
the address space of external process(es), (2) defeat of
address-space-layout randomization, and (3), being able to increase
process RSS and memcg usage, possibly causing memcg OOM.

process_madvise(2) already enforces CAP_SYS_NICE and PTRACE_MODE_READ (in
PTRACE_MODE_FSCREDS mode).  A process with these credentials can already
accomplish (1) and (2) via move_pages(MPOL_MF_MOVE_ALL), and (3) via
process_madvise(MADV_WILLNEED).

process_madvise(MADV_COLLAPSE) may also circumvent sysfs THP settings.
When acting on one's own memory (which is equivalent to
madvise(MADV_COLLAPSE)), this is deemed acceptable, since aside from the
possibility of hoarding available hugepages (which is currently already
possible) no harm to the system can be done.  When acting on an external
process' memory, circumventing sysfs THP settings should provide no
additional threat compared to the ones listed.  As such, imposing
additional capabilities (such as CAP_SETUID, as a way to ensure the caller
could have just altered the sysfs THP settings themselves) provides no
extra protection.

Link: https://lkml.kernel.org/r/20220801210946.3069083-1-zokeefe@google.com
Fixes: 7ec952341312 ("mm/madvise: add MADV_COLLAPSE to process_madvise()")
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/madvise: add MADV_COLLAPSE to process_madvise()

Allow MADV_COLLAPSE behavior for process_madvise(2) if caller has
CAP_SYS_ADMIN or is requesting collapse of it's own memory.

This is useful for the development of userspace agents that seek to
optimize THP utilization system-wide by using userspace signals to
prioritize what memory is most deserving of being THP-backed.

Link: https://lkml.kernel.org/r/20220706235936.2197195-13-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: rename prefix of shared collapse functions

The following functions are shared between khugepaged and madvise collapse
contexts.  Replace the "khugepaged_" prefix with generic "hpage_collapse_"
prefix in such cases:

khugepaged_test_exit() -> hpage_collapse_test_exit()
khugepaged_scan_abort() -> hpage_collapse_scan_abort()
khugepaged_scan_pmd() -> hpage_collapse_scan_pmd()
khugepaged_find_target_node() -> hpage_collapse_find_target_node()
khugepaged_alloc_page() -> hpage_collapse_alloc_page()

The kerenel ABI (e.g.  huge_memory:mm_khugepaged_scan_pmd tracepoint) is
unaltered.

Link: https://lkml.kernel.org/r/20220706235936.2197195-11-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: delay computation of hpage boundaries until use

Only compute hstart/hend once we've passed all checks that would cause
early return in madvise_collapse().

Link: https://lkml.kernel.org/r/20220720140603.1958773-4-zokeefe@google.com
Fixes: c9d968ffd9ba ("mm/madvise: introduce MADV_COLLAPSE sync hugepage collapse")
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: add missing kfree() to madvise_collapse()

smatch warnings:
mm/khugepaged.c:2409 madvise_collapse() warn: possible memory leak of
'cc'

Link: https://lore.kernel.org/linux-mm/202207100715.TBIYQ4fc-lkp@intel.com/
Link: https://lore.kernel.org/linux-mm/20220713024109.62810-1-jrdr.linux@gmail.com/
Link: https://lkml.kernel.org/r/20220713161851.1879439-1-zokeefe@google.com
Fixes: 3f7416127072 ("mm/madvise: introduce MADV_COLLAPSE sync hugepage collapse")
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: avoid possible memory leak in failure path

smatch warnings:
mm/khugepaged.c:2409 madvise_collapse() warn: possible memory
leak of 'cc'

Avoiding possible memory leak.

Link: https://lkml.kernel.org/r/20220713024109.62810-1-jrdr.linux@gmail.com
Signed-off-by: Souptick Joarder (HPE) <jrdr.linux@gmail.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Zach O'Keefe <zokeefe@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/madvise: introduce MADV_COLLAPSE sync hugepage collapse

This idea was introduced by David Rientjes[1].

