shmem: split huge pages beyond i_size under memory pressure
Even if user asked to allocate huge pages always (huge=always), we
should be able to free up some memory by splitting pages which are
partly byound i_size if memory presure comes or once we hit limit on
filesystem size (-o size=).
In order to do this we maintain per-superblock list of inodes, which
potentially have huge pages on the border of file size.
Per-fs shrinker can reclaim memory by splitting such pages.
If we hit -ENOSPC during shmem_getpage_gfp(), we try to split a page to
free up space on the filesystem and retry allocation if it succeed.
For file mappings, we don't deposit page tables on THP allocation
because it's not strictly required to implement split_huge_pmd(): we can
just clear pmd and let following page faults to reconstruct the page
table.
But Power makes use of deposited page table to address MMU quirk.
Let's hide THP page cache, including huge tmpfs, under separate config
option, so it can be forbidden on Power.
We can revert the patch later once solution for Power found.
khugepaged: add support of collapse for tmpfs/shmem pages
This patch extends khugepaged to support collapse of tmpfs/shmem pages.
We share fair amount of infrastructure with anon-THP collapse.
Few design points:
- First we are looking for VMA which can be suitable for mapping huge
page;
- If the VMA maps shmem file, the rest scan/collapse operations
operates on page cache, not on page tables as in anon VMA case.
- khugepaged_scan_shmem() finds a range which is suitable for huge
page. The scan is lockless and shouldn't disturb system too much.
- once the candidate for collapse is found, collapse_shmem() attempts
to create a huge page:
+ scan over radix tree, making the range point to new huge page;
+ new huge page is not-uptodate, locked and freezed (refcount
is 0), so nobody can touch them until we say so.
+ we swap in pages during the scan. khugepaged_scan_shmem()
filters out ranges with more than khugepaged_max_ptes_swap
swapped out pages. It's HPAGE_PMD_NR/8 by default.
+ old pages are isolated, unmapped and put to local list in case
to be restored back if collapse failed.
- if collapse succeed, we retract pte page tables from VMAs where huge
pages mapping is possible. The huge page will be mapped as PMD on
next minor fault into the range.
shmem, thp: respect MADV_{NO,}HUGEPAGE for file mappings
Let's wire up existing madvise() hugepage hints for file mappings.
MADV_HUGEPAGE advise shmem to allocate huge page on page fault in the
VMA. It only has effect if the filesystem is mounted with huge=advise
or huge=within_size.
MADV_NOHUGEPAGE prevents hugepage from being allocated on page fault in
the VMA. It doesn't prevent a huge page from being allocated by other
means, i.e. page fault into different mapping or write(2) into file.
Here's basic implementation of huge pages support for shmem/tmpfs.
It's all pretty streight-forward:
- shmem_getpage() allcoates huge page if it can and try to inserd into
radix tree with shmem_add_to_page_cache();
- shmem_add_to_page_cache() puts the page onto radix-tree if there's
space for it;
- shmem_undo_range() removes huge pages, if it fully within range.
Partial truncate of huge pages zero out this part of THP.
This have visible effect on fallocate(FALLOC_FL_PUNCH_HOLE)
behaviour. As we don't really create hole in this case,
lseek(SEEK_HOLE) may have inconsistent results depending what
pages happened to be allocated.
- no need to change shmem_fault: core-mm will map an compound page as
huge if VMA is suitable;
Provide a shmem_get_unmapped_area method in file_operations, called at
mmap time to decide the mapping address. It could be conditional on
CONFIG_TRANSPARENT_HUGEPAGE, but save #ifdefs in other places by making
it unconditional.
shmem_get_unmapped_area() first calls the usual mm->get_unmapped_area
(which we treat as a black box, highly dependent on architecture and
config and executable layout). Lots of conditions, and in most cases it
just goes with the address that chose; but when our huge stars are
rightly aligned, yet that did not provide a suitable address, go back to
ask for a larger arena, within which to align the mapping suitably.
There have to be some direct calls to shmem_get_unmapped_area(), not via
the file_operations: because of the way shmem_zero_setup() is called to
create a shmem object late in the mmap sequence, when MAP_SHARED is
requested with MAP_ANONYMOUS or /dev/zero. Though this only matters
when /proc/sys/vm/shmem_huge has been set.
This patch adds new mount option "huge=". It can have following values:
- "always":
Attempt to allocate huge pages every time we need a new page;
- "never":
Do not allocate huge pages;
- "within_size":
Only allocate huge page if it will be fully within i_size.
Also respect fadvise()/madvise() hints;
- "advise:
Only allocate huge pages if requested with fadvise()/madvise();
Default is "never" for now.
"mount -o remount,huge= /mountpoint" works fine after mount: remounting
huge=never will not attempt to break up huge pages at all, just stop
more from being allocated.
No new config option: put this under CONFIG_TRANSPARENT_HUGEPAGE, which
is the appropriate option to protect those who don't want the new bloat,
and with which we shall share some pmd code.
Prohibit the option when !CONFIG_TRANSPARENT_HUGEPAGE, just as mpol is
invalid without CONFIG_NUMA (was hidden in mpol_parse_str(): make it
explicit).
Allow enabling THP only if the machine has_transparent_hugepage().
But what about Shmem with no user-visible mount? SysV SHM, memfds,
shared anonymous mmaps (of /dev/zero or MAP_ANONYMOUS), GPU drivers' DRM
objects, Ashmem. Though unlikely to suit all usages, provide sysfs knob
/sys/kernel/mm/transparent_hugepage/shmem_enabled to experiment with
huge on those.
And allow shmem_enabled two further values:
- "deny":
For use in emergencies, to force the huge option off from
all mounts;
- "force":
Force the huge option on for all - very useful for testing;
filemap: prepare find and delete operations for huge pages
For now, we would have HPAGE_PMD_NR entries in radix tree for every huge
page. That's suboptimal and it will be changed to use Matthew's
multi-order entries later.
'add' operation is not changed, because we don't need it to implement
hugetmpfs: shmem uses its own implementation.
The new helper is similar to radix_tree_maybe_preload(), but tries to
preload number of nodes required to insert (1 << order) continuous
naturally-aligned elements.
This is required to push huge pages into pagecache.
thp, mlock: do not mlock PTE-mapped file huge pages
As with anon THP, we only mlock file huge pages if we can prove that the
page is not mapped with PTE. This way we can avoid mlock leak into
non-mlocked vma on split.
We rely on PageDoubleMap() under lock_page() to check if the the page
may be PTE mapped. PG_double_map is set by page_add_file_rmap() when
the page mapped with PTEs.
- File pages are on radix-tree, so we have head->_count offset by
HPAGE_PMD_NR. The count got distributed to small pages during split.
- mapping->tree_lock prevents non-lockless access to pages under split
over radix-tree;
- Lockless access is prevented by setting the head->_count to 0 during
split;
- After split, some pages can be beyond i_size. We drop them from
radix-tree.
- We don't setup migration entries. Just unmap pages. It helps
handling cases when i_size is in the middle of the page: no need
handle unmap pages beyond i_size manually.
thp: run vma_adjust_trans_huge() outside i_mmap_rwsem
vma_addjust_trans_huge() splits pmd if it's crossing VMA boundary.
During split we munlock the huge page which requires rmap walk. rmap
wants to take the lock on its own.
Let's move vma_adjust_trans_huge() outside i_mmap_rwsem to fix this.
change_huge_pmd() has assert which is not relvant for file page. For
shared mapping it's perfectly fine to have page table entry writable,
without explicit mkwrite.
copy_page_range() has a check for "Don't copy ptes where a page fault
will fill them correctly." It works on VMA level. We still copy all
page table entries from private mappings, even if they map page cache.
We can simplify copy_huge_pmd() a bit by skipping file PMDs.
We don't map file private pages with PMDs, so they only can map page
cache. It's safe to skip them as they can be re-faulted later.
Splitting THP PMD is simple: just unmap it as in DAX case. This way we
can avoid memory overhead on page table allocation to deposit.
It's probably a good idea to try to allocation page table with
GFP_ATOMIC in __split_huge_pmd_locked() to avoid refaulting the area,
but clearing pmd should be good enough for now.
Unlike DAX, we also remove the page from rmap and drop reference.
pmd_young() is transfered to PageReferenced().
split_huge_pmd() for file mappings (and DAX too) is implemented by just
clearing pmd entry as we can re-fill this area from page cache on pte
level later.
This means we don't need deposit page tables when file THP is mapped.
Therefore we shouldn't try to withdraw a page table on zap_huge_pmd()
file THP PMD.
Naive approach: on mapping/unmapping the page as compound we update
->_mapcount on each 4k page. That's not efficient, but it's not obvious
how we can optimize this. We can look into optimization later.
PG_double_map optimization doesn't work for file pages since lifecycle
of file pages is different comparing to anon pages: file page can be
mapped again at any time.
mm: postpone page table allocation until we have page to map
The idea (and most of code) is borrowed again: from Hugh's patchset on
huge tmpfs[1].
Instead of allocation pte page table upfront, we postpone this until we
have page to map in hands. This approach opens possibility to map the
page as huge if filesystem supports this.
