Workloads consisting of a large number of processes running the same
program with a very large shared data segment may experience performance
problems when numa balancing attempts to migrate the shared cow pages.
This manifests itself with many processes or tasks in
TASK_UNINTERRUPTIBLE state waiting for the shared pages to be migrated.
The program listed below simulates the conditions with these results
when run with 288 processes on a 144 core/8 socket machine.
Average throughput Average throughput Average throughput
with numa_balancing=0 with numa_balancing=1 with numa_balancing=1
without the patch with the patch
--------------------- --------------------- ---------------------
2118782 2021534 2107979
Complex production environments show less variability and fewer poorly
performing outliers accompanied with a smaller number of processes
waiting on NUMA page migration with this patch applied. In some cases,
%iowait drops from 16%-26% to 0.
// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (c) 2017 Oracle and/or its affiliates. All rights reserved.
*/
#include <sys/time.h>
#include <stdio.h>
#include <wait.h>
#include <sys/mman.h>
int a[1000000] = {13};
int main(int argc, const char **argv)
{
int n = 0;
int i;
pid_t pid;
int stat;
int *count_array;
int cpu_count = 288;
long total = 0;
struct timeval t1, t2 = {(argc > 1 ? atoi(argv[1]) : 10), 0};
if (argc > 2)
cpu_count = atoi(argv[2]);
count_array = mmap(NULL, cpu_count * sizeof(int),
(PROT_READ|PROT_WRITE),
(MAP_SHARED|MAP_ANONYMOUS), 0, 0);
if (count_array == MAP_FAILED) {
perror("mmap:");
return 0;
}
for (i = 0; i < cpu_count; ++i) {
pid = fork();
if (pid <= 0)
break;
if ((i & 0xf) == 0)
usleep(2);
}
if (pid != 0) {
if (i == 0) {
perror("fork:");
return 0;
}
for (;;) {
pid = wait(&stat);
if (pid < 0)
break;
}
for (i = 0; i < cpu_count; ++i)
total += count_array[i];
printf("Total %ld\n", total);
munmap(count_array, cpu_count * sizeof(int));
return 0;
}
gettimeofday(&t1, 0);
timeradd(&t1, &t2, &t1);
while (timercmp(&t2, &t1, <)) {
int b = 0;
int j;
for (j = 0; j < 1000000; j++)
b += a[j];
gettimeofday(&t2, 0);
n++;
}
count_array[i] = n;
return 0;
}
This patch changes change_pte_range() to skip shared copy-on-write pages
when called from change_prot_numa().
NOTE: change_prot_numa() is nominally called from task_numa_work() and
queue_pages_test_walk(). task_numa_work() is the auto NUMA balancing
path, and queue_pages_test_walk() is part of explicit NUMA policy
management. However, queue_pages_test_walk() only calls
change_prot_numa() when MPOL_MF_LAZY is specified and currently that is
not allowed, so change_prot_numa() is only called from auto NUMA
balancing.
In the case of explicit NUMA policy management, shared pages are not
migrated unless MPOL_MF_MOVE_ALL is specified, and MPOL_MF_MOVE_ALL
depends on CAP_SYS_NICE. Currently, there is no way to pass information
about MPOL_MF_MOVE_ALL to change_pte_range. This will have to be fixed
if MPOL_MF_LAZY is enabled and MPOL_MF_MOVE_ALL is to be honored in lazy
migration mode.
task_numa_work() skips the read-only VMAs of programs and shared
libraries.
Link: http://lkml.kernel.org/r/1516751617-7369-1-git-send-email-henry.willard@oracle.com
Signed-off-by: Henry Willard <henry.willard@oracle.com>
Reviewed-by: Håkon Bugge <haakon.bugge@oracle.com>
Reviewed-by: Steve Sistare <steven.sistare@oracle.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Kate Stewart <kstewart@linuxfoundation.org>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Cc: Philippe Ombredanne <pombredanne@nexb.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: "Jérôme Glisse" <jglisse@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Dan Carpenter has noticed that mbind migration callback (new_page) can
get a NULL vma pointer and choke on it inside alloc_huge_page_vma which
relies on the VMA to get the hstate. We used to BUG_ON this case but
the BUG_+ON has been removed recently by "hugetlb, mempolicy: fix the
mbind hugetlb migration".
The proper way to handle this is to get the hstate from the migrated
page and rely on huge_node (resp. get_vma_policy) do the right thing
with null VMA. We are currently falling back to the default mempolicy
in that case which is in line what THP path is doing here.
Link: http://lkml.kernel.org/r/20180110104712.GR1732@dhcp22.suse.cz
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
do_mbind migration code relies on alloc_huge_page_noerr for hugetlb
pages. alloc_huge_page_noerr uses alloc_huge_page which is a highlevel
allocation function which has to take care of reserves, overcommit or
hugetlb cgroup accounting. None of that is really required for the page
migration because the new page is only temporal and either will replace
the original page or it will be dropped. This is essentially as for
other migration call paths and there shouldn't be any reason to handle
mbind in a special way.
The current implementation is even suboptimal because the migration
might fail just because the hugetlb cgroup limit is reached, or the
overcommit is saturated.
Fix this by making mbind like other hugetlb migration paths. Add a new
migration helper alloc_huge_page_vma as a wrapper around
alloc_huge_page_nodemask with additional mempolicy handling.
alloc_huge_page_noerr has no more users and it can go.
Link: http://lkml.kernel.org/r/20180103093213.26329-7-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Reale <ar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Hugetlb allocator has several layer of allocation functions depending
and the purpose of the allocation. There are two allocators depending
on whether the page can be allocated from the page allocator or we need
a contiguous allocator. This is currently opencoded in
alloc_fresh_huge_page which is the only path that might allocate giga
pages which require the later allocator. Create alloc_fresh_huge_page
which hides this implementation detail and use it in all callers which
hardcoded the buddy allocator path (__hugetlb_alloc_buddy_huge_page).
This shouldn't introduce any funtional change because both migration and
surplus allocators exlude giga pages explicitly.
While we are at it let's do some renaming. The current scheme is not
consistent and overly painfull to read and understand. Get rid of
prefix underscores from most functions. There is no real reason to make
names longer.