Introduce a new madvise mode, MADV_COLLAPSE, that allows users to request
a synchronous collapse of memory at their own expense.

The benefits of this approach are:

* CPU is charged to the process that wants to spend the cycles for the
  THP
* Avoid unpredictable timing of khugepaged collapse

Semantics

This call is independent of the system-wide THP sysfs settings, but will
fail for memory marked VM_NOHUGEPAGE.  If the ranges provided span
multiple VMAs, the semantics of the collapse over each VMA is independent
from the others.  This implies a hugepage cannot cross a VMA boundary.  If
collapse of a given hugepage-aligned/sized region fails, the operation may
continue to attempt collapsing the remainder of memory specified.

The memory ranges provided must be page-aligned, but are not required to
be hugepage-aligned.  If the memory ranges are not hugepage-aligned, the
start/end of the range will be clamped to the first/last hugepage-aligned
address covered by said range.  The memory ranges must span at least one
hugepage-sized region.

All non-resident pages covered by the range will first be
swapped/faulted-in, before being internally copied onto a freshly
allocated hugepage.  Unmapped pages will have their data directly
initialized to 0 in the new hugepage.  However, for every eligible
hugepage aligned/sized region to-be collapsed, at least one page must
currently be backed by memory (a PMD covering the address range must
already exist).

Allocation for the new hugepage may enter direct reclaim and/or
compaction, regardless of VMA flags.  When the system has multiple NUMA
nodes, the hugepage will be allocated from the node providing the most
native pages.  This operation operates on the current state of the
specified process and makes no persistent changes or guarantees on how
pages will be mapped, constructed, or faulted in the future

Return Value

If all hugepage-sized/aligned regions covered by the provided range were
either successfully collapsed, or were already PMD-mapped THPs, this
operation will be deemed successful.  On success, process_madvise(2)
returns the number of bytes advised, and madvise(2) returns 0.  Else, -1
is returned and errno is set to indicate the error for the most-recently
attempted hugepage collapse.  Note that many failures might have occurred,
since the operation may continue to collapse in the event a single
hugepage-sized/aligned region fails.

ENOMEM Memory allocation failed or VMA not found
EBUSY Memcg charging failed
EAGAIN Required resource temporarily unavailable.  Try again
might succeed.
EINVAL Other error: No PMD found, subpage doesn't have Present
bit set, "Special" page no backed by struct page, VMA
incorrectly sized, address not page-aligned, ...

Most notable here is ENOMEM and EBUSY (new to madvise) which are intended
to provide the caller with actionable feedback so they may take an
appropriate fallback measure.

Use Cases

An immediate user of this new functionality are malloc() implementations
that manage memory in hugepage-sized chunks, but sometimes subrelease
memory back to the system in native-sized chunks via MADV_DONTNEED;
zapping the pmd.  Later, when the memory is hot, the implementation could
madvise(MADV_COLLAPSE) to re-back the memory by THPs to regain hugepage
coverage and dTLB performance.  TCMalloc is such an implementation that
could benefit from this[2].

Only privately-mapped anon memory is supported for now, but additional
support for file, shmem, and HugeTLB high-granularity mappings[2] is
expected.  File and tmpfs/shmem support would permit:

* Backing executable text by THPs.  Current support provided by
  CONFIG_READ_ONLY_THP_FOR_FS may take a long time on a large system which
  might impair services from serving at their full rated load after
  (re)starting.  Tricks like mremap(2)'ing text onto anonymous memory to
  immediately realize iTLB performance prevents page sharing and demand
  paging, both of which increase steady state memory footprint.  With
  MADV_COLLAPSE, we get the best of both worlds: Peak upfront performance
  and lower RAM footprints.
* Backing guest memory by hugapages after the memory contents have been
  migrated in native-page-sized chunks to a new host, in a
  userfaultfd-based live-migration stack.