Comparing to Hugh's patch I've pushed page table allocation a bit
further: into do_set_pte(). This way we can postpone allocation even in
faultaround case without moving do_fault_around() after __do_fault().
do_set_pte() got renamed to alloc_set_pte() as it can allocate page
table if required.
The idea borrowed from Peter's patch from patchset on speculative page
faults[1]:
Instead of passing around the endless list of function arguments,
replace the lot with a single structure so we can change context without
endless function signature changes.
The changes are mostly mechanical with exception of faultaround code:
filemap_map_pages() got reworked a bit.
mm, thp: fix locking inconsistency in collapse_huge_page
After creating revalidate vma function, locking inconsistency occured
due to directing the code path to wrong label. This patch directs to
correct label and fix the inconsistency.
mm, thp: make swapin readahead under down_read of mmap_sem
Currently khugepaged makes swapin readahead under down_write. This
patch supplies to make swapin readahead under down_read instead of
down_write.
The patch was tested with a test program that allocates 800MB of memory,
writes to it, and then sleeps. The system was forced to swap out all.
Afterwards, the test program touches the area by writing, it skips a
page in each 20 pages of the area.
mm: make swapin readahead to improve thp collapse rate
This patch makes swapin readahead to improve thp collapse rate. When
khugepaged scanned pages, there can be a few of the pages in swap area.
With the patch THP can collapse 4kB pages into a THP when there are up
to max_ptes_swap swap ptes in a 2MB range.
The patch was tested with a test program that allocates 400B of memory,
writes to it, and then sleeps. I force the system to swap out all.
Afterwards, the test program touches the area by writing, it skips a
page in each 20 pages of the area.
Without the patch, system did not swap in readahead. THP rate was %65
of the program of the memory, it did not change over time.
With this patch, after 10 minutes of waiting khugepaged had collapsed
%99 of the program's memory.
[kirill.shutemov@linux.intel.com: trivial cleanup of exit path of the function]
[kirill.shutemov@linux.intel.com: __collapse_huge_page_swapin(): drop unused 'pte' parameter]
[kirill.shutemov@linux.intel.com: do not hold anon_vma lock during swap in] Signed-off-by: Ebru Akagunduz <ebru.akagunduz@gmail.com> Acked-by: Rik van Riel <riel@redhat.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Xie XiuQi <xiexiuqi@huawei.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: David Rientjes <rientjes@google.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Hugh Dickins <hughd@google.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Signed-off-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>
Introduce a new sysfs integer knob
/sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_swap which makes
optimistic check for swapin readahead to increase thp collapse rate.
Before getting swapped out pages to memory, checks them and allows up to a
certain number. It also prints out using tracepoints amount of unmapped
ptes.
[vdavydov@parallels.com: fix scan not aborted on SCAN_EXCEED_SWAP_PTE]
[sfr@canb.auug.org.au: build fix] Link: http://lkml.kernel.org/r/20160616154503.65806e12@canb.auug.org.au Signed-off-by: Ebru Akagunduz <ebru.akagunduz@gmail.com> Acked-by: Rik van Riel <riel@redhat.com> Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Cc: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Xie XiuQi <xiexiuqi@huawei.com> Cc: Cyrill Gorcunov <gorcunov@openvz.org> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: David Rientjes <rientjes@google.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Hugh Dickins <hughd@google.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@suse.cz> Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Michal Hocko [Tue, 26 Jul 2016 22:24:53 +0000 (15:24 -0700)]
mm, memcg: use consistent gfp flags during readahead
Vladimir has noticed that we might declare memcg oom even during
readahead because read_pages only uses GFP_KERNEL (with mapping_gfp
restriction) while __do_page_cache_readahead uses
page_cache_alloc_readahead which adds __GFP_NORETRY to prevent from
OOMs. This gfp mask discrepancy is really unfortunate and easily
fixable. Drop page_cache_alloc_readahead() which only has one user and
outsource the gfp_mask logic into readahead_gfp_mask and propagate this
mask from __do_page_cache_readahead down to read_pages.
This alone would have only very limited impact as most filesystems are
implementing ->readpages and the common implementation mpage_readpages
does GFP_KERNEL (with mapping_gfp restriction) again. We can tell it to
use readahead_gfp_mask instead as this function is called only during
readahead as well. The same applies to read_cache_pages.
ext4 has its own ext4_mpage_readpages but the path which has pages !=
NULL can use the same gfp mask. Btrfs, cifs, f2fs and orangefs are
doing a very similar pattern to mpage_readpages so the same can be
applied to them as well.
[akpm@linux-foundation.org: coding-style fixes]
[mhocko@suse.com: restrict gfp mask in mpage_alloc] Link: http://lkml.kernel.org/r/20160610074223.GC32285@dhcp22.suse.cz Link: http://lkml.kernel.org/r/1465301556-26431-1-git-send-email-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Cc: Vladimir Davydov <vdavydov@parallels.com> Cc: Chris Mason <clm@fb.com> Cc: Steve French <sfrench@samba.org> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Jan Kara <jack@suse.cz> Cc: Mike Marshall <hubcap@omnibond.com> Cc: Jaegeuk Kim <jaegeuk@kernel.org> Cc: Changman Lee <cm224.lee@samsung.com> Cc: Chao Yu <yuchao0@huawei.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
the final __mmput will be executed in the delayed context which might
happen far in the future. Such a race is highly unlikely because the
write holder of mmap_sem would have to be an external task (all direct
holders are already killed or exiting) and it usually have to pin
mm_users in order to do anything reasonable.
We can, however, make sure that the mmput_async is only called when we
do not back off and reap some memory. That would reduce the impact of
the delayed __mmput because the real content would be already freed.
Pin mm_count to keep it alive after we drop task_lock and before we try
to get mmap_sem. If the mmap_sem succeeds we can try to grab mm_users
reference and then go on with unmapping the address space.
It is not clear whether this race is possible at all but it is better to
be more robust and do not pin mm_users unless we are sure we are
actually doing some real work during __oom_reap_task.
Link: http://lkml.kernel.org/r/1465306987-30297-1-git-send-email-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Chen Gang [Tue, 26 Jul 2016 22:24:47 +0000 (15:24 -0700)]
include/linux/memblock.h: Clean up code for several trivial details
Correct the function parameters alignment, since original code already
use both tabs and white spaces together for the incorrect parameters
alignment functions.
If one line can hold one statement within 80 columns, let it in one line
(original code did not consider about the tabs/spaces for 2nd line when
a statement is separated into 2 lines).
Try to let '' aligned within one macro, since all related lines are
short enough.
Remove useless statement "idx = 0;", and always assign rgn within the
'for' statement.
Minchan Kim [Tue, 26 Jul 2016 22:24:45 +0000 (15:24 -0700)]
mm: add NR_ZSMALLOC to vmstat
zram is very popular for some of the embedded world (e.g., TV, mobile
phones). On those system, zsmalloc's consumed memory size is never
trivial (one of example from real product system, total memory: 800M,
zsmalloc consumed: 150M), so we have used this out of tree patch to
monitor system memory behavior via /proc/vmstat.
With zsmalloc in vmstat, it helps in tracking down system behavior due
to memory usage.
[minchan@kernel.org: zsmalloc: follow up zsmalloc vmstat] Link: http://lkml.kernel.org/r/20160607091737.GC23435@bbox
[akpm@linux-foundation.org: fix build with CONFIG_ZSMALLOC=m] Link: http://lkml.kernel.org/r/1464919731-13255-1-git-send-email-minchan@kernel.org Signed-off-by: Minchan Kim <minchan@kernel.org> Cc: Sangseok Lee <sangseok.lee@lge.com> Cc: Chanho Min <chanho.min@lge.com> Cc: Chan Gyun Jeong <chan.jeong@lge.com> Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm, frontswap: convert frontswap_enabled to static key
I have noticed that frontswap.h first declares "frontswap_enabled" as
extern bool variable, and then overrides it with "#define
frontswap_enabled (1)" for CONFIG_FRONTSWAP=Y or (0) when disabled. The
bool variable isn't actually instantiated anywhere.
This all looks like an unfinished attempt to make frontswap_enabled
reflect whether a backend is instantiated. But in the current state,
all frontswap hooks call unconditionally into frontswap.c just to check
if frontswap_ops is non-NULL. This should at least be checked inline,
but we can further eliminate the overhead when CONFIG_FRONTSWAP is
enabled and no backend registered, using a static key that is initially
disabled, and gets enabled only upon first backend registration.
Thus, checks for "frontswap_enabled" are replaced with
"frontswap_enabled()" wrapping the static key check. There are two
exceptions:
- xen's selfballoon_process() was testing frontswap_enabled in code guarded
by #ifdef CONFIG_FRONTSWAP, which was effectively always true when reachable.
The patch just removes this check. Using frontswap_enabled() does not sound
correct here, as this can be true even without xen's own backend being
registered.
- in SYSCALL_DEFINE2(swapon), change the check to IS_ENABLED(CONFIG_FRONTSWAP)
as it seems the bitmap allocation cannot currently be postponed until a
backend is registered. This means that frontswap will still have some
memory overhead by being configured, but without a backend.