* alloc_fresh_huge_page is the new layer to abstract underlying
allocator
* __hugetlb_alloc_buddy_huge_page becomes shorter and neater
alloc_buddy_huge_page.
* Former alloc_fresh_huge_page becomes alloc_pool_huge_page because we put
the new page directly to the pool
* alloc_surplus_huge_page can drop the opencoded prep_new_huge_page code
as it uses alloc_fresh_huge_page now
* others lose their excessive prefix underscores to make names shorter
[dan.carpenter@oracle.com: fix double unlock bug in alloc_surplus_huge_page()]
Link: http://lkml.kernel.org/r/20180109200559.g3iz5kvbdrz7yydp@mwanda
Link: http://lkml.kernel.org/r/20180103093213.26329-6-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Reale <ar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
alloc_surplus_huge_page increases the pool size and the number of
surplus pages opportunistically to prevent from races with the pool size
change. See commit d1c3fb1f8f ("hugetlb: introduce
nr_overcommit_hugepages sysctl") for more details.
The resulting code is unnecessarily hairy, cause code duplication and
doesn't allow to share the allocation paths. Moreover pool size changes
tend to be very seldom so optimizing for them is not really reasonable.
Simplify the code and allow to allocate a fresh surplus page as long as
we are under the overcommit limit and then recheck the condition after
the allocation and drop the new page if the situation has changed. This
should provide a reasonable guarantee that an abrupt allocation requests
will not go way off the limit.
If we consider races with the pool shrinking and enlarging then we
should be reasonably safe as well. In the first case we are off by one
in the worst case and the second case should work OK because the page is
not yet visible. We can waste CPU cycles for the allocation but that
should be acceptable for a relatively rare condition.
Link: http://lkml.kernel.org/r/20180103093213.26329-5-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Reale <ar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
hugepage migration relies on __alloc_buddy_huge_page to get a new page.
This has 2 main disadvantages.
1) it doesn't allow to migrate any huge page if the pool is used
completely which is not an exceptional case as the pool is static and
unused memory is just wasted.
2) it leads to a weird semantic when migration between two numa nodes
might increase the pool size of the destination NUMA node while the
page is in use. The issue is caused by per NUMA node surplus pages
tracking (see free_huge_page).
Address both issues by changing the way how we allocate and account
pages allocated for migration. Those should temporal by definition. So
we mark them that way (we will abuse page flags in the 3rd page) and
update free_huge_page to free such pages to the page allocator. Page
migration path then just transfers the temporal status from the new page
to the old one which will be freed on the last reference. The global
surplus count will never change during this path but we still have to be
careful when migrating a per-node suprlus page. This is now handled in
move_hugetlb_state which is called from the migration path and it copies
the hugetlb specific page state and fixes up the accounting when needed
Rename __alloc_buddy_huge_page to __alloc_surplus_huge_page to better
reflect its purpose. The new allocation routine for the migration path
is __alloc_migrate_huge_page.
The user visible effect of this patch is that migrated pages are really
temporal and they travel between NUMA nodes as per the migration
request:
Before migration
/sys/devices/system/node/node0/hugepages/hugepages-2048kB/free_hugepages:0
/sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages:1
/sys/devices/system/node/node0/hugepages/hugepages-2048kB/surplus_hugepages:0
/sys/devices/system/node/node1/hugepages/hugepages-2048kB/free_hugepages:0
/sys/devices/system/node/node1/hugepages/hugepages-2048kB/nr_hugepages:0
/sys/devices/system/node/node1/hugepages/hugepages-2048kB/surplus_hugepages:0
After
/sys/devices/system/node/node0/hugepages/hugepages-2048kB/free_hugepages:0
/sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages:0
/sys/devices/system/node/node0/hugepages/hugepages-2048kB/surplus_hugepages:0
/sys/devices/system/node/node1/hugepages/hugepages-2048kB/free_hugepages:0
/sys/devices/system/node/node1/hugepages/hugepages-2048kB/nr_hugepages:1
/sys/devices/system/node/node1/hugepages/hugepages-2048kB/surplus_hugepages:0
with the previous implementation, both nodes would have nr_hugepages:1
until the page is freed.
Link: http://lkml.kernel.org/r/20180103093213.26329-4-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Reale <ar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Gigantic hugetlb pages were ingrown to the hugetlb code as an alien
specie with a lot of special casing. The allocation path is not an
exception. Unnecessarily so to be honest. It is true that the
underlying allocator is different but that is an implementation detail.
This patch unifies the hugetlb allocation path that a prepares fresh
pool pages. alloc_fresh_gigantic_page basically copies
alloc_fresh_huge_page logic so we can move everything there. This will
simplify set_max_huge_pages which doesn't have to care about what kind
of huge page we allocate.
Link: http://lkml.kernel.org/r/20180103093213.26329-3-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Reale <ar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Patch series "mm, hugetlb: allocation API and migration improvements"
Motivation:
this is a follow up for [3] for the allocation API and [4] for the
hugetlb migration. It wasn't really easy to split those into two
separate patch series as they share some code.
My primary motivation to touch this code is to make the gigantic pages
migration working. The giga pages allocation code is just too fragile
and hacked into the hugetlb code now. This series tries to move giga
pages closer to the first class citizen. We are not there yet but
having 5 patches is quite a lot already and it will already make the
code much easier to follow. I will come with other changes on top after
this sees some review.
The first two patches should be trivial to review. The third patch
changes the way how we migrate huge pages. Newly allocated pages are a
subject of the overcommit check and they participate surplus accounting
which is quite unfortunate as the changelog explains. This patch
doesn't change anything wrt. giga pages.
Patch #4 removes the surplus accounting hack from
__alloc_surplus_huge_page. I hope I didn't miss anything there and a
deeper review is really due there.
Patch #5 finally unifies allocation paths and giga pages shouldn't be
any special anymore. There is also some renaming going on as well.
This patch (of 6):
hugetlb allocator has two entry points to the page allocator
- alloc_fresh_huge_page_node
- __hugetlb_alloc_buddy_huge_page
The two differ very subtly in two aspects. The first one doesn't care
about HTLB_BUDDY_* stats and it doesn't initialize the huge page.
prep_new_huge_page is not used because it not only initializes hugetlb
specific stuff but because it also put_page and releases the page to the
hugetlb pool which is not what is required in some contexts. This makes
things more complicated than necessary.