[1] https://lore.kernel.org/linux-mm/d098c392-273a-36a4-1a29-59731cdf5d3d@google.com/
[2] https://github.com/google/tcmalloc/tree/master/tcmalloc

Link: https://lkml.kernel.org/r/20220706235936.2197195-10-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Suggested-by: David Rientjes <rientjes@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: record SCAN_PMD_MAPPED when scan_pmd() finds hugepage

When scanning an anon pmd to see if it's eligible for collapse, return
SCAN_PMD_MAPPED if the pmd already maps a hugepage.  Note that
SCAN_PMD_MAPPED is different from SCAN_PAGE_COMPOUND used in the
file-collapse path, since the latter might identify pte-mapped compound
pages.  This is required by MADV_COLLAPSE which necessarily needs to know
what hugepage-aligned/sized regions are already pmd-mapped.

In order to determine if a pmd already maps a hugepage, refactor
mm_find_pmd():

Return mm_find_pmd() to it's pre-commit f72e7dcdd252 ("mm: let mm_find_pmd
fix buggy race with THP fault") behavior.  ksm was the only caller that
explicitly wanted a pte-mapping pmd, so open code the pte-mapping logic
there (pmd_present() and pmd_trans_huge() checks).

Undo revert change in commit f72e7dcdd252 ("mm: let mm_find_pmd fix buggy
race with THP fault") that open-coded split_huge_pmd_address() pmd lookup
and use mm_find_pmd() instead.

Link: https://lkml.kernel.org/r/20220706235936.2197195-9-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: consistently order cc->is_khugepaged and pte_* checks

cc->is_khugepaged is used to predicate the khugepaged-only behavior of
enforcing khugepaged heuristics limited by the sysfs knobs
khugepaged_max_ptes_[none|swap|shared].

In branches where khugepaged_max_ptes_* is checked, consistently check
cc->is_khugepaged first.  Also, local counters (for comparison vs
khugepaged_max_ptes_* limits) were previously incremented in the
comparison expression.  Some of these counters (unmapped) are additionally
used outside of khugepaged_max_ptes_* enforcement, and all counters are
communicated in tracepoints.  Move the correct accounting of these
counters before branching statements to avoid future errors due to C's
short-circuiting evaluation.

Link: https://lkml.kernel.org/r/20220720140603.1958773-3-zokeefe@google.com
Link: https://lore.kernel.org/linux-mm/Ys2qJm6FaOQcxkha@google.com/
Fixes: 9fab4752a181 ("mm/khugepaged: add flag to predicate khugepaged-only behavior")
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/thp: add flag to enforce sysfs THP in hugepage_vma_check()

MADV_COLLAPSE is not coupled to the kernel-oriented sysfs THP settings[1].

hugepage_vma_check() is the authority on determining if a VMA is eligible
for THP allocation/collapse, and currently enforces the sysfs THP
settings.  Add a flag to disable these checks.  For now, only apply this
arg to anon and file, which use /sys/kernel/transparent_hugepage/enabled.
We can expand this to shmem, which uses
/sys/kernel/transparent_hugepage/shmem_enabled, later.

Use this flag in collapse_pte_mapped_thp() where previously the VMA flags
passed to hugepage_vma_check() were OR'd with VM_HUGEPAGE to elide the
VM_HUGEPAGE check in "madvise" THP mode.  Prior to "mm: khugepaged: check
THP flag in hugepage_vma_check()", this check also didn't check "never"
THP mode.  As such, this restores the previous behavior of
collapse_pte_mapped_thp() where sysfs THP settings are ignored.  See
comment in code for justification why this is OK.

[1] https://lore.kernel.org/linux-mm/CAAa6QmQxay1_=Pmt8oCX2-Va18t44FV-Vs-WsQt_6+qBks4nZA@mail.gmail.com/

Link: https://lkml.kernel.org/r/20220706235936.2197195-8-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: add flag to predicate khugepaged-only behavior

Add .is_khugepaged flag to struct collapse_control so khugepaged-specific
behavior can be elided by MADV_COLLAPSE context.