After the patch, we can expect that some functions in frontswap.c are
called only when frontswap_ops is non-NULL. Change the checks there to
VM_BUG_ONs. While at it, convert other BUG_ONs to VM_BUG_ONs as
frontswap has been stable for some time.
[akpm@linux-foundation.org: coding-style fixes] Link: http://lkml.kernel.org/r/1463152235-9717-1-git-send-email-vbabka@suse.cz Signed-off-by: Vlastimil Babka <vbabka@suse.cz> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Cc: Boris Ostrovsky <boris.ostrovsky@oracle.com> Cc: David Vrabel <david.vrabel@citrix.com> Cc: Juergen Gross <jgross@suse.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vladimir Davydov [Tue, 26 Jul 2016 22:24:36 +0000 (15:24 -0700)]
af_unix: charge buffers to kmemcg
Unix sockets can consume a significant amount of system memory, hence
they should be accounted to kmemcg.
Since unix socket buffers are always allocated from process context, all
we need to do to charge them to kmemcg is set __GFP_ACCOUNT in
sock->sk_allocation mask.
Eric asked:
> 1) What happens when a buffer, allocated from socket <A> lands in a
> different socket <B>, maybe owned by another user/process.
>
> Who owns it now, in term of kmemcg accounting ?
We never move memcg charges. E.g. if two processes from different
cgroups are sharing a memory region, each page will be charged to the
process which touched it first. Or if two processes are working with
the same directory tree, inodes and dentries will be charged to the
first user. The same is fair for unix socket buffers - they will be
charged to the sender.
> 2) Has performance impact been evaluated ?
I ran netperf STREAM_STREAM with default options in a kmemcg on a 4 core
x2 HT box. The results are below:
So the accounting doesn't come for free - it takes ~4% of performance.
I believe we could optimize it by using per cpu batching not only on
charge, but also on uncharge in memcg core, but that's beyond the scope
of this patch set - I'll take a look at this later.
Anyway, if performance impact is found to be unacceptable, it is always
possible to disable kmem accounting at boot time (cgroup.memory=nokmem)
or not use memory cgroups at runtime at all (thanks to jump labels
there'll be no overhead even if they are compiled in).
Link: http://lkml.kernel.org/r/fcfe6cae27a59fbc5e40145664b3cf085a560c68.1464079538.git.vdavydov@virtuozzo.com Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vladimir Davydov [Tue, 26 Jul 2016 22:24:33 +0000 (15:24 -0700)]
pipe: account to kmemcg
Pipes can consume a significant amount of system memory, hence they
should be accounted to kmemcg.
This patch marks pipe_inode_info and anonymous pipe buffer page
allocations as __GFP_ACCOUNT so that they would be charged to kmemcg.
Note, since a pipe buffer page can be "stolen" and get reused for other
purposes, including mapping to userspace, we clear PageKmemcg thus
resetting page->_mapcount and uncharge it in anon_pipe_buf_steal, which
is introduced by this patch.
A note regarding anon_pipe_buf_steal implementation. We allow to steal
the page if its ref count equals 1. It looks racy, but it is correct
for anonymous pipe buffer pages, because:
- We lock out all other pipe users, because ->steal is called with
pipe_lock held, so the page can't be spliced to another pipe from
under us.
- The page is not on LRU and it never was.
- Thus a parallel thread can access it only by PFN. Although this is
quite possible (e.g. see page_idle_get_page and balloon_page_isolate)
this is not dangerous, because all such functions do is increase page
ref count, check if the page is the one they are looking for, and
decrease ref count if it isn't. Since our page is clean except for
PageKmemcg mark, which doesn't conflict with other _mapcount users,
the worst that can happen is we see page_count > 2 due to a transient
ref, in which case we false-positively abort ->steal, which is still
fine, because ->steal is not guaranteed to succeed.
Link: http://lkml.kernel.org/r/20160527150313.GD26059@esperanza Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Alexander Viro <viro@zeniv.linux.org.uk> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vladimir Davydov [Tue, 26 Jul 2016 22:24:30 +0000 (15:24 -0700)]
arch: x86: charge page tables to kmemcg
Page tables can bite a relatively big chunk off system memory and their
allocations are easy to trigger from userspace, so they should be
accounted to kmemcg.
This patch marks page table allocations as __GFP_ACCOUNT for x86. Note
we must not charge allocations of kernel page tables, because they can
be shared among processes from different cgroups so accounting them to a
particular one can pin other cgroups for indefinitely long. So we clear
__GFP_ACCOUNT flag if a page table is allocated for the kernel.
Link: http://lkml.kernel.org/r/7d5c54f6a2bcbe76f03171689440003d87e6c742.1464079538.git.vdavydov@virtuozzo.com Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@redhat.com> Cc: "H. Peter Anvin" <hpa@zytor.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vladimir Davydov [Tue, 26 Jul 2016 22:24:27 +0000 (15:24 -0700)]
mm: memcontrol: teach uncharge_list to deal with kmem pages
Page table pages are batched-freed in release_pages on most
architectures. If we want to charge them to kmemcg (this is what is
done later in this series), we need to teach mem_cgroup_uncharge_list to
handle kmem pages.
Link: http://lkml.kernel.org/r/18d5c09e97f80074ed25b97a7d0f32b95d875717.1464079538.git.vdavydov@virtuozzo.com Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vladimir Davydov [Tue, 26 Jul 2016 22:24:24 +0000 (15:24 -0700)]
mm: charge/uncharge kmemcg from generic page allocator paths
Currently, to charge a non-slab allocation to kmemcg one has to use
alloc_kmem_pages helper with __GFP_ACCOUNT flag. A page allocated with
this helper should finally be freed using free_kmem_pages, otherwise it
won't be uncharged.
This API suits its current users fine, but it turns out to be impossible
to use along with page reference counting, i.e. when an allocation is
supposed to be freed with put_page, as it is the case with pipe or unix
socket buffers.
To overcome this limitation, this patch moves charging/uncharging to
generic page allocator paths, i.e. to __alloc_pages_nodemask and
free_pages_prepare, and zaps alloc/free_kmem_pages helpers. This way,
one can use any of the available page allocation functions to get the
allocated page charged to kmemcg - it's enough to pass __GFP_ACCOUNT,
just like in case of kmalloc and friends. A charged page will be
automatically uncharged on free.
To make it possible, we need to mark pages charged to kmemcg somehow.
To avoid introducing a new page flag, we make use of page->_mapcount for
marking such pages. Since pages charged to kmemcg are not supposed to
be mapped to userspace, it should work just fine. There are other
(ab)users of page->_mapcount - buddy and balloon pages - but we don't
conflict with them.
In case kmemcg is compiled out or not used at runtime, this patch
introduces no overhead to generic page allocator paths. If kmemcg is
used, it will be plus one gfp flags check on alloc and plus one
page->_mapcount check on free, which shouldn't hurt performance, because
the data accessed are hot.
Link: http://lkml.kernel.org/r/a9736d856f895bcb465d9f257b54efe32eda6f99.1464079538.git.vdavydov@virtuozzo.com Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vladimir Davydov [Tue, 26 Jul 2016 22:24:21 +0000 (15:24 -0700)]
mm: memcontrol: cleanup kmem charge functions
- Handle memcg_kmem_enabled check out to the caller. This reduces the
number of function definitions making the code easier to follow. At
the same time it doesn't result in code bloat, because all of these
functions are used only in one or two places.
- Move __GFP_ACCOUNT check to the caller as well so that one wouldn't
have to dive deep into memcg implementation to see which allocations
are charged and which are not.
- Refresh comments.
Link: http://lkml.kernel.org/r/52882a28b542c1979fd9a033b4dc8637fc347399.1464079537.git.vdavydov@virtuozzo.com Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vladimir Davydov [Tue, 26 Jul 2016 22:24:18 +0000 (15:24 -0700)]
mm: clean up non-standard page->_mapcount users
- Add a proper comment to page->_mapcount.
- Introduce a macro for generating helper functions.
- Place all special page->_mapcount values next to each other so that
readers can see all possible values and so we don't get duplicates.
Link: http://lkml.kernel.org/r/502f49000e0b63e6c62e338fac6b420bf34fb526.1464079537.git.vdavydov@virtuozzo.com Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Vladimir Davydov [Tue, 26 Jul 2016 22:24:16 +0000 (15:24 -0700)]
mm: remove pointless struct in struct page definition
This patchset implements per kmemcg accounting of page tables
(x86-only), pipe buffers, and unix socket buffers.
Patches 1-3 are just cleanups that are not supposed to introduce any
functional changes. Patches 4 and 5 move charge/uncharge to generic
page allocator paths for the sake of accounting pipe and unix socket
buffers. Patches 5-7 make x86 page tables, pipe buffers, and unix
socket buffers accountable.
This patch (of 8):
... to reduce indentation level thus leaving more space for comments.
Link: http://lkml.kernel.org/r/f34ffe70fce2b0b9220856437f77972d67c14275.1464079537.git.vdavydov@virtuozzo.com Signed-off-by: Vladimir Davydov <vdavydov@virtuozzo.com> Cc: Johannes Weiner <hannes@cmpxchg.org> Cc: Michal Hocko <mhocko@kernel.org> Cc: Eric Dumazet <eric.dumazet@gmail.com> Cc: Minchan Kim <minchan@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This updates the generic and arch specific implementation to return true
if we need to do a tlb flush. That means if a __tlb_remove_page
indicate a flush is needed, the page we try to remove need to be tracked
and added again after the flush. We need to track it because we have
already update the pte to none and we can't just loop back.
