It's pointless to kill a task if another thread sharing its mm cannot be
killed to allow future memory freeing. A subsequent patch will prevent
kills in such cases, but first it's necessary to have a way to flag a task
that shares memory with an OOM_DISABLE task that doesn't incur an
additional tasklist scan, which would make select_bad_process() an O(n^2)
function.
This patch adds an atomic counter to struct mm_struct that follows how
many threads attached to it have an oom_score_adj of OOM_SCORE_ADJ_MIN.
They cannot be killed by the kernel, so their memory cannot be freed in
oom conditions.
This only requires task_lock() on the task that we're operating on, it
does not require mm->mmap_sem since task_lock() pins the mm and the
operation is atomic.
[rientjes@google.com: changelog and sys_unshare() code]
[rientjes@google.com: protect oom_disable_count with task_lock in fork]
[rientjes@google.com: use old_mm for oom_disable_count in exec]
Signed-off-by: Ying Han <yinghan@google.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's a really simple list, and several of the users want to go backwards
in it to find the previous vma. So rather than have to look up the
previous entry with 'find_vma_prev()' or something similar, just make it
doubly linked instead.
Tested-by: Ian Campbell <ijc@hellion.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This adds annotations for RCU operations in core kernel components
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
Commit 34e55232e5 ("mm: avoid false sharing
of mm_counter") added sync_mm_rss() for syncing loosely accounted rss
counters. It's for CONFIG_MMU but sync_mm_rss is called even in NOMMU
enviroment (kerne/exit.c, fs/exec.c). Above commit doesn't handle it
well.
This patch changes
SPLIT_RSS_COUNTING depends on SPLIT_PTLOCKS && CONFIG_MMU
And for avoid unnecessary function calls, sync_mm_rss changed to be inlined
noop function in header file.
Reported-by: David Howells <dhowells@redhat.com>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Michal Simek <monstr@monstr.eu>
Signed-off-by: David Howells <dhowells@redhat.com>
Cc: Greg Ungerer <gerg@snapgear.com>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The old anon_vma code can lead to scalability issues with heavily forking
workloads. Specifically, each anon_vma will be shared between the parent
process and all its child processes.
In a workload with 1000 child processes and a VMA with 1000 anonymous
pages per process that get COWed, this leads to a system with a million
anonymous pages in the same anon_vma, each of which is mapped in just one
of the 1000 processes. However, the current rmap code needs to walk them
all, leading to O(N) scanning complexity for each page.
This can result in systems where one CPU is walking the page tables of
1000 processes in page_referenced_one, while all other CPUs are stuck on
the anon_vma lock. This leads to catastrophic failure for a benchmark
like AIM7, where the total number of processes can reach in the tens of
thousands. Real workloads are still a factor 10 less process intensive
than AIM7, but they are catching up.
This patch changes the way anon_vmas and VMAs are linked, which allows us
to associate multiple anon_vmas with a VMA. At fork time, each child
process gets its own anon_vmas, in which its COWed pages will be
instantiated. The parents' anon_vma is also linked to the VMA, because
non-COWed pages could be present in any of the children.
This reduces rmap scanning complexity to O(1) for the pages of the 1000
child processes, with O(N) complexity for at most 1/N pages in the system.
This reduces the average scanning cost in heavily forking workloads from
O(N) to 2.
The only real complexity in this patch stems from the fact that linking a
VMA to anon_vmas now involves memory allocations. This means vma_adjust
can fail, if it needs to attach a VMA to anon_vma structures. This in
turn means error handling needs to be added to the calling functions.
A second source of complexity is that, because there can be multiple
anon_vmas, the anon_vma linking in vma_adjust can no longer be done under
"the" anon_vma lock. To prevent the rmap code from walking up an
incomplete VMA, this patch introduces the VM_LOCK_RMAP VMA flag. This bit
flag uses the same slot as the NOMMU VM_MAPPED_COPY, with an ifdef in mm.h
to make sure it is impossible to compile a kernel that needs both symbolic
values for the same bitflag.
Some test results:
Without the anon_vma changes, when AIM7 hits around 9.7k users (on a test
box with 16GB RAM and not quite enough IO), the system ends up running
>99% in system time, with every CPU on the same anon_vma lock in the
pageout code.
With these changes, AIM7 hits the cross-over point around 29.7k users.
This happens with ~99% IO wait time, there never seems to be any spike in
system time. The anon_vma lock contention appears to be resolved.
[akpm@linux-foundation.org: cleanups]
Signed-off-by: Rik van Riel <riel@redhat.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Larry Woodman <lwoodman@redhat.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Hugh Dickins <hugh.dickins@tiscali.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A frequent questions from users about memory management is what numbers of
swap ents are user for processes. And this information will give some
hints to oom-killer.
Besides we can count the number of swapents per a process by scanning
/proc/<pid>/smaps, this is very slow and not good for usual process
information handler which works like 'ps' or 'top'. (ps or top is now
enough slow..)