Start by protecting khugepaged-specific heuristics by this flag.  In
MADV_COLLAPSE, the user presumably has reason to believe the collapse will
be beneficial and khugepaged heuristics shouldn't prevent the user from
doing so:

1) sysfs-controlled knobs khugepaged_max_ptes_[none|swap|shared]

2) requirement that some pages in region being collapsed be young or
   referenced

Link: https://lkml.kernel.org/r/20220706235936.2197195-7-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: propagate enum scan_result codes back to callers

Propagate enum scan_result codes back through return values of
functions downstream of khugepaged_scan_file() and
khugepaged_scan_pmd() to inform callers if the operation was
successful, and if not, why.

Since khugepaged_scan_pmd()'s return value already has a specific meaning
(whether mmap_lock was unlocked or not), add a bool* argument to
khugepaged_scan_pmd() to retrieve this information.

Change khugepaged to take action based on the return values of
khugepaged_scan_file() and khugepaged_scan_pmd() instead of acting deep
within the collapsing functions themselves.

hugepage_vma_revalidate() now returns SCAN_SUCCEED on success to be more
consistent with enum scan_result propagation.

Remove dependency on error pointers to communicate to khugepaged that
allocation failed and it should sleep; instead just use the result of the
scan (SCAN_ALLOC_HUGE_PAGE_FAIL if allocation fails).

Link: https://lkml.kernel.org/r/20220706235936.2197195-6-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: dedup and simplify hugepage alloc and charging

The following code is duplicated in collapse_huge_page() and
collapse_file():

        gfp = alloc_hugepage_khugepaged_gfpmask() | __GFP_THISNODE;

new_page = khugepaged_alloc_page(hpage, gfp, node);
        if (!new_page) {
                result = SCAN_ALLOC_HUGE_PAGE_FAIL;
                goto out;
        }

        if (unlikely(mem_cgroup_charge(page_folio(new_page), mm, gfp))) {
                result = SCAN_CGROUP_CHARGE_FAIL;
                goto out;
        }
        count_memcg_page_event(new_page, THP_COLLAPSE_ALLOC);

Also, "node" is passed as an argument to both collapse_huge_page() and
collapse_file() and obtained the same way, via
khugepaged_find_target_node().

Move all this into a new helper, alloc_charge_hpage(), and remove the
duplicate code from collapse_huge_page() and collapse_file().  Also,
simplify khugepaged_alloc_page() by returning a bool indicating allocation
success instead of a copy of the allocated struct page *.

Link: https://lkml.kernel.org/r/20220706235936.2197195-5-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Suggested-by: Peter Xu <peterx@redhat.com>
Acked-by: David Rientjes <rientjes@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm-khugepaged-add-struct-collapse_control-fix-fix

fix build

Link: https://lkml.kernel.org/r/20220721195508.15f1e07a@canb.auug.org.au
Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Reviewed-by: Zach O'Keefe <zokeefe@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: use minimal bits to store num page < HPAGE_PMD_NR

Minimally, node_load[] entries just need to be able to hold the maximum
value of HPAGE_PMD_NR, which is compile-time defined per-arch based on
PMD_SHIFT and PAGE_SHIFT.  node_load[] is only written either via
memset(), or with via post-increment.  struct collapse_control may be
allocated via kmalloc() in other collapse contexts, and MAX_NUMNODES may
be arbitrarily large.  #define the underlying type of node_load[] based
off HPAGE_PMD_NR to avoid excessive memory allocated for this struct.

Link: https://lkml.kernel.org/r/20220720140603.1958773-2-zokeefe@google.com
Link: https://lore.kernel.org/linux-mm/Ys2CeIm%2FQmQwWh9a@google.com/
Fixes: 3b07f3bb225a ("mm/khugepaged: add struct collapse_control")
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Reviewed-by: Yang Shi <shy828301@gmail.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/khugepaged: add struct collapse_control

Modularize hugepage collapse by introducing struct collapse_control.  This
structure serves to describe the properties of the requested collapse, as
well as serve as a local scratch pad to use during the collapse itself.