Simplify things by a) removing the page allocator entry point duplicity
and only keep __hugetlb_alloc_buddy_huge_page and b) make
prep_new_huge_page more reusable by removing the put_page which moves
the page to the allocator pool. All current callers are updated to call
put_page explicitly. Later patches will add new callers which won't
need it.
This patch shouldn't introduce any functional change.
Link: http://lkml.kernel.org/r/20180103093213.26329-2-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Andrea Reale <ar@linux.vnet.ibm.com>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Zi Yan <zi.yan@cs.rutgers.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mem_cgroup_resize_[memsw]_limit() tries to free only 32
(SWAP_CLUSTER_MAX) pages on each iteration. This makes it practically
impossible to decrease limit of memory cgroup. Tasks could easily
allocate back 32 pages, so we can't reduce memory usage, and once
retry_count reaches zero we return -EBUSY.
Easy to reproduce the problem by running the following commands:
mkdir /sys/fs/cgroup/memory/test
echo $$ >> /sys/fs/cgroup/memory/test/tasks
cat big_file > /dev/null &
sleep 1 && echo $((100*1024*1024)) > /sys/fs/cgroup/memory/test/memory.limit_in_bytes
-bash: echo: write error: Device or resource busy
Instead of relying on retry_count, keep retrying the reclaim until the
desired limit is reached or fail if the reclaim doesn't make any
progress or a signal is pending.
Link: http://lkml.kernel.org/r/20180119132544.19569-1-aryabinin@virtuozzo.com
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Shakeel Butt <shakeelb@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Vladimir Davydov <vdavydov.dev@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix the following sparse warning:
mm/memcontrol.c:1097:14: warning: symbol 'memcg1_stats' was not declared. Should it be static?
Link: http://lkml.kernel.org/r/20180118193327.14200-1-chrisadr@gentoo.org
Signed-off-by: Christopher Díaz Riveros <chrisadr@gentoo.org>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The variable 'entry' is used before being initialized in
hmm_vma_walk_pmd().
No bad effect (beside performance hit) so !non_swap_entry(0) evaluate to
true which trigger a fault as if CPU was trying to access migrated
memory and migrate memory back from device memory to regular memory.
This function (hmm_vma_walk_pmd()) is called when a device driver tries
to populate its own page table. For migrated memory it should not
happen as the device driver should already have populated its page table
correctly during the migration.
Only case I can think of is multi-GPU where a second GPU triggers
migration back to regular memory. Again this would just result in a
performance hit, nothing bad would happen.
Link: http://lkml.kernel.org/r/20180122185759.26286-1-jglisse@redhat.com
Signed-off-by: Ralph Campbell <rcampbell@nvidia.com>
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The comment is confusing. On the one hand, it refers to 32-bit
alignment (struct page alignment on 32-bit platforms), but this would
only guarantee that the 2 lowest bits must be zero. On the other hand,
it claims that at least 3 bits are available, and 3 bits are actually
used.
This is not broken, because there is a stronger alignment guarantee,
just less obvious. Let's fix the comment to make it clear how many bits
are available and why.
Although memmap arrays are allocated in various places, the resulting
pointer is encoded eventually, so I am adding a BUG_ON() here to enforce
at runtime that all expected bits are indeed available.
I have also added a BUILD_BUG_ON to check that PFN_SECTION_SHIFT is
sufficient, because this part of the calculation can be easily checked
at build time.
[ptesarik@suse.com: v2]
Link: http://lkml.kernel.org/r/20180125100516.589ea6af@ezekiel.suse.cz
Link: http://lkml.kernel.org/r/20180119080908.3a662e6f@ezekiel.suse.cz
Signed-off-by: Petr Tesarik <ptesarik@suse.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Kemi Wang <kemi.wang@intel.com>
Cc: YASUAKI ISHIMATSU <yasu.isimatu@gmail.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
"mode" argument is not used by try_to_compact_pages() and sub functions
anymore, it has been replaced by "prio". Fix the comment to explain the
use of "prio" argument.
Link: http://lkml.kernel.org/r/1515801336-20611-1-git-send-email-yang.shi@linux.alibaba.com
Signed-off-by: Yang Shi <yang.shi@linux.alibaba.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
static struct page_ext_operations *page_ext_ops[] always contains debug_guardpage_ops,
static struct page_ext_operations *page_ext_ops[] = {
&debug_guardpage_ops,
#ifdef CONFIG_PAGE_OWNER
&page_owner_ops,
#endif
...
}
but for it to work, CONFIG_DEBUG_PAGEALLOC must be enabled first. If
someone has CONFIG_PAGE_EXTENSION, but has none of its users, eg:
(CONFIG_PAGE_OWNER, CONFIG_DEBUG_PAGEALLOC, CONFIG_IDLE_PAGE_TRACKING),
we can shrink page_ext_init() to a simple retq.
$ size vmlinux (before patch)
text data bss dec hex filename
14356698 5681582 1687748 21726028 14b834c vmlinux
$ size vmlinux (after patch)
text data bss dec hex filename
14356008 5681538 1687748 21725294 14b806e vmlinux
On the other hand, it might does not even make sense, since if someone
enables CONFIG_PAGE_EXTENSION, I would expect him to enable also at
least one of its users.
Link: http://lkml.kernel.org/r/20180105130235.GA21241@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@techadventures.net>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Jaewon Kim <jaewon31.kim@samsung.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mem_cgroup_resize_limit() and mem_cgroup_resize_memsw_limit() have
identical logics. Refactor code so we don't need to keep two pieces of
code that does same thing.
Link: http://lkml.kernel.org/r/20180108224238.14583-1-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We waste sizeof(swp_entry_t) for zswap header when using zsmalloc as
zpool driver because zsmalloc doesn't support eviction.
Add zpool_evictable() to detect if zpool is potentially evictable, and
use it in zswap to avoid waste memory for zswap header.