This change is done to enable us to do a tlb_flush when we try to flush
a range that consists of different page sizes. For architectures like
ppc64, we can do a range based tlb flush and we need to track page size
for that. When we try to remove a huge page, we will force a tlb flush
and starts a new mmu gather.
[aneesh.kumar@linux.vnet.ibm.com: mm-change-the-interface-for-__tlb_remove_page-v3] Link: http://lkml.kernel.org/r/1465049193-22197-2-git-send-email-aneesh.kumar@linux.vnet.ibm.com Link: http://lkml.kernel.org/r/1464860389-29019-2-git-send-email-aneesh.kumar@linux.vnet.ibm.com Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Hugh Dickins <hughd@google.com> Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com> Cc: Andrea Arcangeli <aarcange@redhat.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mgorman@suse.de> Cc: David Rientjes <rientjes@google.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Minchan Kim <minchan.kim@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
For hugetlb like THP (and unlike regular page), we do tlb flush after
dropping ptl. Because of the above, we don't need to track force_flush
like we do now. Instead we can simply call tlb_remove_page() which will
do the flush if needed.
No functionality change in this patch.
Link: http://lkml.kernel.org/r/1465049193-22197-1-git-send-email-aneesh.kumar@linux.vnet.ibm.com Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: "Kirill A. Shutemov" <kirill@shutemov.name> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm: thp: check pmd_trans_unstable() after split_huge_pmd()
split_huge_pmd() doesn't guarantee that the pmd is normal pmd pointing
to pte entries, which can be checked with pmd_trans_unstable(). Some
callers make this assertion and some do it differently and some not, so
let's do it in a unified manner.
Joonsoo Kim [Tue, 26 Jul 2016 22:24:01 +0000 (15:24 -0700)]
mm/page_isolation: clean up confused code
When there is an isolated_page, post_alloc_hook() is called with page
but __free_pages() is called with isolated_page. Since they are the
same so no problem but it's very confusing. To reduce it, this patch
changes isolated_page to boolean type and uses page variable
consistently.
Joonsoo Kim [Tue, 26 Jul 2016 22:23:58 +0000 (15:23 -0700)]
mm/page_alloc: introduce post allocation processing on page allocator
This patch is motivated from Hugh and Vlastimil's concern [1].
There are two ways to get freepage from the allocator. One is using
normal memory allocation API and the other is __isolate_free_page()
which is internally used for compaction and pageblock isolation. Later
usage is rather tricky since it doesn't do whole post allocation
processing done by normal API.
One problematic thing I already know is that poisoned page would not be
checked if it is allocated by __isolate_free_page(). Perhaps, there
would be more.
We could add more debug logic for allocated page in the future and this
separation would cause more problem. I'd like to fix this situation at
this time. Solution is simple. This patch commonize some logic for
newly allocated page and uses it on all sites. This will solve the
problem.
Joonsoo Kim [Tue, 26 Jul 2016 22:23:55 +0000 (15:23 -0700)]
mm/page_owner: use stackdepot to store stacktrace
Currently, we store each page's allocation stacktrace on corresponding
page_ext structure and it requires a lot of memory. This causes the
problem that memory tight system doesn't work well if page_owner is
enabled. Moreover, even with this large memory consumption, we cannot
get full stacktrace because we allocate memory at boot time and just
maintain 8 stacktrace slots to balance memory consumption. We could
increase it to more but it would make system unusable or change system
behaviour.
To solve the problem, this patch uses stackdepot to store stacktrace.
It obviously provides memory saving but there is a drawback that
stackdepot could fail.
stackdepot allocates memory at runtime so it could fail if system has
not enough memory. But, most of allocation stack are generated at very
early time and there are much memory at this time. So, failure would
not happen easily. And, one failure means that we miss just one page's
allocation stacktrace so it would not be a big problem. In this patch,
when memory allocation failure happens, we store special stracktrace
handle to the page that is failed to save stacktrace. With it, user can
guess memory usage properly even if failure happens.
Memory saving looks as following. (4GB memory system with page_owner)
(before the patch -> after the patch)
Note that implementation looks complex than someone would imagine
because there is recursion issue. stackdepot uses page allocator and
page_owner is called at page allocation. Using stackdepot in page_owner
could re-call page allcator and then page_owner. That is a recursion.
To detect and avoid it, whenever we obtain stacktrace, recursion is
checked and page_owner is set to dummy information if found. Dummy
information means that this page is allocated for page_owner feature
itself (such as stackdepot) and it's understandable behavior for user.
Joonsoo Kim [Tue, 26 Jul 2016 22:23:49 +0000 (15:23 -0700)]
mm/page_owner: introduce split_page_owner and replace manual handling
split_page() calls set_page_owner() to set up page_owner to each pages.
But, it has a drawback that head page and the others have different
stacktrace because callsite of set_page_owner() is slightly differnt.
To avoid this problem, this patch copies head page's page_owner to the
others. It needs to introduce new function, split_page_owner() but it
also remove the other function, get_page_owner_gfp() so looks good to
do.
Link: http://lkml.kernel.org/r/1464230275-25791-4-git-send-email-iamjoonsoo.kim@lge.com Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Minchan Kim <minchan@kernel.org> Cc: Alexander Potapenko <glider@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Joonsoo Kim [Tue, 26 Jul 2016 22:23:46 +0000 (15:23 -0700)]
mm/page_owner: copy last_migrate_reason in copy_page_owner()
Currently, copy_page_owner() doesn't copy all the owner information. It
skips last_migrate_reason because copy_page_owner() is used for
migration and it will be properly set soon. But, following patch will
use copy_page_owner() and this skip will cause the problem that
allocated page has uninitialied last_migrate_reason. To prevent it,
this patch also copy last_migrate_reason in copy_page_owner().
Link: http://lkml.kernel.org/r/1464230275-25791-3-git-send-email-iamjoonsoo.kim@lge.com Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Minchan Kim <minchan@kernel.org> Cc: Alexander Potapenko <glider@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Joonsoo Kim [Tue, 26 Jul 2016 22:23:43 +0000 (15:23 -0700)]
mm/page_owner: initialize page owner without holding the zone lock
It's not necessary to initialized page_owner with holding the zone lock.
It would cause more contention on the zone lock although it's not a big
problem since it is just debug feature. But, it is better than before
so do it. This is also preparation step to use stackdepot in page owner
feature. Stackdepot allocates new pages when there is no reserved space
and holding the zone lock in this case will cause deadlock.
Link: http://lkml.kernel.org/r/1464230275-25791-2-git-send-email-iamjoonsoo.kim@lge.com Signed-off-by: Joonsoo Kim <iamjoonsoo.kim@lge.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Mel Gorman <mgorman@techsingularity.net> Cc: Minchan Kim <minchan@kernel.org> Cc: Alexander Potapenko <glider@google.com> Cc: Hugh Dickins <hughd@google.com> Cc: Michal Hocko <mhocko@kernel.org> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Minchan Kim [Tue, 26 Jul 2016 22:23:37 +0000 (15:23 -0700)]
zsmalloc: use OBJ_TAG_BIT for bit shifter
Static check warns using tag as bit shifter. It doesn't break current
working but not good for redability. Let's use OBJ_TAG_BIT as bit
shifter instead of OBJ_ALLOCATED_TAG.
Link: http://lkml.kernel.org/r/20160607045146.GF26230@bbox Signed-off-by: Minchan Kim <minchan@kernel.org> Reported-by: Dan Carpenter <dan.carpenter@oracle.com> Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Minchan Kim [Tue, 26 Jul 2016 22:23:31 +0000 (15:23 -0700)]
zsmalloc: page migration support
This patch introduces run-time migration feature for zspage.
For migration, VM uses page.lru field so it would be better to not use
page.next field which is unified with page.lru for own purpose. For
that, firstly, we can get first object offset of the page via runtime
calculation instead of using page.index so we can use page.index as link
for page chaining instead of page.next.
In case of huge object, it stores handle to page.index instead of next
link of page chaining because huge object doesn't need to next link for
page chaining. So get_next_page need to identify huge object to return
NULL. For it, this patch uses PG_owner_priv_1 flag of the page flag.
For migration, it supports three functions
* zs_page_isolate
It isolates a zspage which includes a subpage VM want to migrate from
class so anyone cannot allocate new object from the zspage.
We could try to isolate a zspage by the number of subpage so subsequent
isolation trial of other subpage of the zpsage shouldn't fail. For
that, we introduce zspage.isolated count. With that, zs_page_isolate
can know whether zspage is already isolated or not for migration so if
it is isolated for migration, subsequent isolation trial can be
successful without trying further isolation.
* zs_page_migrate
First of all, it holds write-side zspage->lock to prevent migrate other
subpage in zspage. Then, lock all objects in the page VM want to
migrate. The reason we should lock all objects in the page is due to
race between zs_map_object and zs_page_migrate.