Start by moving global per-node khugepaged statistics into this new
structure.  Note that this structure is still statically allocated since
CONFIG_NODES_SHIFT might be arbitrary large, and stack-allocating a
MAX_NUMNODES-sized array could cause -Wframe-large-than= errors.

Link: https://lkml.kernel.org/r/20220706235936.2197195-4-zokeefe@google.com
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Yang Shi <shy828301@gmail.com>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm: khugepaged: don't carry huge page to the next loop for !CONFIG_NUMA

Patch series "mm: userspace hugepage collapse", v7.

Introduction
--------------------------------

This series provides a mechanism for userspace to induce a collapse of
eligible ranges of memory into transparent hugepages in process context,
thus permitting users to more tightly control their own hugepage
utilization policy at their own expense.

This idea was introduced by David Rientjes[5].

Interface
--------------------------------

The proposed interface adds a new madvise(2) mode, MADV_COLLAPSE, and
leverages the new process_madvise(2) call.

process_madvise(2)

Performs a synchronous collapse of the native pages
mapped by the list of iovecs into transparent hugepages.

This operation is independent of the system THP sysfs settings,
but attempts to collapse VMAs marked VM_NOHUGEPAGE will still fail.

THP allocation may enter direct reclaim and/or compaction.

When a range spans multiple VMAs, the semantics of the collapse
over of each VMA is independent from the others.

Caller must have CAP_SYS_ADMIN if not acting on self.

Return value follows existing process_madvise(2) conventions.  A
“success” indicates that all hugepage-sized/aligned regions
covered by the provided range were either successfully
collapsed, or were already pmd-mapped THPs.

madvise(2)

Equivalent to process_madvise(2) on self, with 0 returned on
“success”.

Current Use-Cases
--------------------------------

(1) Immediately back executable text by THPs.  Current support provided
by CONFIG_READ_ONLY_THP_FOR_FS may take a long time on a large
system which might impair services from serving at their full rated
load after (re)starting.  Tricks like mremap(2)'ing text onto
anonymous memory to immediately realize iTLB performance prevents
page sharing and demand paging, both of which increase steady state
memory footprint.  With MADV_COLLAPSE, we get the best of both
worlds: Peak upfront performance and lower RAM footprints.  Note
that subsequent support for file-backed memory is required here.

(2) malloc() implementations that manage memory in hugepage-sized
chunks, but sometimes subrelease memory back to the system in
native-sized chunks via MADV_DONTNEED; zapping the pmd.  Later,
when the memory is hot, the implementation could
madvise(MADV_COLLAPSE) to re-back the memory by THPs to regain
hugepage coverage and dTLB performance.  TCMalloc is such an
implementation that could benefit from this[6].  A prior study of
Google internal workloads during evaluation of Temeraire, a
hugepage-aware enhancement to TCMalloc, showed that nearly 20% of
all cpu cycles were spent in dTLB stalls, and that increasing
hugepage coverage by even small amount can help with that[7].

(3) userfaultfd-based live migration of virtual machines satisfy UFFD
faults by fetching native-sized pages over the network (to avoid
latency of transferring an entire hugepage).  However, after guest
memory has been fully copied to the new host, MADV_COLLAPSE can
be used to immediately increase guest performance.  Note that
subsequent support for file/shmem-backed memory is required here.

(4) HugeTLB high-granularity mapping allows HugeTLB a HugeTLB page to
be mapped at different levels in the page tables[8].  As it's not
"transparent" like THP, HugeTLB high-granularity mappings require
an explicit user API. It is intended that MADV_COLLAPSE be co-opted
for this use case[9].  Note that subsequent support for HugeTLB
memory is required here.

Future work
--------------------------------

Only private anonymous memory is supported by this series. File and
shmem memory support will be added later.