[yuzhao@google.com: The zpool->" prefix is a result of copy & paste]
Link: http://lkml.kernel.org/r/20180110225626.110330-1-yuzhao@google.com
Link: http://lkml.kernel.org/r/20180110224741.83751-1-yuzhao@google.com
Signed-off-by: Yu Zhao <yuzhao@google.com>
Acked-by: Dan Streetman <ddstreet@ieee.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Seth Jennings <sjenning@redhat.com>
Cc: Minchan Kim <minchan@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
During our recent testing with fadvise(FADV_DONTNEED), we find that if
given offset/length is not page-aligned, the last page will not be
discarded. The tool we use is vmtouch (https://hoytech.com/vmtouch/),
we map a 10KB-sized file into memory and then try to run this tool to
evict the whole file mapping, but the last single page always remains
staying in the memory:
$./vmtouch -e test_10K
Files: 1
Directories: 0
Evicted Pages: 3 (12K)
Elapsed: 2.1e-05 seconds
$./vmtouch test_10K
Files: 1
Directories: 0
Resident Pages: 1/3 4K/12K 33.3%
Elapsed: 5.5e-05 seconds
However when we test with an older kernel, say 3.10, this problem is
gone. So we wonder if this is a regression:
$./vmtouch -e test_10K
Files: 1
Directories: 0
Evicted Pages: 3 (12K)
Elapsed: 8.2e-05 seconds
$./vmtouch test_10K
Files: 1
Directories: 0
Resident Pages: 0/3 0/12K 0% <-- partial page also discarded
Elapsed: 5e-05 seconds
After digging a little bit into this problem, we find it seems not a
regression. Not discarding partial page is likely to be on purpose
according to commit 441c228f81 ("mm: fadvise: document the
fadvise(FADV_DONTNEED) behaviour for partial pages") written by Mel
Gorman. He explained why partial pages should be preserved instead of
being discarded when using fadvise(FADV_DONTNEED).
However, the interesting part is that the actual code did NOT work as
the same as it was described, the partial page was still discarded
anyway, due to a calculation mistake of `end_index' passed to
invalidate_mapping_pages(). This mistake has not been fixed until
recently, that's why we fail to reproduce our problem in old kernels.
The fix is done in commit 18aba41cbf ("mm/fadvise.c: do not discard
partial pages with POSIX_FADV_DONTNEED") by Oleg Drokin.
Back to the original testing, our problem becomes that there is a
special case that, if the page-unaligned `endbyte' is also the end of
file, it is not necessary at all to preserve the last partial page, as
we all know no one else will use the rest of it. It should be safe
enough if we just discard the whole page. So we add an EOF check in
this patch.
We also find a poosbile real world issue in mainline kernel. Assume
such scenario: A userspace backup application want to backup a huge
amount of small files (<4k) at once, the developer might (I guess) want
to use fadvise(FADV_DONTNEED) to save memory. However, FADV_DONTNEED
won't really happen since the only page mapped is a partial page, and
kernel will preserve it. Our patch also fixes this problem, since we
know the endbyte is EOF, so we discard it.
Here is a simple reproducer to reproduce and verify each scenario we
described above:
test_fadvise.c
==============================
#include <sys/mman.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <stdlib.h>
#include <string.h>
#include <stdio.h>
#include <unistd.h>
int main(int argc, char **argv)
{
int i, fd, ret, len;
struct stat buf;
void *addr;
unsigned char *vec;
char *strbuf;
ssize_t pagesize = getpagesize();
ssize_t filesize;
fd = open(argv[1], O_RDWR|O_CREAT, S_IRUSR|S_IWUSR);
if (fd < 0)
return -1;
filesize = strtoul(argv[2], NULL, 10);
strbuf = malloc(filesize);
memset(strbuf, 42, filesize);
write(fd, strbuf, filesize);
free(strbuf);
fsync(fd);
len = (filesize + pagesize - 1) / pagesize;
printf("length of pages: %d\n", len);
addr = mmap(NULL, filesize, PROT_READ, MAP_SHARED, fd, 0);
if (addr == MAP_FAILED)
return -1;
ret = posix_fadvise(fd, 0, filesize, POSIX_FADV_DONTNEED);
if (ret < 0)
return -1;
vec = malloc(len);
ret = mincore(addr, filesize, (void *)vec);
if (ret < 0)
return -1;
for (i = 0; i < len; i++)
printf("pages[%d]: %x\n", i, vec[i] & 0x1);
free(vec);
close(fd);
return 0;
}
==============================
Test 1: running on kernel with commit 18aba41cbf reverted:
[root@caspar ~]# uname -r
4.15.0-rc6.revert+
[root@caspar ~]# ./test_fadvise file1 1024
length of pages: 1
pages[0]: 0 # <-- partial page discarded
[root@caspar ~]# ./test_fadvise file2 8192
length of pages: 2
pages[0]: 0
pages[1]: 0
[root@caspar ~]# ./test_fadvise file3 10240
length of pages: 3
pages[0]: 0
pages[1]: 0
pages[2]: 0 # <-- partial page discarded
Test 2: running on mainline kernel:
[root@caspar ~]# uname -r
4.15.0-rc6+
[root@caspar ~]# ./test_fadvise test1 1024
length of pages: 1
pages[0]: 1 # <-- partial and the only page not discarded
[root@caspar ~]# ./test_fadvise test2 8192
length of pages: 2
pages[0]: 0
pages[1]: 0
[root@caspar ~]# ./test_fadvise test3 10240
length of pages: 3
pages[0]: 0
pages[1]: 0
pages[2]: 1 # <-- partial page not discarded
Test 3: running on kernel with this patch:
[root@caspar ~]# uname -r
4.15.0-rc6.patched+
[root@caspar ~]# ./test_fadvise test1 1024
length of pages: 1
pages[0]: 0 # <-- partial page and EOF, discarded
[root@caspar ~]# ./test_fadvise test2 8192
length of pages: 2
pages[0]: 0
pages[1]: 0
[root@caspar ~]# ./test_fadvise test3 10240
length of pages: 3
pages[0]: 0
pages[1]: 0
pages[2]: 0 # <-- partial page and EOF, discarded
[akpm@linux-foundation.org: tweak code comment]
Link: http://lkml.kernel.org/r/5222da9ee20e1695eaabb69f631f200d6e6b8876.1515132470.git.jinli.zjl@alibaba-inc.com
Signed-off-by: shidao.ytt <shidao.ytt@alibaba-inc.com>
Signed-off-by: Caspar Zhang <jinli.zjl@alibaba-inc.com>
Reviewed-by: Oliver Yang <zhiche.yy@alibaba-inc.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Minchan Kim asked the following question -- what locks protects
address_space destroying when race happens between inode trauncation and
__isolate_lru_page? Jan Kara clarified by describing the race as follows
CPU1 CPU2
truncate(inode) __isolate_lru_page()
...
truncate_inode_page(mapping, page);
delete_from_page_cache(page)
spin_lock_irqsave(&mapping->tree_lock, flags);
__delete_from_page_cache(page, NULL)
page_cache_tree_delete(..)