This patch adds a counter of swapents to mm_counter and update is at each
swap events. Information is exported via /proc/<pid>/status file as
[kamezawa@bluextal memory]$ cat /proc/self/status
Name: cat
State: R (running)
Tgid: 2910
Pid: 2910
PPid: 2823
TracerPid: 0
Uid: 500 500 500 500
Gid: 500 500 500 500
FDSize: 256
Groups: 500
VmPeak: 82696 kB
VmSize: 82696 kB
VmLck: 0 kB
VmHWM: 432 kB
VmRSS: 432 kB
VmData: 172 kB
VmStk: 84 kB
VmExe: 48 kB
VmLib: 1568 kB
VmPTE: 40 kB
VmSwap: 0 kB <=============== this.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Considering the nature of per mm stats, it's the shared object among
threads and can be a cache-miss point in the page fault path.
This patch adds per-thread cache for mm_counter. RSS value will be
counted into a struct in task_struct and synchronized with mm's one at
events.
Now, in this patch, the event is the number of calls to handle_mm_fault.
Per-thread value is added to mm at each 64 calls.
rough estimation with small benchmark on parallel thread (2threads) shows
[before]
4.5 cache-miss/faults
[after]
4.0 cache-miss/faults
Anyway, the most contended object is mmap_sem if the number of threads grows.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Presently, per-mm statistics counter is defined by macro in sched.h
This patch modifies it to
- defined in mm.h as inlinf functions
- use array instead of macro's name creation.
This patch is for reducing patch size in future patch to modify
implementation of per-mm counter.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
get_unmapped_area() is unnecessary for NOMMU as no-one calls it.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Greg Ungerer <gerg@snapgear.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The vm_usage count field in struct vm_region does not need to be atomic as
it's only even modified whilst nommu_region_sem is write locked.
Signed-off-by: David Howells <dhowells@redhat.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Cc: Greg Ungerer <gerg@snapgear.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When working with FDPIC, there are many shared mappings of read-only
code regions between applications (the C library, applet packages like
busybox, etc.), but the current do_mmap_pgoff() function will issue an
icache flush whenever a VMA is added to an MM instead of only doing it
when the map is initially created.
The flush can instead be done when a region is first mmapped PROT_EXEC.
Note that we may not rely on the first mapping of a region being
executable - it's possible for it to be PROT_READ only, so we have to
remember whether we've flushed the region or not, and then flush the
entire region when a bit of it is made executable.
However, this also affects the brk area. That will no longer be
executable. We can mprotect() it to PROT_EXEC on MPU-mode kernels, but
for NOMMU mode kernels, when it increases the brk allocation, making
sys_brk() flush the extra from the icache should suffice. The brk area
probably isn't used by NOMMU programs since the brk area can only use up
the leavings from the stack allocation, where the stack allocation is
larger than requested.
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* mark struct vm_area_struct::vm_ops as const
* mark vm_ops in AGP code
But leave TTM code alone, something is fishy there with global vm_ops
being used.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Because the binfmt is not different between threads in the same process,
it can be moved from task_struct to mm_struct. And binfmt moudle is
handled per mm_struct instead of task_struct.
Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
->ioctx_lock and ->ioctx_list are used only under CONFIG_AIO.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The commit 2ff05b2b (oom: move oom_adj value) moveed the oom_adj value to
the mm_struct. It was a very good first step for sanitize OOM.
However Paul Menage reported the commit makes regression to his job
scheduler. Current OOM logic can kill OOM_DISABLED process.
Why? His program has the code of similar to the following.
...
set_oom_adj(OOM_DISABLE); /* The job scheduler never killed by oom */
...
if (vfork() == 0) {
set_oom_adj(0); /* Invoked child can be killed */
execve("foo-bar-cmd");
}
....
vfork() parent and child are shared the same mm_struct. then above
set_oom_adj(0) doesn't only change oom_adj for vfork() child, it's also
change oom_adj for vfork() parent. Then, vfork() parent (job scheduler)
lost OOM immune and it was killed.
Actually, fork-setting-exec idiom is very frequently used in userland program.
We must not break this assumption.
Then, this patch revert commit 2ff05b2b and related commit.