If zs_page_migrate doesn't do trypin_tag, zs_map_object's page can be
stale by migration so it goes crash.
If it locks all of objects successfully, it copies content from old page
to new one, finally, create new zspage chain with new page. And if it's
last isolated subpage in the zspage, put the zspage back to class.
* zs_page_putback
It returns isolated zspage to right fullness_group list if it fails to
migrate a page. If it find a zspage is ZS_EMPTY, it queues zspage
freeing to workqueue. See below about async zspage freeing.
This patch introduces asynchronous zspage free. The reason to need it
is we need page_lock to clear PG_movable but unfortunately, zs_free path
should be atomic so the apporach is try to grab page_lock. If it got
page_lock of all of pages successfully, it can free zspage immediately.
Otherwise, it queues free request and free zspage via workqueue in
process context.
If zs_free finds the zspage is isolated when it try to free zspage, it
delays the freeing until zs_page_putback finds it so it will free free
the zspage finally.
In this patch, we expand fullness_list from ZS_EMPTY to ZS_FULL. First
of all, it will use ZS_EMPTY list for delay freeing. And with adding
ZS_FULL list, it makes to identify whether zspage is isolated or not via
list_empty(&zspage->list) test.
Minchan Kim [Tue, 26 Jul 2016 22:23:28 +0000 (15:23 -0700)]
zsmalloc: use freeobj for index
Zsmalloc stores first free object's <PFN, obj_idx> position into freeobj
in each zspage. If we change it with index from first_page instead of
position, it makes page migration simple because we don't need to
correct other entries for linked list if a page is migrated out.
Minchan Kim [Tue, 26 Jul 2016 22:23:26 +0000 (15:23 -0700)]
zsmalloc: separate free_zspage from putback_zspage
Currently, putback_zspage does free zspage under class->lock if fullness
become ZS_EMPTY but it makes trouble to implement locking scheme for new
zspage migration. So, this patch is to separate free_zspage from
putback_zspage and free zspage out of class->lock which is preparation
for zspage migration.
Minchan Kim [Tue, 26 Jul 2016 22:23:23 +0000 (15:23 -0700)]
zsmalloc: introduce zspage structure
We have squeezed meta data of zspage into first page's descriptor. So,
to get meta data from subpage, we should get first page first of all.
But it makes trouble to implment page migration feature of zsmalloc
because any place where to get first page from subpage can be raced with
first page migration. IOW, first page it got could be stale. For
preventing it, I have tried several approahces but it made code
complicated so finally, I concluded to separate metadata from first
page. Of course, it consumes more memory. IOW, 16bytes per zspage on
32bit at the moment. It means we lost 1% at *worst case*(40B/4096B)
which is not bad I think at the cost of maintenance.
Minchan Kim [Tue, 26 Jul 2016 22:23:14 +0000 (15:23 -0700)]
zsmalloc: use bit_spin_lock
Use kernel standard bit spin-lock instead of custom mess. Even, it has
a bug which doesn't disable preemption. The reason we don't have any
problem is that we have used it during preemption disable section by
class->lock spinlock. So no need to go to stable.
Minchan Kim [Tue, 26 Jul 2016 22:23:09 +0000 (15:23 -0700)]
mm: balloon: use general non-lru movable page feature
Now, VM has a feature to migrate non-lru movable pages so balloon
doesn't need custom migration hooks in migrate.c and compaction.c.
Instead, this patch implements the page->mapping->a_ops->
{isolate|migrate|putback} functions.
With that, we could remove hooks for ballooning in general migration
functions and make balloon compaction simple.
[akpm@linux-foundation.org: compaction.h requires that the includer first include node.h] Link: http://lkml.kernel.org/r/1464736881-24886-4-git-send-email-minchan@kernel.org Signed-off-by: Gioh Kim <gi-oh.kim@profitbricks.com> Signed-off-by: Minchan Kim <minchan@kernel.org> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Rafael Aquini <aquini@redhat.com> Cc: Konstantin Khlebnikov <koct9i@gmail.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Minchan Kim [Tue, 26 Jul 2016 22:23:05 +0000 (15:23 -0700)]
mm: migrate: support non-lru movable page migration
We have allowed migration for only LRU pages until now and it was enough
to make high-order pages. But recently, embedded system(e.g., webOS,
android) uses lots of non-movable pages(e.g., zram, GPU memory) so we
have seen several reports about troubles of small high-order allocation.
For fixing the problem, there were several efforts (e,g,. enhance
compaction algorithm, SLUB fallback to 0-order page, reserved memory,
vmalloc and so on) but if there are lots of non-movable pages in system,
their solutions are void in the long run.
So, this patch is to support facility to change non-movable pages with
movable. For the feature, this patch introduces functions related to
migration to address_space_operations as well as some page flags.
If a driver want to make own pages movable, it should define three
functions which are function pointers of struct
address_space_operations.
What VM expects on isolate_page function of driver is to return *true*
if driver isolates page successfully. On returing true, VM marks the
page as PG_isolated so concurrent isolation in several CPUs skip the
page for isolation. If a driver cannot isolate the page, it should
return *false*.
Once page is successfully isolated, VM uses page.lru fields so driver
shouldn't expect to preserve values in that fields.
After isolation, VM calls migratepage of driver with isolated page. The
function of migratepage is to move content of the old page to new page
and set up fields of struct page newpage. Keep in mind that you should
indicate to the VM the oldpage is no longer movable via
__ClearPageMovable() under page_lock if you migrated the oldpage
successfully and returns 0. If driver cannot migrate the page at the
moment, driver can return -EAGAIN. On -EAGAIN, VM will retry page
migration in a short time because VM interprets -EAGAIN as "temporal
migration failure". On returning any error except -EAGAIN, VM will give
up the page migration without retrying in this time.
Driver shouldn't touch page.lru field VM using in the functions.
3. void (*putback_page)(struct page *);
If migration fails on isolated page, VM should return the isolated page
to the driver so VM calls driver's putback_page with migration failed
page. In this function, driver should put the isolated page back to the
own data structure.
4. non-lru movable page flags
There are two page flags for supporting non-lru movable page.
* PG_movable
Driver should use the below function to make page movable under
page_lock.
It needs argument of address_space for registering migration family
functions which will be called by VM. Exactly speaking, PG_movable is
not a real flag of struct page. Rather than, VM reuses page->mapping's
lower bits to represent it.
so driver shouldn't access page->mapping directly. Instead, driver
should use page_mapping which mask off the low two bits of page->mapping
so it can get right struct address_space.
For testing of non-lru movable page, VM supports __PageMovable function.
However, it doesn't guarantee to identify non-lru movable page because
page->mapping field is unified with other variables in struct page. As
well, if driver releases the page after isolation by VM, page->mapping
doesn't have stable value although it has PAGE_MAPPING_MOVABLE (Look at
__ClearPageMovable). But __PageMovable is cheap to catch whether page
is LRU or non-lru movable once the page has been isolated. Because LRU
pages never can have PAGE_MAPPING_MOVABLE in page->mapping. It is also
good for just peeking to test non-lru movable pages before more
expensive checking with lock_page in pfn scanning to select victim.
For guaranteeing non-lru movable page, VM provides PageMovable function.
Unlike __PageMovable, PageMovable functions validates page->mapping and
mapping->a_ops->isolate_page under lock_page. The lock_page prevents
sudden destroying of page->mapping.
Driver using __SetPageMovable should clear the flag via
__ClearMovablePage under page_lock before the releasing the page.
* PG_isolated
To prevent concurrent isolation among several CPUs, VM marks isolated
page as PG_isolated under lock_page. So if a CPU encounters PG_isolated
non-lru movable page, it can skip it. Driver doesn't need to manipulate
the flag because VM will set/clear it automatically. Keep in mind that
if driver sees PG_isolated page, it means the page have been isolated by
VM so it shouldn't touch page.lru field. PG_isolated is alias with
PG_reclaim flag so driver shouldn't use the flag for own purpose.
[opensource.ganesh@gmail.com: mm/compaction: remove local variable is_lru] Link: http://lkml.kernel.org/r/20160618014841.GA7422@leo-test Link: http://lkml.kernel.org/r/1464736881-24886-3-git-send-email-minchan@kernel.org Signed-off-by: Gioh Kim <gi-oh.kim@profitbricks.com> Signed-off-by: Minchan Kim <minchan@kernel.org> Signed-off-by: Ganesh Mahendran <opensource.ganesh@gmail.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Cc: Rik van Riel <riel@redhat.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: Mel Gorman <mgorman@suse.de> Cc: Hugh Dickins <hughd@google.com> Cc: Rafael Aquini <aquini@redhat.com> Cc: Jonathan Corbet <corbet@lwn.net> Cc: John Einar Reitan <john.reitan@foss.arm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Minchan Kim [Tue, 26 Jul 2016 22:23:02 +0000 (15:23 -0700)]
mm: use put_page() to free page instead of putback_lru_page()
Recently, I got many reports about perfermance degradation in embedded
system(Android mobile phone, webOS TV and so on) and easy fork fail.
The problem was fragmentation caused by zram and GPU driver mainly.