One possible user of this functionality is a userspace agent that
attempts to optimize THP utilization system-wide by allocating THPs
based on, for example, task priority, task performance requirements, or
heatmaps.  For the latter, one idea that has already surfaced is using
DAMON to identify hot regions, and driving THP collapse through a new
DAMOS_COLLAPSE scheme[10].

This patch (of 17):

The khugepaged has optimization to reduce huge page allocation calls for
!CONFIG_NUMA by carrying the allocated but failed to collapse huge page to
the next loop.  CONFIG_NUMA doesn't do so since the next loop may try to
collapse huge page from a different node, so it doesn't make too much
sense to carry it.

But when NUMA=n, the huge page is allocated by khugepaged_prealloc_page()
before scanning the address space, so it means huge page may be allocated
even though there is no suitable range for collapsing.  Then the page
would be just freed if khugepaged already made enough progress.  This
could make NUMA=n run have 5 times as much thp_collapse_alloc as NUMA=y
run.  This problem actually makes things worse due to the way more
pointless THP allocations and makes the optimization pointless.

This could be fixed by carrying the huge page across scans, but it will
complicate the code further and the huge page may be carried indefinitely.
But if we take one step back, the optimization itself seems not worth
keeping nowadays since:

  * Not too many users build NUMA=n kernel nowadays even though the kernel is
    actually running on a non-NUMA machine. Some small devices may run NUMA=n
    kernel, but I don't think they actually use THP.
  * Since commit 44042b449872 ("mm/page_alloc: allow high-order pages to be
    stored on the per-cpu lists"), THP could be cached by pcp.  This actually
    somehow does the job done by the optimization.

Link: https://lkml.kernel.org/r/20220706235936.2197195-1-zokeefe@google.com
Link: https://lkml.kernel.org/r/20220706235936.2197195-3-zokeefe@google.com
Signed-off-by: Yang Shi <shy828301@gmail.com>
Signed-off-by: Zach O'Keefe <zokeefe@google.com>
Co-developed-by: Peter Xu <peterx@redhat.com>
Signed-off-by: Peter Xu <peterx@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Alex Shi <alex.shi@linux.alibaba.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Chris Kennelly <ckennelly@google.com>
Cc: Chris Zankel <chris@zankel.net>
Cc: David Hildenbrand <david@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Helge Deller <deller@gmx.de>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Max Filippov <jcmvbkbc@gmail.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Pasha Tatashin <pasha.tatashin@soleen.com>
Cc: Pavel Begunkov <asml.silence@gmail.com>
Cc: Rongwei Wang <rongwei.wang@linux.alibaba.com>
Cc: SeongJae Park <sj@kernel.org>
Cc: Song Liu <songliubraving@fb.com>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <ziy@nvidia.com>
Cc: Dan Carpenter <dan.carpenter@oracle.com>
Cc: "Souptick Joarder (HPE)" <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

procfs-add-path-to-proc-pid-fdinfo-fix

warning: Local variable 'anon_aops' shadows outer variable

Reported-by: kernel test robot <lkp@intel.com>
Cc: Kalesh Singh <kaleshsingh@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

procfs: add 'path' to /proc/<pid>/fdinfo/

In order to identify the type of memory a process has pinned through its
open fds, add the file path to fdinfo output.  This allows identifying
memory types based on common prefixes: e.g.  "/memfd...", "/dmabuf...",
"/dev/ashmem...".

To be cautious, only expose the paths for anonymous inodes, and this also
avoids printing path names with strange characters.

Access to /proc/<pid>/fdinfo is governed by PTRACE_MODE_READ_FSCREDS the
same as /proc/<pid>/maps which also exposes the file path of mappings; so
the security permissions for accessing path is consistent with that of
/proc/<pid>/maps.