Reverted commit list
---------------------
- commit 2ff05b2b4e (oom: move oom_adj value from task_struct to mm_struct)
- commit 4d8b9135c3 (oom: avoid unnecessary mm locking and scanning for OOM_DISABLE)
- commit 8123681022 (oom: only oom kill exiting tasks with attached memory)
- commit 933b787b57 (mm: copy over oom_adj value at fork time)
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Cc: David Rientjes <rientjes@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* akpm: (182 commits)
fbdev: bf54x-lq043fb: use kzalloc over kmalloc/memset
fbdev: *bfin*: fix __dev{init,exit} markings
fbdev: *bfin*: drop unnecessary calls to memset
fbdev: bfin-t350mcqb-fb: drop unused local variables
fbdev: blackfin has __raw I/O accessors, so use them in fb.h
fbdev: s1d13xxxfb: add accelerated bitblt functions
tcx: use standard fields for framebuffer physical address and length
fbdev: add support for handoff from firmware to hw framebuffers
intelfb: fix a bug when changing video timing
fbdev: use framebuffer_release() for freeing fb_info structures
radeon: P2G2CLK_ALWAYS_ONb tested twice, should 2nd be P2G2CLK_DAC_ALWAYS_ONb?
s3c-fb: CPUFREQ frequency scaling support
s3c-fb: fix resource releasing on error during probing
carminefb: fix possible access beyond end of carmine_modedb[]
acornfb: remove fb_mmap function
mb862xxfb: use CONFIG_OF instead of CONFIG_PPC_OF
mb862xxfb: restrict compliation of platform driver to PPC
Samsung SoC Framebuffer driver: add Alpha Channel support
atmel-lcdc: fix pixclock upper bound detection
offb: use framebuffer_alloc() to allocate fb_info struct
...
Manually fix up conflicts due to kmemcheck in mm/slab.c
The per-task oom_adj value is a characteristic of its mm more than the
task itself since it's not possible to oom kill any thread that shares the
mm. If a task were to be killed while attached to an mm that could not be
freed because another thread were set to OOM_DISABLE, it would have
needlessly been terminated since there is no potential for future memory
freeing.
This patch moves oomkilladj (now more appropriately named oom_adj) from
struct task_struct to struct mm_struct. This requires task_lock() on a
task to check its oom_adj value to protect against exec, but it's already
necessary to take the lock when dereferencing the mm to find the total VM
size for the badness heuristic.
This fixes a livelock if the oom killer chooses a task and another thread
sharing the same memory has an oom_adj value of OOM_DISABLE. This occurs
because oom_kill_task() repeatedly returns 1 and refuses to kill the
chosen task while select_bad_process() will repeatedly choose the same
task during the next retry.
Taking task_lock() in select_bad_process() to check for OOM_DISABLE and in
oom_kill_task() to check for threads sharing the same memory will be
removed in the next patch in this series where it will no longer be
necessary.
Writing to /proc/pid/oom_adj for a kthread will now return -EINVAL since
these threads are immune from oom killing already. They simply report an
oom_adj value of OOM_DISABLE.
Cc: Nick Piggin <npiggin@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mel@csn.ul.ie>
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>
General description: kmemcheck is a patch to the linux kernel that
detects use of uninitialized memory. It does this by trapping every
read and write to memory that was allocated dynamically (e.g. using
kmalloc()). If a memory address is read that has not previously been
written to, a message is printed to the kernel log.
Thanks to Andi Kleen for the set_memory_4k() solution.
Andrew Morton suggested documenting the shadow member of struct page.
Signed-off-by: Vegard Nossum <vegardno@ifi.uio.no>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
[export kmemcheck_mark_initialized]
[build fix for setup_max_cpus]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
[rebased for mainline inclusion]
Signed-off-by: Vegard Nossum <vegardno@ifi.uio.no>
This fixes a build failure with generic debug pagealloc:
mm/debug-pagealloc.c: In function 'set_page_poison':
mm/debug-pagealloc.c:8: error: 'struct page' has no member named 'debug_flags'
mm/debug-pagealloc.c: In function 'clear_page_poison':
mm/debug-pagealloc.c:13: error: 'struct page' has no member named 'debug_flags'
mm/debug-pagealloc.c: In function 'page_poison':
mm/debug-pagealloc.c:18: error: 'struct page' has no member named 'debug_flags'
mm/debug-pagealloc.c: At top level:
mm/debug-pagealloc.c:120: error: redefinition of 'kernel_map_pages'
include/linux/mm.h:1278: error: previous definition of 'kernel_map_pages' was here
mm/debug-pagealloc.c: In function 'kernel_map_pages':
mm/debug-pagealloc.c:122: error: 'debug_pagealloc_enabled' undeclared (first use in this function)
by fixing
- debug_flags should be in struct page
- define DEBUG_PAGEALLOC config option for all architectures
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Reported-by: Alexander Beregalov <a.beregalov@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
CONFIG_DEBUG_PAGEALLOC is now supported by x86, powerpc, sparc64, and
s390. This patch implements it for the rest of the architectures by
filling the pages with poison byte patterns after free_pages() and
verifying the poison patterns before alloc_pages().
This generic one cannot detect invalid page accesses immediately but
invalid read access may cause invalid dereference by poisoned memory and
invalid write access can be detected after a long delay.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This allows us to change the representation (to a dangling bitmap or
cpumask_var_t) without breaking all the callers: they can use
mm_cpumask() now and won't see a difference as the changes roll into
linux-next.
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
NOMMU mmap allocates a piece of memory for an mmap that's rounded up in size to
the nearest power-of-2 number of pages. Currently it then discards the excess
pages back to the page allocator, making that memory available for use by other
things. This can, however, cause greater amount of fragmentation.