With memory pressure, their pages were spread out all of pageblock and
it cannot be migrated with current compaction algorithm which supports
only LRU pages. In the end, compaction cannot work well so reclaimer
shrinks all of working set pages. It made system very slow and even to
fail to fork easily which requires order-[2 or 3] allocations.
Other pain point is that they cannot use CMA memory space so when OOM
kill happens, I can see many free pages in CMA area, which is not memory
efficient. In our product which has big CMA memory, it reclaims zones
too exccessively to allocate GPU and zram page although there are lots
of free space in CMA so system becomes very slow easily.
To solve these problem, this patch tries to add facility to migrate
non-lru pages via introducing new functions and page flags to help
migration.
For details, please read description in "mm: migrate: support non-lru
movable page migration".
Originally, Gioh Kim had tried to support this feature but he moved so I
took over the work. I took many code from his work and changed a little
bit and Konstantin Khlebnikov helped Gioh a lot so he should deserve to
have many credit, too.
And I should mention Chulmin who have tested this patchset heavily so I
can find many bugs from him. :)
Thanks, Gioh, Konstantin and Chulmin!
This patchset consists of five parts.
1. clean up migration
mm: use put_page to free page instead of putback_lru_page
3. rework KVM memory-ballooning
mm: balloon: use general non-lru movable page feature
4. zsmalloc refactoring for preparing page migration
zsmalloc: keep max_object in size_class
zsmalloc: use bit_spin_lock
zsmalloc: use accessor
zsmalloc: factor page chain functionality out
zsmalloc: introduce zspage structure
zsmalloc: separate free_zspage from putback_zspage
zsmalloc: use freeobj for index
5. zsmalloc page migration
zsmalloc: page migration support
zram: use __GFP_MOVABLE for memory allocation
This patch (of 12):
Procedure of page migration is as follows:
First of all, it should isolate a page from LRU and try to migrate the
page. If it is successful, it releases the page for freeing.
Otherwise, it should put the page back to LRU list.
For LRU pages, we have used putback_lru_page for both freeing and
putback to LRU list. It's okay because put_page is aware of LRU list so
if it releases last refcount of the page, it removes the page from LRU
list. However, It makes unnecessary operations (e.g., lru_cache_add,
pagevec and flags operations. It would be not significant but no worth
to do) and harder to support new non-lru page migration because put_page
isn't aware of non-lru page's data structure.
To solve the problem, we can add new hook in put_page with PageMovable
flags check but it can increase overhead in hot path and needs new
locking scheme to stabilize the flag check with put_page.
So, this patch cleans it up to divide two semantic(ie, put and putback).
If migration is successful, use put_page instead of putback_lru_page and
use putback_lru_page only on failure. That makes code more readable and
doesn't add overhead in put_page.
Comment from Vlastimil
"Yeah, and compaction (perhaps also other migration users) has to drain
the lru pvec... Getting rid of this stuff is worth even by itself."
We now allocate streams from CPU_UP hot-plug path, there are no
context-dependent stream allocations anymore and we can schedule from
zcomp_strm_alloc(). Use GFP_KERNEL directly and drop a gfp_t parameter.
Link: http://lkml.kernel.org/r/20160531122017.2878-9-sergey.senozhatsky@gmail.com Signed-off-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com> Acked-by: Minchan Kim <minchan@kernel.org> 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>
Add "deflate", "lz4hc", "842" algorithms to the list of known
compression backends. The real availability of those algorithms,
however, depends on the corresponding CONFIG_CRYPTO_FOO config options.
There is no way to get a string with all the crypto comp algorithms
supported by the crypto comp engine, so we need to maintain our own
backends list. At the same time we additionally need to use
crypto_has_comp() to make sure that the user has requested a compression
algorithm that is recognized by the crypto comp engine. Relying on
/proc/crypto is not an options here, because it does not show
not-yet-inserted compression modules.
So the user can't tell exactly if the lz4 is really supported from
/proc/crypto output, unless someone or something has loaded it.
This patch also adds crypto_has_comp() to zcomp_available_show(). We
store all the compression algorithms names in zcomp's `backends' array,
regardless the CONFIG_CRYPTO_FOO configuration, but show only those that
are also supported by crypto engine. This helps user to know the exact
list of compression algorithms that can be used.
Example:
module lz4 is not loaded yet, but is supported by the crypto
engine. /proc/crypto has no information on this module, while
zram's `comp_algorithm' lists it:
We still use the `backends' array to determine if the requested
compression backend is known to crypto api. This array, however, may not
contain some entries, therefore as the last step we call crypto_has_comp()
function which attempts to insmod the requested compression algorithm to
determine if crypto api supports it. The advantage of this method is that
now we permit the usage of out-of-tree crypto compression modules
(implementing S/W or H/W compression).
We don't have an idle zstreams list anymore and our write path now works
absolutely differently, preventing preemption during compression. This
removes possibilities of read paths preempting writes at wrong places
(which could badly affect the performance of both paths) and at the same
time opens the door for a move from custom LZO/LZ4 compression backends
implementation to a more generic one, using crypto compress API.
Joonsoo Kim [1] attempted to do this a while ago, but faced with the
need of introducing a new crypto API interface. The root cause was the
fact that crypto API compression algorithms require a compression stream
structure (in zram terminology) for both compression and decompression
ops, while in reality only several of compression algorithms really need
it. This resulted in a concept of context-less crypto API compression
backends [2]. Both write and read paths, though, would have been
executed with the preemption enabled, which in the worst case could have
resulted in a decreased worst-case performance, e.g. consider the
following case:
CPU0
zram_write()
spin_lock()
take the last idle stream
spin_unlock()
<< preempted >>
zram_read()
spin_lock()
no idle streams
spin_unlock()
schedule()
resuming zram_write compression()
but it took me some time to realize that, and it took even longer to
evolve zram and to make it ready for crypto API. The key turned out to be
-- drop the idle streams list entirely. Without the idle streams list we
are free to use compression algorithms that require compression stream for
decompression (read), because streams are now placed in per-cpu data and
each write path has to disable preemption for compression op, almost
completely eliminating the aforementioned case (technically, we still have
a small chance, because write path has a fast and a slow paths and the
slow path is executed with the preemption enabled; but the frequency of
failed fast path is too low).
This has started as a 'add zlib support' work, but after some thinking I
saw no blockers for a bigger change -- a switch to crypto API.
We don't have an idle zstreams list anymore and our write path now works
absolutely differently, preventing preemption during compression. This
removes possibilities of read paths preempting writes at wrong places
and opens the door for a move from custom LZO/LZ4 compression backends
implementation to a more generic one, using crypto compress API.
This patch set also eliminates the need of a new context-less crypto API
interface, which was quite hard to sell, so we can move along faster.
benchmarks:
(x86_64, 4GB, zram-perf script)
perf reported run-time fio (max jobs=3). I performed fio test with the
increasing number of parallel jobs (max to 3) on a 3G zram device, using
`static' data and the following crypto comp algorithms:
842, deflate, lz4, lz4hc, lzo
the output was:
- test running time (which can tell us what algorithms performs faster)
and
- zram mm_stat (which tells the compressed memory size, max used memory, etc).
It's just for information. for example, LZ4HC has twice the running
time of LZO, but the compressed memory size is: 23592960 vs 34603008
bytes.
test-fio-zram-842
197.907655282 seconds time elapsed
201.623142884 seconds time elapsed
226.854291345 seconds time elapsed
test-fio-zram-DEFLATE
253.259516155 seconds time elapsed
258.148563401 seconds time elapsed
290.251909365 seconds time elapsed
test-fio-zram-LZ4
27.022598717 seconds time elapsed
29.580522717 seconds time elapsed
33.293463430 seconds time elapsed
test-fio-zram-LZ4HC
56.393954615 seconds time elapsed
74.904659747 seconds time elapsed
101.940998564 seconds time elapsed
test-fio-zram-LZO
28.155948075 seconds time elapsed
30.390036330 seconds time elapsed
34.455773159 seconds time elapsed
powerpc/mm: check for irq disabled() only if DEBUG_VM is enabled
We don't need to check this always. The idea here is to capture the
wrong usage of find_linux_pte_or_hugepte and we can do that by
occasionally running with DEBUG_VM enabled.
Link: http://lkml.kernel.org/r/1464692688-6612-2-git-send-email-aneesh.kumar@linux.vnet.ibm.com Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com> Cc: Michael Ellerman <mpe@ellerman.id.au> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Reviewed-by: 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>
memory-hotplug: more general validation of zone during online
When memory is onlined, we are only able to rezone from ZONE_MOVABLE to
ZONE_KERNEL, or from (ZONE_MOVABLE - 1) to ZONE_MOVABLE.
To be more flexible, use the following criteria instead; to online
memory from zone X into zone Y,
* Any zones between X and Y must be unused.
* If X is lower than Y, the onlined memory must lie at the end of X.
* If X is higher than Y, the onlined memory must lie at the start of X.
Add zone_can_shift() to make this determination.