... mapping = page_mapping(page);
page->mapping = NULL;
...
spin_unlock_irqrestore(&mapping->tree_lock, flags);
page_cache_free_page(mapping, page)
put_page(page)
if (put_page_testzero(page)) -> false
- inode now has no pages and can be freed including embedded address_space
if (mapping && !mapping->a_ops->migratepage)
- we've dereferenced mapping which is potentially already free.
The race is theoretically possible but unlikely. Before the
delete_from_page_cache, truncate_cleanup_page is called so the page is
likely to be !PageDirty or PageWriteback which gets skipped by the only
caller that checks the mappping in __isolate_lru_page. Even if the race
occurs, a substantial amount of work has to happen during a tiny window
with no preemption but it could potentially be done using a virtual
machine to artifically slow one CPU or halt it during the critical
window.
This patch should eliminate the race with truncation by try-locking the
page before derefencing mapping and aborting if the lock was not
acquired. There was a suggestion from Huang Ying to use RCU as a
side-effect to prevent mapping being freed. However, I do not like the
solution as it's an unconventional means of preserving a mapping and
it's not a context where rcu_read_lock is obviously protecting rcu data.
Link: http://lkml.kernel.org/r/20180104102512.2qos3h5vqzeisrek@techsingularity.net
Fixes: c824493528 ("mm: compaction: make isolate_lru_page() filter-aware again")
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: "Huang, Ying" <ying.huang@intel.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Adapt add_seals()/get_seals() to work with hugetbfs-backed memory.
Teach memfd_create() to allow sealing operations on MFD_HUGETLB.
Link: http://lkml.kernel.org/r/20171107122800.25517-6-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Those functions are called for memfd files, backed by shmem or hugetlb
(the next patches will handle hugetlb).
Link: http://lkml.kernel.org/r/20171107122800.25517-3-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Patch series "memfd: add sealing to hugetlb-backed memory", v3.
Recently, Mike Kravetz added hugetlbfs support to memfd. However, he
didn't add sealing support. One of the reasons to use memfd is to have
shared memory sealing when doing IPC or sharing memory with another
process with some extra safety. qemu uses shared memory & hugetables
with vhost-user (used by dpdk), so it is reasonable to use memfd now
instead for convenience and security reasons.
This patch (of 9):
The functions are called through shmem_fcntl() only. And no danger in
removing the EXPORTs as the routines only work with shmem file structs.
Link: http://lkml.kernel.org/r/20171107122800.25517-2-marcandre.lureau@redhat.com
Signed-off-by: Marc-André Lureau <marcandre.lureau@redhat.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Structure zs_pool has special flag to indicate success of shrinker
initialization. unregister_shrinker() has improved and can detect by
itself whether actual deinitialization should be performed or not, so
extra flag becomes redundant.
[akpm@linux-foundation.org: update comment (Aliaksei), remove unneeded cast]
Link: http://lkml.kernel.org/r/1513680552-9798-1-git-send-email-akaraliou.dev@gmail.com
Signed-off-by: Aliaksei Karaliou <akaraliou.dev@gmail.com>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: Michal Hocko <mhocko@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This uses the new annotation to determine if an mm has mmu notifiers
with blockable invalidate range callbacks to avoid oom reaping.
Otherwise, the callbacks are used around unmap_page_range().
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1712141330120.74052@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Paolo Bonzini <pbonzini@redhat.com>
Cc: Christian König <christian.koenig@amd.com>
Cc: Dimitri Sivanich <sivanich@hpe.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Doug Ledford <dledford@redhat.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Mike Marciniszyn <mike.marciniszyn@intel.com>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Jérôme Glisse <jglisse@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Commit 4d4bbd8526 ("mm, oom_reaper: skip mm structs with mmu
notifiers") prevented the oom reaper from unmapping private anonymous
memory with the oom reaper when the oom victim mm had mmu notifiers
registered.
The rationale is that doing mmu_notifier_invalidate_range_{start,end}()
around the unmap_page_range(), which is needed, can block and the oom
killer will stall forever waiting for the victim to exit, which may not
be possible without reaping.
That concern is real, but only true for mmu notifiers that have
blockable invalidate_range_{start,end}() callbacks. This patch adds a
"flags" field to mmu notifier ops that can set a bit to indicate that
these callbacks do not block.
The implementation is steered toward an expensive slowpath, such as
after the oom reaper has grabbed mm->mmap_sem of a still alive oom
victim.
[rientjes@google.com: mmu_notifier_invalidate_range_end() can also call the invalidate_range() must not block, fix comment]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1801091339570.240101@chino.kir.corp.google.com
[akpm@linux-foundation.org: make mm_has_blockable_invalidate_notifiers() return bool, use rwsem_is_locked()]
Link: http://lkml.kernel.org/r/alpine.DEB.2.10.1712141329500.74052@chino.kir.corp.google.com
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Paolo Bonzini <pbonzini@redhat.com>
Acked-by: Christian König <christian.koenig@amd.com>
Acked-by: Dimitri Sivanich <sivanich@hpe.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Oded Gabbay <oded.gabbay@gmail.com>
Cc: Alex Deucher <alexander.deucher@amd.com>
Cc: David Airlie <airlied@linux.ie>
Cc: Joerg Roedel <joro@8bytes.org>
Cc: Doug Ledford <dledford@redhat.com>
Cc: Jani Nikula <jani.nikula@linux.intel.com>
Cc: Mike Marciniszyn <mike.marciniszyn@intel.com>
Cc: Sean Hefty <sean.hefty@intel.com>
Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Cc: Jérôme Glisse <jglisse@redhat.com>
Cc: Radim Krčmář <rkrcmar@redhat.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In the current design, khugepaged needs to acquire mmap_sem before
scanning an mm. But in some corner cases, khugepaged may scan a process
which is modifying its memory mapping, so khugepaged blocks in
uninterruptible state. But the process might hold the mmap_sem for a
long time when modifying a huge memory space and it may trigger the
below khugepaged hung issue:
INFO: task khugepaged:270 blocked for more than 120 seconds.