To counter this, a sysctl is added in order to fine-tune the trimming
behaviour. The default behaviour remains to trim pages aggressively, while
this can either be disabled completely or set to a higher page-granular
watermark in order to have finer-grained control.
vm region vm_top bits taken from an earlier patch by David Howells.
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
Make VMAs per mm_struct as for MMU-mode linux. This solves two problems:
(1) In SYSV SHM where nattch for a segment does not reflect the number of
shmat's (and forks) done.
(2) In mmap() where the VMA's vm_mm is set to point to the parent mm by an
exec'ing process when VM_EXECUTABLE is specified, regardless of the fact
that a VMA might be shared and already have its vm_mm assigned to another
process or a dead process.
A new struct (vm_region) is introduced to track a mapped region and to remember
the circumstances under which it may be shared and the vm_list_struct structure
is discarded as it's no longer required.
This patch makes the following additional changes:
(1) Regions are now allocated with alloc_pages() rather than kmalloc() and
with no recourse to __GFP_COMP, so the pages are not composite. Instead,
each page has a reference on it held by the region. Anything else that is
interested in such a page will have to get a reference on it to retain it.
When the pages are released due to unmapping, each page is passed to
put_page() and will be freed when the page usage count reaches zero.
(2) Excess pages are trimmed after an allocation as the allocation must be
made as a power-of-2 quantity of pages.
(3) VMAs are added to the parent MM's R/B tree and mmap lists. As an MM may
end up with overlapping VMAs within the tree, the VMA struct address is
appended to the sort key.
(4) Non-anonymous VMAs are now added to the backing inode's prio list.
(5) Holes may be punched in anonymous VMAs with munmap(), releasing parts of
the backing region. The VMA and region structs will be split if
necessary.
(6) sys_shmdt() only releases one attachment to a SYSV IPC shared memory
segment instead of all the attachments at that addresss. Multiple
shmat()'s return the same address under NOMMU-mode instead of different
virtual addresses as under MMU-mode.
(7) Core dumping for ELF-FDPIC requires fewer exceptions for NOMMU-mode.
(8) /proc/maps is now the global list of mapped regions, and may list bits
that aren't actually mapped anywhere.
(9) /proc/meminfo gains a line (tagged "MmapCopy") that indicates the amount
of RAM currently allocated by mmap to hold mappable regions that can't be
mapped directly. These are copies of the backing device or file if not
anonymous.
These changes make NOMMU mode more similar to MMU mode. The downside is that
NOMMU mode requires some extra memory to track things over NOMMU without this
patch (VMAs are no longer shared, and there are now region structs).
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Mike Frysinger <vapier.adi@gmail.com>
Acked-by: Paul Mundt <lethal@linux-sh.org>
The mm->ioctx_list is currently protected by a reader-writer lock,
so we always grab that lock on the read side for doing ioctx
lookups. As the workload is extremely reader biased, turn this into
an rcu hlist so we can make lookup_ioctx() lockless. Get rid of
the rwlock and use a spinlock for providing update side exclusion.
There's usually only 1 entry on this list, so it doesn't make sense
to look into fancier data structures.
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
Allocate all page_cgroup at boot and remove page_cgroup poitner from
struct page. This patch adds an interface as
struct page_cgroup *lookup_page_cgroup(struct page*)
All FLATMEM/DISCONTIGMEM/SPARSEMEM and MEMORY_HOTPLUG is supported.
Remove page_cgroup pointer reduces the amount of memory by
- 4 bytes per PAGE_SIZE.
- 8 bytes per PAGE_SIZE
if memory controller is disabled. (even if configured.)
On usual 8GB x86-32 server, this saves 8MB of NORMAL_ZONE memory.
On my x86-64 server with 48GB of memory, this saves 96MB of memory.
I think this reduction makes sense.
By pre-allocation, kmalloc/kfree in charge/uncharge are removed.
This means
- we're not necessary to be afraid of kmalloc faiulre.
(this can happen because of gfp_mask type.)
- we can avoid calling kmalloc/kfree.
- we can avoid allocating tons of small objects which can be fragmented.
- we can know what amount of memory will be used for this extra-lru handling.
I added printk message as
"allocated %ld bytes of page_cgroup"
"please try cgroup_disable=memory option if you don't want"
maybe enough informative for users.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Define USE_SPLIT_PTLOCKS as a constant expression rather than repeating
"NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS" all over the place.
Signed-off-by: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Try to comment away a little of the confusion between mm's vm_area_struct
vm_flags and vmalloc's vm_struct flags: based on an idea by Ulrich Drepper.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With KVM/GFP/XPMEM there isn't just the primary CPU MMU pointing to pages.