Link: http://lkml.kernel.org/r/1462816419-4479-3-git-send-email-arbab@linux.vnet.ibm.com Signed-off-by: Reza Arbab <arbab@linux.vnet.ibm.com> Reviewd-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com> Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Cc: Daniel Kiper <daniel.kiper@oracle.com> Cc: Dan Williams <dan.j.williams@intel.com> Cc: Vlastimil Babka <vbabka@suse.cz> Cc: Tang Chen <tangchen@cn.fujitsu.com> Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com> Cc: David Vrabel <david.vrabel@citrix.com> Cc: Vitaly Kuznetsov <vkuznets@redhat.com> Cc: David Rientjes <rientjes@google.com> Cc: Andrew Banman <abanman@sgi.com> Cc: Chen Yucong <slaoub@gmail.com> Cc: Yasunori Goto <y-goto@jp.fujitsu.com> Cc: Zhang Zhen <zhenzhang.zhang@huawei.com> Cc: Shaohua Li <shaohua.li@intel.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
As a part of memory initialisation the architecture passes an array to
free_area_init_nodes() which specifies the max PFN of each memory zone.
This array is not necessarily monotonic (due to unused zones) so this
array is parsed to build monotonic lists of the min and max PFN for each
zone. ZONE_MOVABLE is special cased here as its limits are managed by
the mm subsystem rather than the architecture. Unfortunately, this
special casing is broken when ZONE_MOVABLE is the not the last zone in
the zone list. The core of the issue is:
if (i == ZONE_MOVABLE)
continue;
arch_zone_lowest_possible_pfn[i] =
arch_zone_highest_possible_pfn[i-1];
As ZONE_MOVABLE is skipped the lowest_possible_pfn of the next zone will
be set to zero. This patch fixes this bug by adding explicitly tracking
where the next zone should start rather than relying on the contents
arch_zone_highest_possible_pfn[].
Thie is low priority. To get bitten by this you need to enable a zone
that appears after ZONE_MOVABLE in the zone_type enum. As far as I can
tell this means running a kernel with ZONE_DEVICE or ZONE_CMA enabled,
so I can't see this affecting too many people.
I only noticed this because I've been fiddling with ZONE_DEVICE on
powerpc and 4.6 broke my test kernel. This bug, in conjunction with the
changes in Taku Izumi's kernelcore=mirror patch (d91749c1dda71) and
powerpc being the odd architecture which initialises max_zone_pfn[] to
~0ul instead of 0 caused all of system memory to be placed into
ZONE_DEVICE at boot, followed a panic since device memory cannot be used
for kernel allocations. I've already submitted a patch to fix the
powerpc specific bits, but I figured this should be fixed too.
Link: http://lkml.kernel.org/r/1462435033-15601-1-git-send-email-oohall@gmail.com Signed-off-by: Oliver O'Halloran <oohall@gmail.com> Cc: Anton Blanchard <anton@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Mel Gorman <mgorman@techsingularity.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Li RongQing [Tue, 26 Jul 2016 22:22:14 +0000 (15:22 -0700)]
mm/memcontrol.c: remove the useless parameter for mc_handle_swap_pte
It seems like this parameter has never been used since being introduced
by 90254a65833b ("memcg: clean up move charge"). Not a big deal because
I assume the function would get inlined into the caller anyway but why
not get rid of it.
When both arguments to kmalloc_array() or kcalloc() are known at compile
time then their product is known at compile time but search for kmalloc
cache happens at runtime not at compile time.
Link: http://lkml.kernel.org/r/20160627213454.GA2440@p183.telecom.by Signed-off-by: Alexey Dobriyan <adobriyan@gmail.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> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Michal Hocko [Tue, 26 Jul 2016 22:22:05 +0000 (15:22 -0700)]
slab: do not panic on invalid gfp_mask
Both SLAB and SLUB BUG() when a caller provides an invalid gfp_mask.
This is a rather harsh way to announce a non-critical issue. Allocator
is free to ignore invalid flags. Let's simply replace BUG() by
dump_stack to tell the offender and fixup the mask to move on with the
allocation request.
This is an example for kmalloc(GFP_KERNEL|__GFP_HIGHMEM) from a test
module:
Unexpected gfp: 0x2 (__GFP_HIGHMEM). Fixing up to gfp: 0x24000c0 (GFP_KERNEL). Fix your code!
CPU: 0 PID: 2916 Comm: insmod Tainted: G O 4.6.0-slabgfp2-00002-g4cdfc2ef4892-dirty #936
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Debian-1.8.2-1 04/01/2014
Call Trace:
dump_stack+0x67/0x90
cache_alloc_refill+0x201/0x617
kmem_cache_alloc_trace+0xa7/0x24a
? 0xffffffffa0005000
mymodule_init+0x20/0x1000 [test_slab]
do_one_initcall+0xe7/0x16c
? rcu_read_lock_sched_held+0x61/0x69
? kmem_cache_alloc_trace+0x197/0x24a
do_init_module+0x5f/0x1d9
load_module+0x1a3d/0x1f21
? retint_kernel+0x2d/0x2d
SyS_init_module+0xe8/0x10e
? SyS_init_module+0xe8/0x10e
do_syscall_64+0x68/0x13f
entry_SYSCALL64_slow_path+0x25/0x25
Link: http://lkml.kernel.org/r/1465548200-11384-2-git-send-email-mhocko@kernel.org Signed-off-by: Michal Hocko <mhocko@suse.com> Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.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> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Thomas Garnier [Tue, 26 Jul 2016 22:21:59 +0000 (15:21 -0700)]
mm: SLUB freelist randomization
Implements freelist randomization for the SLUB allocator. It was
previous implemented for the SLAB allocator. Both use the same
configuration option (CONFIG_SLAB_FREELIST_RANDOM).
The list is randomized during initialization of a new set of pages. The
order on different freelist sizes is pre-computed at boot for
performance. Each kmem_cache has its own randomized freelist.
This security feature reduces the predictability of the kernel SLUB
allocator against heap overflows rendering attacks much less stable.
For example these attacks exploit the predictability of the heap:
- Linux Kernel CAN SLUB overflow (https://goo.gl/oMNWkU)
- Exploiting Linux Kernel Heap corruptions (http://goo.gl/EXLn95)
Performance results:
slab_test impact is between 3% to 4% on average for 100000 attempts
without smp. It is a very focused testing, kernbench show the overall
impact on the system is way lower.
Before:
Single thread testing
=====================
1. Kmalloc: Repeatedly allocate then free test
100000 times kmalloc(8) -> 49 cycles kfree -> 77 cycles
100000 times kmalloc(16) -> 51 cycles kfree -> 79 cycles
100000 times kmalloc(32) -> 53 cycles kfree -> 83 cycles
100000 times kmalloc(64) -> 62 cycles kfree -> 90 cycles
100000 times kmalloc(128) -> 81 cycles kfree -> 97 cycles
100000 times kmalloc(256) -> 98 cycles kfree -> 121 cycles
100000 times kmalloc(512) -> 95 cycles kfree -> 122 cycles
100000 times kmalloc(1024) -> 96 cycles kfree -> 126 cycles
100000 times kmalloc(2048) -> 115 cycles kfree -> 140 cycles
100000 times kmalloc(4096) -> 149 cycles kfree -> 171 cycles
2. Kmalloc: alloc/free test
100000 times kmalloc(8)/kfree -> 70 cycles
100000 times kmalloc(16)/kfree -> 70 cycles
100000 times kmalloc(32)/kfree -> 70 cycles
100000 times kmalloc(64)/kfree -> 70 cycles
100000 times kmalloc(128)/kfree -> 70 cycles
100000 times kmalloc(256)/kfree -> 69 cycles
100000 times kmalloc(512)/kfree -> 70 cycles
100000 times kmalloc(1024)/kfree -> 73 cycles
100000 times kmalloc(2048)/kfree -> 72 cycles
100000 times kmalloc(4096)/kfree -> 71 cycles
After:
Single thread testing
=====================
1. Kmalloc: Repeatedly allocate then free test
100000 times kmalloc(8) -> 57 cycles kfree -> 78 cycles
100000 times kmalloc(16) -> 61 cycles kfree -> 81 cycles
100000 times kmalloc(32) -> 76 cycles kfree -> 93 cycles
100000 times kmalloc(64) -> 83 cycles kfree -> 94 cycles
100000 times kmalloc(128) -> 106 cycles kfree -> 107 cycles
100000 times kmalloc(256) -> 118 cycles kfree -> 117 cycles
100000 times kmalloc(512) -> 114 cycles kfree -> 116 cycles
100000 times kmalloc(1024) -> 115 cycles kfree -> 118 cycles
100000 times kmalloc(2048) -> 147 cycles kfree -> 131 cycles
100000 times kmalloc(4096) -> 214 cycles kfree -> 161 cycles
2. Kmalloc: alloc/free test
100000 times kmalloc(8)/kfree -> 66 cycles
100000 times kmalloc(16)/kfree -> 66 cycles
100000 times kmalloc(32)/kfree -> 66 cycles
100000 times kmalloc(64)/kfree -> 66 cycles
100000 times kmalloc(128)/kfree -> 65 cycles
100000 times kmalloc(256)/kfree -> 67 cycles
100000 times kmalloc(512)/kfree -> 67 cycles
100000 times kmalloc(1024)/kfree -> 64 cycles
100000 times kmalloc(2048)/kfree -> 67 cycles
100000 times kmalloc(4096)/kfree -> 67 cycles
Kernbench, before:
Average Optimal load -j 12 Run (std deviation):
Elapsed Time 101.873 (1.16069)
User Time 1045.22 (1.60447)
System Time 88.969 (0.559195)
Percent CPU 1112.9 (13.8279)
Context Switches 189140 (2282.15)
Sleeps 99008.6 (768.091)
After:
Average Optimal load -j 12 Run (std deviation):
Elapsed Time 102.47 (0.562732)
User Time 1045.3 (1.34263)
System Time 88.311 (0.342554)
Percent CPU 1105.8 (6.49444)
Context Switches 189081 (2355.78)
Sleeps 99231.5 (800.358)
Link: http://lkml.kernel.org/r/1464295031-26375-3-git-send-email-thgarnie@google.com Signed-off-by: Thomas Garnier <thgarnie@google.com> Reviewed-by: Kees Cook <keescook@chromium.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>
Thomas Garnier [Tue, 26 Jul 2016 22:21:56 +0000 (15:21 -0700)]
mm: reorganize SLAB freelist randomization
The kernel heap allocators are using a sequential freelist making their
allocation predictable. This predictability makes kernel heap overflow
easier to exploit. An attacker can careful prepare the kernel heap to
control the following chunk overflowed.