Tainted: G E 4.9.65-006.ali3000.alios7.x86_64 #1
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
khugepaged D 0 270 2 0x00000000
ffff883f3deae4c0 0000000000000000 ffff883f610596c0 ffff883f7d359440
ffff883f63818000 ffffc90019adfc78 ffffffff817079a5 d67e5aa8c1860a64
0000000000000246 ffff883f7d359440 ffffc90019adfc88 ffff883f610596c0
Call Trace:
schedule+0x36/0x80
rwsem_down_read_failed+0xf0/0x150
call_rwsem_down_read_failed+0x18/0x30
down_read+0x20/0x40
khugepaged+0x476/0x11d0
kthread+0xe6/0x100
ret_from_fork+0x25/0x30
So it sounds pointless to just block khugepaged waiting for the
semaphore so replace down_read() with down_read_trylock() to move to
scan the next mm quickly instead of just blocking on the semaphore so
that other processes can get more chances to install THP. Then
khugepaged can come back to scan the skipped mm when it has finished the
current round full_scan.
And it appears that the change can improve khugepaged efficiency a
little bit.
Below is the test result when running LTP on a 24 cores 4GB memory 2
nodes NUMA VM:
pristine w/ trylock
full_scan 197 187
pages_collapsed 21 26
thp_fault_alloc 40818 44466
thp_fault_fallback 18413 16679
thp_collapse_alloc 21 150
thp_collapse_alloc_failed 14 16
thp_file_alloc 369 369
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: tweak comment]
[arnd@arndb.de: avoid uninitialized variable use]
Link: http://lkml.kernel.org/r/20171215125129.2948634-1-arnd@arndb.de
Link: http://lkml.kernel.org/r/1513281203-54878-1-git-send-email-yang.s@alibaba-inc.com
Signed-off-by: Yang Shi <yang.s@alibaba-inc.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Instead of marking the pmd ready for split, invalidate the pmd. This
should take care of powerpc requirement. Only side effect is that we
mark the pmd invalid early. This can result in us blocking access to
the page a bit longer if we race against a thp split.
[kirill.shutemov@linux.intel.com: rebased, dirty THP once]
Link: http://lkml.kernel.org/r/20171213105756.69879-13-kirill.shutemov@linux.intel.com
Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Daney <david.daney@cavium.com>
Cc: David Miller <davem@davemloft.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nitin Gupta <nitin.m.gupta@oracle.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use the modifed pmdp_invalidate() that returns the previous value of pmd
to transfer dirty and accessed bits.
Link: http://lkml.kernel.org/r/20171213105756.69879-12-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Daney <david.daney@cavium.com>
Cc: David Miller <davem@davemloft.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nitin Gupta <nitin.m.gupta@oracle.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vlastimil noted that pmdp_invalidate() is not atomic and we can lose
dirty and access bits if CPU sets them after pmdp dereference, but
before set_pmd_at().
The patch change pmdp_invalidate() to make the entry non-present
atomically and return previous value of the entry. This value can be
used to check if CPU set dirty/accessed bits under us.
The race window is very small and I haven't seen any reports that can be
attributed to the bug. For this reason, I don't think backporting to
stable trees needed.
Link: http://lkml.kernel.org/r/20171213105756.69879-11-kirill.shutemov@linux.intel.com
Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Reported-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: David Daney <david.daney@cavium.com>
Cc: David Miller <davem@davemloft.net>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Michal Hocko <mhocko@kernel.org>
Cc: Nitin Gupta <nitin.m.gupta@oracle.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vineet Gupta <vgupta@synopsys.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Several users of unmap_mapping_range() would prefer to express their
range in pages rather than bytes. Unfortuately, on a 32-bit kernel, you
have to remember to cast your page number to a 64-bit type before
shifting it, and four places in the current tree didn't remember to do
that. That's a sign of a bad interface.
Conveniently, unmap_mapping_range() actually converts from bytes into
pages, so hoist the guts of unmap_mapping_range() into a new function
unmap_mapping_pages() and convert the callers which want to use pages.
Link: http://lkml.kernel.org/r/20171206142627.GD32044@bombadil.infradead.org
Signed-off-by: Matthew Wilcox <mawilcox@microsoft.com>
Reported-by: "zhangyi (F)" <yi.zhang@huawei.com>
Reviewed-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
pmd_trans_splitting() was removed after THP refcounting redesign,
therefore related comment should be updated.
Link: http://lkml.kernel.org/r/1512625745-59451-1-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In register_page_bootmem_info_section() we call __nr_to_section() in
order to get the mem_section struct at the beginning of the function.
Since we already got it, there is no need for a second call to
__nr_to_section().
Link: http://lkml.kernel.org/r/20171207102914.GA12396@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@techadventures.net>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The comment describes @fullmm argument, but the function has no such
parameter.
Update the comment to match the code and convert it to kernel-doc
markup.
Link: http://lkml.kernel.org/r/1512394531-2264-1-git-send-email-rppt@linux.vnet.ibm.com
Signed-off-by: Mike Rapoport <rppt@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When we call register_page_bootmem_info_section() having
CONFIG_SPARSEMEM_VMEMMAP enabled, we check if the pfn is valid.
This check is redundant as we already checked this in
register_page_bootmem_info_node() before calling
register_page_bootmem_info_section(), so let's get rid of it.
Link: http://lkml.kernel.org/r/20171205143422.GA31458@techadventures.net
Signed-off-by: Oscar Salvador <osalvador@techadventures.net>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
hugepages_treat_as_movable has been introduced by 396faf0303 ("Allow
huge page allocations to use GFP_HIGH_MOVABLE") to allow hugetlb
allocations from ZONE_MOVABLE even when hugetlb pages were not
migrateable. The purpose of the movable zone was different at the time.
It aimed at reducing memory fragmentation and hugetlb pages being long
lived and large werre not contributing to the fragmentation so it was
acceptable to use the zone back then.