There are secondary MMUs (with secondary sptes and secondary tlbs) too.
sptes in the kvm case are shadow pagetables, but when I say spte in
mmu-notifier context, I mean "secondary pte". In GRU case there's no
actual secondary pte and there's only a secondary tlb because the GRU
secondary MMU has no knowledge about sptes and every secondary tlb miss
event in the MMU always generates a page fault that has to be resolved by
the CPU (this is not the case of KVM where the a secondary tlb miss will
walk sptes in hardware and it will refill the secondary tlb transparently
to software if the corresponding spte is present). The same way
zap_page_range has to invalidate the pte before freeing the page, the spte
(and secondary tlb) must also be invalidated before any page is freed and
reused.
Currently we take a page_count pin on every page mapped by sptes, but that
means the pages can't be swapped whenever they're mapped by any spte
because they're part of the guest working set. Furthermore a spte unmap
event can immediately lead to a page to be freed when the pin is released
(so requiring the same complex and relatively slow tlb_gather smp safe
logic we have in zap_page_range and that can be avoided completely if the
spte unmap event doesn't require an unpin of the page previously mapped in
the secondary MMU).
The mmu notifiers allow kvm/GRU/XPMEM to attach to the tsk->mm and know
when the VM is swapping or freeing or doing anything on the primary MMU so
that the secondary MMU code can drop sptes before the pages are freed,
avoiding all page pinning and allowing 100% reliable swapping of guest
physical address space. Furthermore it avoids the code that teardown the
mappings of the secondary MMU, to implement a logic like tlb_gather in
zap_page_range that would require many IPI to flush other cpu tlbs, for
each fixed number of spte unmapped.
To make an example: if what happens on the primary MMU is a protection
downgrade (from writeable to wrprotect) the secondary MMU mappings will be
invalidated, and the next secondary-mmu-page-fault will call
get_user_pages and trigger a do_wp_page through get_user_pages if it
called get_user_pages with write=1, and it'll re-establishing an updated
spte or secondary-tlb-mapping on the copied page. Or it will setup a
readonly spte or readonly tlb mapping if it's a guest-read, if it calls
get_user_pages with write=0. This is just an example.
This allows to map any page pointed by any pte (and in turn visible in the
primary CPU MMU), into a secondary MMU (be it a pure tlb like GRU, or an
full MMU with both sptes and secondary-tlb like the shadow-pagetable layer
with kvm), or a remote DMA in software like XPMEM (hence needing of
schedule in XPMEM code to send the invalidate to the remote node, while no
need to schedule in kvm/gru as it's an immediate event like invalidating
primary-mmu pte).
At least for KVM without this patch it's impossible to swap guests
reliably. And having this feature and removing the page pin allows
several other optimizations that simplify life considerably.
Dependencies:
1) mm_take_all_locks() to register the mmu notifier when the whole VM
isn't doing anything with "mm". This allows mmu notifier users to keep
track if the VM is in the middle of the invalidate_range_begin/end
critical section with an atomic counter incraese in range_begin and
decreased in range_end. No secondary MMU page fault is allowed to map
any spte or secondary tlb reference, while the VM is in the middle of
range_begin/end as any page returned by get_user_pages in that critical
section could later immediately be freed without any further
->invalidate_page notification (invalidate_range_begin/end works on
ranges and ->invalidate_page isn't called immediately before freeing
the page). To stop all page freeing and pagetable overwrites the
mmap_sem must be taken in write mode and all other anon_vma/i_mmap
locks must be taken too.
2) It'd be a waste to add branches in the VM if nobody could possibly
run KVM/GRU/XPMEM on the kernel, so mmu notifiers will only enabled if
CONFIG_KVM=m/y. In the current kernel kvm won't yet take advantage of
mmu notifiers, but this already allows to compile a KVM external module
against a kernel with mmu notifiers enabled and from the next pull from
kvm.git we'll start using them. And GRU/XPMEM will also be able to
continue the development by enabling KVM=m in their config, until they
submit all GRU/XPMEM GPLv2 code to the mainline kernel. Then they can
also enable MMU_NOTIFIERS in the same way KVM does it (even if KVM=n).
This guarantees nobody selects MMU_NOTIFIER=y if KVM and GRU and XPMEM
are all =n.
The mmu_notifier_register call can fail because mm_take_all_locks may be
interrupted by a signal and return -EINTR. Because mmu_notifier_reigster
is used when a driver startup, a failure can be gracefully handled. Here
an example of the change applied to kvm to register the mmu notifiers.
Usually when a driver startups other allocations are required anyway and
-ENOMEM failure paths exists already.
struct kvm *kvm_arch_create_vm(void)
{
struct kvm *kvm = kzalloc(sizeof(struct kvm), GFP_KERNEL);
+ int err;
if (!kvm)
return ERR_PTR(-ENOMEM);
INIT_LIST_HEAD(&kvm->arch.active_mmu_pages);
+ kvm->arch.mmu_notifier.ops = &kvm_mmu_notifier_ops;
+ err = mmu_notifier_register(&kvm->arch.mmu_notifier, current->mm);
+ if (err) {
+ kfree(kvm);
+ return ERR_PTR(err);
+ }
+
return kvm;
}
mmu_notifier_unregister returns void and it's reliable.