Link: https://lkml.kernel.org/r/20220623220613.3014268-3-kaleshsingh@google.com
Signed-off-by: Kalesh Singh <kaleshsingh@google.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christian König <christian.koenig@amd.com>
Cc: Christoph Anton Mitterer <mail@christoph.anton.mitterer.name>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Colin Cross <ccross@google.com>
Cc: David Laight <David.Laight@ACULAB.COM>
Cc: Ioannis Ilkos <ilkos@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Paul Gortmaker<paul.gortmaker@windriver.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Stephen Brennan <stephen.s.brennan@oracle.com>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: T.J. Mercier <tjmercier@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

procfs: add 'size' to /proc/<pid>/fdinfo/

Patch series "procfs: Add file path and size to /proc/<pid>/fdinfo", v2.

Processes can pin shared memory by keeping a handle to it through a
file descriptor; for instance dmabufs, memfd, and ashmem (in Android).

In the case of a memory leak, to identify the process pinning the
memory, userspace needs to:
  - Iterate the /proc/<pid>/fd/* for each process
  - Do a readlink on each entry to identify the type of memory from
    the file path.
  - stat() each entry to get the size of the memory.

The file permissions on /proc/<pid>/fd/* only allows for the owner
or root to perform the operations above; and so is not suitable for
capturing the system-wide state in a production environment.

This issue was addressed for dmabufs by making /proc/*/fdinfo/*
accessible to a process with PTRACE_MODE_READ_FSCREDS credentials[1]
To allow the same kind of tracking for other types of shared memory,
add the following fields to /proc/<pid>/fdinfo/<fd>:

path - This allows identifying the type of memory based on common
       prefixes: e.g. "/memfd...", "/dmabuf...", "/dev/ashmem..."

       This was not an issued when dmabuf tracking was introduced
       because the exp_name field of dmabuf fdinfo could be used
       to distinguish dmabuf fds from other types.

size - To track the amount of memory that is being pinned.

       dmabufs expose size as an additional field in fdinfo. Remove
       this and make it a common field for all fds.

Access to /proc/<pid>/fdinfo is governed by PTRACE_MODE_READ_FSCREDS
-- the same as for /proc/<pid>/maps which also exposes the path and
size for mapped memory regions.

This allows for a system process with PTRACE_MODE_READ_FSCREDS to
account the pinned per-process memory via fdinfo.

This patch (of 2):

To be able to account the amount of memory a process is keeping pinned by
open file descriptors add a 'size' field to fdinfo output.

dmabufs fds already expose a 'size' field for this reason, remove this and
make it a common field for all fds.  This allows tracking of other types
of memory (e.g.  memfd and ashmem in Android).

Link: https://lkml.kernel.org/r/20220623220613.3014268-1-kaleshsingh@google.com
Link: https://lkml.kernel.org/r/20220623220613.3014268-2-kaleshsingh@google.com
Signed-off-by: Kalesh Singh <kaleshsingh@google.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Stephen Brennan <stephen.s.brennan@oracle.com>
Cc: David Laight <David.Laight@ACULAB.COM>
Cc: Ioannis Ilkos <ilkos@google.com>
Cc: T.J. Mercier <tjmercier@google.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Jonathan Corbet <corbet@lwn.net>
Cc: Sumit Semwal <sumit.semwal@linaro.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Christoph Anton Mitterer <mail@christoph.anton.mitterer.name>
Cc: Colin Cross <ccross@google.com>
Cc: Paul Gortmaker<paul.gortmaker@windriver.com>
Cc: Randy Dunlap <rdunlap@infradead.org>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>

mm/page_alloc: minor clean up for memmap_init_compound()

Since commit 5232c63f46fd ("mm: Make compound_pincount always available"),
compound_pincount_ptr is stored at first tail page now.  So we should call
prep_compound_head() after the first tail page is initialized to take
advantage of the likelihood of that tail struct page being cached given
that we will read them right after in prep_compound_head().

Link: https://lkml.kernel.org/r/20220611021352.13529-1-linmiaohe@huawei.com
Signed-off-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Joao Martins <joao.m.martins@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>