Things have changed though and the primary purpose of the zone became
migratability guarantee. If we allow non migrateable hugetlb pages to
be in ZONE_MOVABLE memory hotplug might fail to offline the memory.
Remove the knob and only rely on hugepage_migration_supported to allow
movable zones.
Mel said:
: Primarily it was aimed at allowing the hugetlb pool to safely shrink with
: the ability to grow it again. The use case was for batched jobs, some of
: which needed huge pages and others that did not but didn't want the memory
: useless pinned in the huge pages pool.
:
: I suspect that more users rely on THP than hugetlbfs for flexible use of
: huge pages with fallback options so I think that removing the option
: should be ok.
Link: http://lkml.kernel.org/r/20171003072619.8654-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Alexandru Moise <00moses.alexander00@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Alexandru Moise <00moses.alexander00@gmail.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Remove unused function pgdat_reclaimable_pages() and
node_page_state_snapshot() which becomes unused as well.
Link: http://lkml.kernel.org/r/20171122094416.26019-1-jack@suse.cz
Signed-off-by: Jan Kara <jack@suse.cz>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Shakeel Butt reported he has observed in production systems that the job
loader gets stuck for 10s of seconds while doing a mount operation. It
turns out that it was stuck in register_shrinker() because some
unrelated job was under memory pressure and was spending time in
shrink_slab(). Machines have a lot of shrinkers registered and jobs
under memory pressure have to traverse all of those memcg-aware
shrinkers and affect unrelated jobs which want to register their own
shrinkers.
To solve the issue, this patch simply bails out slab shrinking if it is
found that someone wants to register a shrinker in parallel. A downside
is it could cause unfair shrinking between shrinkers. However, it
should be rare and we can add compilcated logic if we find it's not
enough.
[akpm@linux-foundation.org: tweak code comment]
Link: http://lkml.kernel.org/r/20171115005602.GB23810@bbox
Link: http://lkml.kernel.org/r/1511481899-20335-1-git-send-email-minchan@kernel.org
Signed-off-by: Minchan Kim <minchan@kernel.org>
Signed-off-by: Shakeel Butt <shakeelb@google.com>
Reported-by: Shakeel Butt <shakeelb@google.com>
Tested-by: Shakeel Butt <shakeelb@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Anshuman Khandual <khandual@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
__get_free_pages() will return a virtual address, but it is not just a
32-bit address, for example on a 64-bit system. And this comment really
confuses new readers of mm.
Link: http://lkml.kernel.org/r/1511780964-64864-1-git-send-email-chenjiankang1@huawei.com
Signed-off-by: Jiankang Chen <chenjiankang1@huawei.com>
Reported-by: Hanjun Guo <guohanjun@huawei.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Yisheng Xie <xieyisheng1@huawei.com>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We've seen memory.stat reads in top-level cgroups take up to fourteen
seconds during a userspace bug that created tens of thousands of ghost
cgroups pinned by lingering page cache.
Even with a more reasonable number of cgroups, aggregating memory.stat
is unnecessarily heavy. The complexity is this:
nr_cgroups * nr_stat_items * nr_possible_cpus
where the stat items are ~70 at this point. With 128 cgroups and 128
CPUs - decent, not enormous setups - reading the top-level memory.stat
has to aggregate over a million per-cpu counters. This doesn't scale.
Instead of spreading the source of truth across all CPUs, use the
per-cpu counters merely to batch updates to shared atomic counters.
This is the same as the per-cpu stocks we use for charging memory to the
shared atomic page_counters, and also the way the global vmstat counters
are implemented.
Vmstat has elaborate spilling thresholds that depend on the number of
CPUs, amount of memory, and memory pressure - carefully balancing the
cost of counter updates with the amount of per-cpu error. That's
because the vmstat counters are system-wide, but also used for decisions
inside the kernel (e.g. NR_FREE_PAGES in the allocator). Neither is
true for the memory controller.
Use the same static batch size we already use for page_counter updates
during charging. The per-cpu error in the stats will be 128k, which is
an acceptable ratio of cores to memory accounting granularity.
[hannes@cmpxchg.org: fix warning in __this_cpu_xchg() calls]
Link: http://lkml.kernel.org/r/20171201135750.GB8097@cmpxchg.org
Link: http://lkml.kernel.org/r/20171103153336.24044-3-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Replace all raw 'this_cpu_' modifications of the stat and event per-cpu
counters with API functions such as mod_memcg_state().
This makes the code easier to read, but is also in preparation for the
next patch, which changes the per-cpu implementation of those counters.
Link: http://lkml.kernel.org/r/20171103153336.24044-1-hannes@cmpxchg.org
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Vladimir Davydov <vdavydov.dev@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
in_atomic() has been moved to include/linux/preempt.h, and the filemap.c
doesn't use in_atomic() directly at all, so it sounds unnecessary to
include hardirq.h.
Link: http://lkml.kernel.org/r/1509985319-38633-1-git-send-email-yang.s@alibaba-inc.com
Signed-off-by: Yang Shi <yang.s@alibaba-inc.com>
Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
In deferred_init_range() we initialize struct pages, and also free them
to buddy allocator. We do it in separate loops, because buddy page is
computed ahead, so we do not want to access a struct page that has not
been initialized yet.
There is still, however, a corner case where it is potentially possible
to access uninitialized struct page: this is when buddy page is from the
next memblock range.
This patch fixes this problem by splitting deferred_init_range() into
two functions: one to initialize struct pages, and another to free them.
In addition, this patch brings the following improvements:
- Get rid of __def_free() helper function. And simplifies loop logic by
adding a new pfn validity check function: deferred_pfn_valid().
- Reduces number of variables that we track. So, there is a higher
chance that we will avoid using stack to store/load variables inside
hot loops.
- Enables future multi-threading of these functions: do initialization
in multiple threads, wait for all threads to finish, do freeing part
in multithreading.
Tested on x86 with 1T of memory to make sure no regressions are
introduced.