The patch also adds a few needed but missing includes that would prevent
kernel to compile after these changes on non-x86 archs (x86 didn't need
them by luck).
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: fix mm/filemap_xip.c build]
[akpm@linux-foundation.org: fix mm/mmu_notifier.c build]
Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: Jack Steiner <steiner@sgi.com>
Cc: Robin Holt <holt@sgi.com>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Kanoj Sarcar <kanojsarcar@yahoo.com>
Cc: Roland Dreier <rdreier@cisco.com>
Cc: Steve Wise <swise@opengridcomputing.com>
Cc: Avi Kivity <avi@qumranet.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Chris Wright <chrisw@redhat.com>
Cc: Marcelo Tosatti <marcelo@kvack.org>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: "Paul E. McKenney" <paulmck@us.ibm.com>
Cc: Izik Eidus <izike@qumranet.com>
Cc: Anthony Liguori <aliguori@us.ibm.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Now that we have core_state->dumper list we can use it to wake up the
sub-threads waiting for the coredump completion.
This uglifies the code and .text grows by 47 bytes, but otoh mm_struct
lessens by sizeof(struct completion). Also, with this change we can
decouple exit_mm() from the coredumping code.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
binfmt->core_dump() has to iterate over the all threads in system in order
to find the coredumping threads and construct the list using the
GFP_ATOMIC allocations.
With this patch each thread allocates the list node on exit_mm()'s stack and
adds itself to the list.
This allows us to do further changes:
- simplify ->core_dump()
- change exit_mm() to clear ->mm first, then wait for ->core_done.
this makes the coredumping process visible to oom_kill
- kill mm->core_done
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Acked-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Turn core_state->nr_threads into atomic_t and kill now unneeded
down_write(&mm->mmap_sem) in exit_mm().
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Move mm->core_waiters into "struct core_state" allocated on stack. This
shrinks mm_struct a little bit and allows further changes.
This patch mostly does s/core_waiters/core_state. The only essential
change is that coredump_wait() must clear mm->core_state before return.
The coredump_wait()'s path is uglified and .text grows by 30 bytes, this
is fixed by the next patch.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm->core_startup_done points to "struct completion startup_done" allocated
on the coredump_wait()'s stack. Introduce the new structure, core_state,
which holds this "struct completion". This way we can add more info
visible to the threads participating in coredump without enlarging
mm_struct.
No changes in affected .o files.
Signed-off-by: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When mm destruction happens, we should pass mm_update_next_owner() the old mm.
But unfortunately new mm is passed in exec_mmap().
Thus, kernel panic is possible when a multi-threaded process uses exec().
Also, the owner member comment description is wrong. mm->owner does not
necessarily point to the thread group leader.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: "Paul Menage" <menage@google.com>
Cc: "KAMEZAWA Hiroyuki" <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The kernel implements readlink of /proc/pid/exe by getting the file from
the first executable VMA. Then the path to the file is reconstructed and
reported as the result.
Because of the VMA walk the code is slightly different on nommu systems.
This patch avoids separate /proc/pid/exe code on nommu systems. Instead of
walking the VMAs to find the first executable file-backed VMA we store a
reference to the exec'd file in the mm_struct.
That reference would prevent the filesystem holding the executable file
from being unmounted even after unmapping the VMAs. So we track the number
of VM_EXECUTABLE VMAs and drop the new reference when the last one is
unmapped. This avoids pinning the mounted filesystem.
[akpm@linux-foundation.org: improve comments]
[yamamoto@valinux.co.jp: fix dup_mmap]
Signed-off-by: Matt Helsley <matthltc@us.ibm.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: David Howells <dhowells@redhat.com>
Cc:"Eric W. Biederman" <ebiederm@xmission.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Remove the mem_cgroup member from mm_struct and instead adds an owner.
This approach was suggested by Paul Menage. The advantage of this approach
is that, once the mm->owner is known, using the subsystem id, the cgroup
can be determined. It also allows several control groups that are
virtually grouped by mm_struct, to exist independent of the memory
controller i.e., without adding mem_cgroup's for each controller, to
mm_struct.
A new config option CONFIG_MM_OWNER is added and the memory resource
controller selects this config option.
This patch also adds cgroup callbacks to notify subsystems when mm->owner
changes. The mm_cgroup_changed callback is called with the task_lock() of
the new task held and is called just prior to changing the mm->owner.
I am indebted to Paul Menage for the several reviews of this patchset and
helping me make it lighter and simpler.
This patch was tested on a powerpc box, it was compiled with both the
MM_OWNER config turned on and off.
After the thread group leader exits, it's moved to init_css_state by
cgroup_exit(), thus all future charges from runnings threads would be
redirected to the init_css_set's subsystem.