For example these attacks exploit the predictability of the heap:
- Linux Kernel CAN SLUB overflow (https://goo.gl/oMNWkU)
- Exploiting Linux Kernel Heap corruptions (http://goo.gl/EXLn95)
***Problems that needed solving:
- Randomize the Freelist (singled linked) used in the SLUB allocator.
- Ensure good performance to encourage usage.
- Get best entropy in early boot stage.
***Parts:
- 01/02 Reorganize the SLAB Freelist randomization to share elements
with the SLUB implementation.
- 02/02 The SLUB Freelist randomization implementation. Similar approach
than the SLAB but tailored to the singled freelist used in SLUB.
***Performance data:
slab_test impact is between 3% to 4% on average for 100000 attempts
without smp. It is a very focused testing, kernbench show the overall
impact on the system is way lower.
Before:
Single thread testing
=====================
1. Kmalloc: Repeatedly allocate then free test
100000 times kmalloc(8) -> 49 cycles kfree -> 77 cycles
100000 times kmalloc(16) -> 51 cycles kfree -> 79 cycles
100000 times kmalloc(32) -> 53 cycles kfree -> 83 cycles
100000 times kmalloc(64) -> 62 cycles kfree -> 90 cycles
100000 times kmalloc(128) -> 81 cycles kfree -> 97 cycles
100000 times kmalloc(256) -> 98 cycles kfree -> 121 cycles
100000 times kmalloc(512) -> 95 cycles kfree -> 122 cycles
100000 times kmalloc(1024) -> 96 cycles kfree -> 126 cycles
100000 times kmalloc(2048) -> 115 cycles kfree -> 140 cycles
100000 times kmalloc(4096) -> 149 cycles kfree -> 171 cycles
2. Kmalloc: alloc/free test
100000 times kmalloc(8)/kfree -> 70 cycles
100000 times kmalloc(16)/kfree -> 70 cycles
100000 times kmalloc(32)/kfree -> 70 cycles
100000 times kmalloc(64)/kfree -> 70 cycles
100000 times kmalloc(128)/kfree -> 70 cycles
100000 times kmalloc(256)/kfree -> 69 cycles
100000 times kmalloc(512)/kfree -> 70 cycles
100000 times kmalloc(1024)/kfree -> 73 cycles
100000 times kmalloc(2048)/kfree -> 72 cycles
100000 times kmalloc(4096)/kfree -> 71 cycles
After:
Single thread testing
=====================
1. Kmalloc: Repeatedly allocate then free test
100000 times kmalloc(8) -> 57 cycles kfree -> 78 cycles
100000 times kmalloc(16) -> 61 cycles kfree -> 81 cycles
100000 times kmalloc(32) -> 76 cycles kfree -> 93 cycles
100000 times kmalloc(64) -> 83 cycles kfree -> 94 cycles
100000 times kmalloc(128) -> 106 cycles kfree -> 107 cycles
100000 times kmalloc(256) -> 118 cycles kfree -> 117 cycles
100000 times kmalloc(512) -> 114 cycles kfree -> 116 cycles
100000 times kmalloc(1024) -> 115 cycles kfree -> 118 cycles
100000 times kmalloc(2048) -> 147 cycles kfree -> 131 cycles
100000 times kmalloc(4096) -> 214 cycles kfree -> 161 cycles
2. Kmalloc: alloc/free test
100000 times kmalloc(8)/kfree -> 66 cycles
100000 times kmalloc(16)/kfree -> 66 cycles
100000 times kmalloc(32)/kfree -> 66 cycles
100000 times kmalloc(64)/kfree -> 66 cycles
100000 times kmalloc(128)/kfree -> 65 cycles
100000 times kmalloc(256)/kfree -> 67 cycles
100000 times kmalloc(512)/kfree -> 67 cycles
100000 times kmalloc(1024)/kfree -> 64 cycles
100000 times kmalloc(2048)/kfree -> 67 cycles
100000 times kmalloc(4096)/kfree -> 67 cycles
Kernbench, before:
Average Optimal load -j 12 Run (std deviation):
Elapsed Time 101.873 (1.16069)
User Time 1045.22 (1.60447)
System Time 88.969 (0.559195)
Percent CPU 1112.9 (13.8279)
Context Switches 189140 (2282.15)
Sleeps 99008.6 (768.091)
After:
Average Optimal load -j 12 Run (std deviation):
Elapsed Time 102.47 (0.562732)
User Time 1045.3 (1.34263)
System Time 88.311 (0.342554)
Percent CPU 1105.8 (6.49444)
Context Switches 189081 (2355.78)
Sleeps 99231.5 (800.358)
This patch (of 2):
This commit reorganizes the previous SLAB freelist randomization to
prepare for the SLUB implementation. It moves functions that will be
shared to slab_common.
The entropy functions are changed to align with the SLUB implementation,
now using get_random_(int|long) functions. These functions were chosen
because they provide a bit more entropy early on boot and better
performance when specific arch instructions are not available.
[akpm@linux-foundation.org: fix build] Link: http://lkml.kernel.org/r/1464295031-26375-2-git-send-email-thgarnie@google.com Signed-off-by: Thomas Garnier <thgarnie@google.com> Reviewed-by: Kees Cook <keescook@chromium.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>
Brian Foster [Tue, 26 Jul 2016 22:21:53 +0000 (15:21 -0700)]
fs/fs-writeback.c: inode writeback list tracking tracepoints
The per-sb inode writeback list tracks inodes currently under writeback
to facilitate efficient sync processing. In particular, it ensures that
sync only needs to walk through a list of inodes that were cleaned by
the sync.
Add a couple tracepoints to help identify when inodes are added/removed
to and from the writeback lists. Piggyback off of the writeback
lazytime tracepoint template as it already tracks the relevant inode
information.
Link: http://lkml.kernel.org/r/1466594593-6757-3-git-send-email-bfoster@redhat.com Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Jan Kara <jack@suse.cz> Cc: Dave Chinner <dchinner@redhat.com>
cc: Josef Bacik <jbacik@fb.com> Cc: Holger Hoffstätte <holger.hoffstaette@applied-asynchrony.com> Cc: Al Viro <viro@ZenIV.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Dave Chinner [Tue, 26 Jul 2016 22:21:50 +0000 (15:21 -0700)]
fs/fs-writeback.c: add a new writeback list for sync
wait_sb_inodes() currently does a walk of all inodes in the filesystem
to find dirty one to wait on during sync. This is highly inefficient
and wastes a lot of CPU when there are lots of clean cached inodes that
we don't need to wait on.
To avoid this "all inode" walk, we need to track inodes that are
currently under writeback that we need to wait for. We do this by
adding inodes to a writeback list on the sb when the mapping is first
tagged as having pages under writeback. wait_sb_inodes() can then walk
this list of "inodes under IO" and wait specifically just for the inodes
that the current sync(2) needs to wait for.
Define a couple helpers to add/remove an inode from the writeback list
and call them when the overall mapping is tagged for or cleared from
writeback. Update wait_sb_inodes() to walk only the inodes under
writeback due to the sync.
With this change, filesystem sync times are significantly reduced for
fs' with largely populated inode caches and otherwise no other work to
do. For example, on a 16xcpu 2GHz x86-64 server, 10TB XFS filesystem
with a ~10m entry inode cache, sync times are reduced from ~7.3s to less
than 0.1s when the filesystem is fully clean.
Link: http://lkml.kernel.org/r/1466594593-6757-2-git-send-email-bfoster@redhat.com Signed-off-by: Dave Chinner <dchinner@redhat.com> Signed-off-by: Josef Bacik <jbacik@fb.com> Signed-off-by: Brian Foster <bfoster@redhat.com> Reviewed-by: Jan Kara <jack@suse.cz> Tested-by: Holger Hoffstätte <holger.hoffstaette@applied-asynchrony.com> Cc: Al Viro <viro@ZenIV.linux.org.uk> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
projects / karo-tx-linux.git / log