[akpm@linux-foundation.org: fix spello in comment]
Link: http://lkml.kernel.org/r/20171107150446.32055-2-pasha.tatashin@oracle.com
Signed-off-by: Pavel Tatashin <pasha.tatashin@oracle.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: Steven Sistare <steven.sistare@oracle.com>
Cc: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Previously we were using the ratio of the number of lru pages scanned to
the number of eligible lru pages to determine the number of slab objects
to scan. The problem with this is that these two things have nothing to
do with each other, so in slab heavy work loads where there is little to
no page cache we can end up with the pages scanned being a very low
number. This means that we reclaim next to no slab pages and waste a
lot of time reclaiming small amounts of space.
Consider the following scenario, where we have the following values and
the rest of the memory usage is in slab
Active: 58840 kB
Inactive: 46860 kB
Every time we do a get_scan_count() we do this
scan = size >> sc->priority
where sc->priority starts at DEF_PRIORITY, which is 12. The first loop
through reclaim would result in a scan target of 2 pages to 11715 total
inactive pages, and 3 pages to 14710 total active pages. This is a
really really small target for a system that is entirely slab pages.
And this is super optimistic, this assumes we even get to scan these
pages. We don't increment sc->nr_scanned unless we 1) isolate the page,
which assumes it's not in use, and 2) can lock the page. Under pressure
these numbers could probably go down, I'm sure there's some random pages
from daemons that aren't actually in use, so the targets get even
smaller.
Instead use sc->priority in the same way we use it to determine scan
amounts for the lru's. This generally equates to pages. Consider the
following
slab_pages = (nr_objects * object_size) / PAGE_SIZE
What we would like to do is
scan = slab_pages >> sc->priority
but we don't know the number of slab pages each shrinker controls, only
the objects. However say that theoretically we knew how many pages a
shrinker controlled, we'd still have to convert this to objects, which
would look like the following
scan = shrinker_pages >> sc->priority
scan_objects = (PAGE_SIZE / object_size) * scan
or written another way
scan_objects = (shrinker_pages >> sc->priority) *
(PAGE_SIZE / object_size)
which can thus be written
scan_objects = ((shrinker_pages * PAGE_SIZE) / object_size) >>
sc->priority
which is just
scan_objects = nr_objects >> sc->priority
We don't need to know exactly how many pages each shrinker represents,
it's objects are all the information we need. Making this change allows
us to place an appropriate amount of pressure on the shrinker pools for
their relative size.
Link: http://lkml.kernel.org/r/1510780549-6812-1-git-send-email-josef@toxicpanda.com
Signed-off-by: Josef Bacik <jbacik@fb.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Dave Chinner <david@fromorbit.com>
Acked-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Michal Hocko <mhocko@kernel.org>
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>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently we display some hugepage statistics (total, free, etc) in
/proc/meminfo, but only for default hugepage size (e.g. 2Mb).
If hugepages of different sizes are used (like 2Mb and 1Gb on x86-64),
/proc/meminfo output can be confusing, as non-default sized hugepages
are not reflected at all, and there are no signs that they are existing
and consuming system memory.
To solve this problem, let's display the total amount of memory,
consumed by hugetlb pages of all sized (both free and used). Let's call
it "Hugetlb", and display size in kB to match generic /proc/meminfo
style.
For example, (1024 2Mb pages and 2 1Gb pages are pre-allocated):
$ cat /proc/meminfo
MemTotal: 8168984 kB
MemFree: 3789276 kB
<...>
CmaFree: 0 kB
HugePages_Total: 1024
HugePages_Free: 1024
HugePages_Rsvd: 0
HugePages_Surp: 0
Hugepagesize: 2048 kB
Hugetlb: 4194304 kB
DirectMap4k: 32632 kB
DirectMap2M: 4161536 kB
DirectMap1G: 6291456 kB
Also, this patch updates corresponding docs to reflect Hugetlb entry
meaning and difference between Hugetlb and HugePages_Total * Hugepagesize.
Link: http://lkml.kernel.org/r/20171115231409.12131-1-guro@fb.com
Signed-off-by: Roman Gushchin <guro@fb.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Mike Kravetz <mike.kravetz@oracle.com>
Cc: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Dave Hansen <dave.hansen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pulling cpu hotplug locks inside the mm core function like
lru_add_drain_all just asks for problems and the recent lockdep splat
[1] just proves this. While the usage in that particular case might be
wrong we should avoid the locking as lru_add_drain_all() is used in many
places. It seems that this is not all that hard to achieve actually.
We have done the same thing for drain_all_pages which is analogous by
commit a459eeb7b8 ("mm, page_alloc: do not depend on cpu hotplug locks
inside the allocator"). All we have to care about is to handle
- the work item might be executed on a different cpu in worker from
unbound pool so it doesn't run on pinned on the cpu
- we have to make sure that we do not race with page_alloc_cpu_dead
calling lru_add_drain_cpu
the first part is already handled because the worker calls lru_add_drain
which disables preemption when calling lru_add_drain_cpu on the local
cpu it is draining. The later is true because page_alloc_cpu_dead is
called on the controlling CPU after the hotplugged CPU vanished
completely.
[1] http://lkml.kernel.org/r/089e0825eec8955c1f055c83d476@google.com
[add a cpu hotplug locking interaction as per tglx]
Link: http://lkml.kernel.org/r/20171116120535.23765-1-mhocko@kernel.org
Signed-off-by: Michal Hocko <mhocko@suse.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Tejun Heo <tj@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
As in manpage of migrate_pages, the errno should be set to EINVAL when
none of the node IDs specified by new_nodes are on-line and allowed by
the process's current cpuset context, or none of the specified nodes
contain memory. However, when test by following case:
new_nodes = 0;
old_nodes = 0xf;
ret = migrate_pages(pid, old_nodes, new_nodes, MAX);
The ret will be 0 and no errno is set. As the new_nodes is empty, we
should expect EINVAL as documented.
To fix the case like above, this patch check whether target nodes AND
current task_nodes is empty, and then check whether AND
node_states[N_MEMORY] is empty.
Link: http://lkml.kernel.org/r/1510882624-44342-4-git-send-email-xieyisheng1@huawei.com
Signed-off-by: Yisheng Xie <xieyisheng1@huawei.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: Andi Kleen <ak@linux.intel.com>
Cc: Chris Salls <salls@cs.ucsb.edu>
Cc: Christopher Lameter <cl@linux.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: Tan Xiaojun <tanxiaojun@huawei.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