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Sudhir Kumar <skumar@linux.vnet.ibm.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: David Rientjes <rientjes@google.com>,
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Reviewed-by: Paul Menage <menage@google.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/slab-2.6:
slub: pack objects denser
slub: Calculate min_objects based on number of processors.
slub: Drop DEFAULT_MAX_ORDER / DEFAULT_MIN_OBJECTS
slub: Simplify any_slab_object checks
slub: Make the order configurable for each slab cache
slub: Drop fallback to page allocator method
slub: Fallback to minimal order during slab page allocation
slub: Update statistics handling for variable order slabs
slub: Add kmem_cache_order_objects struct
slub: for_each_object must be passed the number of objects in a slab
slub: Store max number of objects in the page struct.
slub: Dump list of objects not freed on kmem_cache_close()
slub: free_list() cleanup
slub: improve kmem_cache_destroy() error message
slob: fix bug - when slob allocates "struct kmem_cache", it does not force alignment.
Save some bytes in mm_struct by filling holes
Putting int values together for better packing on 64bit shrinks sizeof(struct
mm_struct) from 776 bytes to 764 bytes.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Split the inuse field up to be able to store the number of objects in this
page in the page struct as well. Necessary if we want to have pages of
various orders for a slab. Also avoids touching struct kmem_cache cachelines in
__slab_alloc().
Update diagnostic code to check the number of objects and make sure that
the number of objects always stays within the bounds of a 16 bit unsigned
integer.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Rename Memory Controller to Memory Resource Controller. Reflect the same
changes in the CONFIG definition for the Memory Resource Controller. Group
together the config options for Resource Counters and Memory Resource
Controller.
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This only made sense for the alternate fastpath which was reverted last week.
Mathieu is working on a new version that addresses the fastpath issues but that
new code first needs to go through mm and it is not clear if we need the
unique end pointers with his new scheme.
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Christoph Lameter <clameter@sgi.com>
We use a NULL pointer on freelists to signal that there are no more objects.
However the NULL pointers of all slabs match in contrast to the pointers to
the real objects which are in different ranges for different slab pages.
Change the end pointer to be a pointer to the first object and set bit 0.
Every slab will then have a different end pointer. This is necessary to ensure
that end markers can be matched to the source slab during cmpxchg_local.
Bring back the use of the mapping field by SLUB since we would otherwise have
to call a relatively expensive function page_address() in __slab_alloc(). Use
of the mapping field allows avoiding a call to page_address() in various other
functions as well.
There is no need to change the page_mapping() function since bit 0 is set on
the mapping as also for anonymous pages. page_mapping(slab_page) will
therefore still return NULL although the mapping field is overloaded.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Basic setup routines, the mm_struct has a pointer to the cgroup that
it belongs to and the the page has a page_cgroup associated with it.
Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
include/asm-powerpc/elf.h has 6 entries in ARCH_DLINFO. fs/binfmt_elf.c
has 14 unconditional NEW_AUX_ENT entries and 2 conditional NEW_AUX_ENT
entries. So in the worst case, saved_auxv does not get an AT_NULL entry at
the end.
The saved_auxv array must be terminated with an AT_NULL entry. Make the
size of mm_struct->saved_auxv arch dependend, based on the number of
ARCH_DLINFO entries.
Signed-off-by: Olaf Hering <olh@suse.de>
Cc: Roland McGrath <roland@redhat.com>
Cc: Jakub Jelinek <jakub@redhat.com>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
We need the offset from the page struct during slab_alloc and slab_free. In
both cases we also reference the cacheline of the kmem_cache_cpu structure.
We can therefore move the offset field into the kmem_cache_cpu structure
freeing up 16 bits in the page struct.
Moving the offset allows an allocation from slab_alloc() without touching the
page struct in the hot path.
The only thing left in slab_free() that touches the page struct cacheline for
per cpu freeing is the checking of SlabDebug(page). The next patch deals with
that.
Use the available 16 bits to broaden page->inuse. More than 64k objects per
slab become possible and we can get rid of the checks for that limitation.
No need anymore to shrink the order of slabs if we boot with 2M sized slabs
(slub_min_order=9).
No need anymore to switch off the offset calculation for very large slabs
since the field in the kmem_cache_cpu structure is 32 bits and so the offset
field can now handle slab sizes of up to 8GB.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After moving the lockless_freelist to kmem_cache_cpu we no longer need
page->lockless_freelist. Restructure the use of the struct page fields in
such a way that we never touch the mapping field.
This is turn allows us to remove the special casing of SLUB when determining
the mapping of a page (needed for corner cases of virtual caches machines that
need to flush caches of processors mapping a page).
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Move the definitions of struct mm_struct and struct vma_area_struct to
include/mm_types.h. This allows to define more function in asm/pgtable.h
and friends with inline assemblies instead of macros. Compile tested on
i386, powerpc, powerpc64, s390-32, s390-64 and x86_64.
[aurelien@aurel32.net: build fix]
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Aurelien Jarno <aurelien@aurel32.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Avoid atomic overhead in slab_alloc and slab_free
SLUB needs to use the slab_lock for the per cpu slabs to synchronize with
potential kfree operations. This patch avoids that need by moving all free
objects onto a lockless_freelist. The regular freelist continues to exist
and will be used to free objects. So while we consume the
lockless_freelist the regular freelist may build up objects.
If we are out of objects on the lockless_freelist then we may check the
regular freelist. If it has objects then we move those over to the
lockless_freelist and do this again. There is a significant savings in
terms of atomic operations that have to be performed.
We can even free directly to the lockless_freelist if we know that we are
running on the same processor. So this speeds up short lived objects.
They may be allocated and freed without taking the slab_lock. This is
particular good for netperf.
In order to maximize the effect of the new faster hotpath we extract the
hottest performance pieces into inlined functions. These are then inlined
into kmem_cache_alloc and kmem_cache_free. So hotpath allocation and
freeing no longer requires a subroutine call within SLUB.
[I am not sure that it is worth doing this because it changes the easy to
read structure of slub just to reduce atomic ops. However, there is
someone out there with a benchmark on 4 way and 8 way processor systems
that seems to show a 5% regression vs. Slab. Seems that the regression is
due to increased atomic operations use vs. SLAB in SLUB). I wonder if
this is applicable or discernable at all in a real workload?]
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is a new slab allocator which was motivated by the complexity of the
existing code in mm/slab.c. It attempts to address a variety of concerns
with the existing implementation.
A. Management of object queues
A particular concern was the complex management of the numerous object
queues in SLAB. SLUB has no such queues. Instead we dedicate a slab for
each allocating CPU and use objects from a slab directly instead of
queueing them up.
B. Storage overhead of object queues
SLAB Object queues exist per node, per CPU. The alien cache queue even
has a queue array that contain a queue for each processor on each
node. For very large systems the number of queues and the number of
objects that may be caught in those queues grows exponentially. On our
systems with 1k nodes / processors we have several gigabytes just tied up
for storing references to objects for those queues This does not include
the objects that could be on those queues. One fears that the whole
memory of the machine could one day be consumed by those queues.
C. SLAB meta data overhead
SLAB has overhead at the beginning of each slab. This means that data
cannot be naturally aligned at the beginning of a slab block. SLUB keeps
all meta data in the corresponding page_struct. Objects can be naturally
aligned in the slab. F.e. a 128 byte object will be aligned at 128 byte
boundaries and can fit tightly into a 4k page with no bytes left over.
SLAB cannot do this.
D. SLAB has a complex cache reaper
SLUB does not need a cache reaper for UP systems. On SMP systems
the per CPU slab may be pushed back into partial list but that
operation is simple and does not require an iteration over a list
of objects. SLAB expires per CPU, shared and alien object queues
during cache reaping which may cause strange hold offs.
E. SLAB has complex NUMA policy layer support
SLUB pushes NUMA policy handling into the page allocator. This means that
allocation is coarser (SLUB does interleave on a page level) but that
situation was also present before 2.6.13. SLABs application of
policies to individual slab objects allocated in SLAB is
certainly a performance concern due to the frequent references to
memory policies which may lead a sequence of objects to come from
one node after another. SLUB will get a slab full of objects
from one node and then will switch to the next.
F. Reduction of the size of partial slab lists
SLAB has per node partial lists. This means that over time a large
number of partial slabs may accumulate on those lists. These can
only be reused if allocator occur on specific nodes. SLUB has a global
pool of partial slabs and will consume slabs from that pool to
decrease fragmentation.
G. Tunables
SLAB has sophisticated tuning abilities for each slab cache. One can
manipulate the queue sizes in detail. However, filling the queues still
requires the uses of the spin lock to check out slabs. SLUB has a global
parameter (min_slab_order) for tuning. Increasing the minimum slab
order can decrease the locking overhead. The bigger the slab order the
less motions of pages between per CPU and partial lists occur and the
better SLUB will be scaling.
G. Slab merging
We often have slab caches with similar parameters. SLUB detects those
on boot up and merges them into the corresponding general caches. This
leads to more effective memory use. About 50% of all caches can
be eliminated through slab merging. This will also decrease
slab fragmentation because partial allocated slabs can be filled
up again. Slab merging can be switched off by specifying
slub_nomerge on boot up.
Note that merging can expose heretofore unknown bugs in the kernel
because corrupted objects may now be placed differently and corrupt
differing neighboring objects. Enable sanity checks to find those.
H. Diagnostics
The current slab diagnostics are difficult to use and require a
recompilation of the kernel. SLUB contains debugging code that
is always available (but is kept out of the hot code paths).
SLUB diagnostics can be enabled via the "slab_debug" option.
Parameters can be specified to select a single or a group of
slab caches for diagnostics. This means that the system is running
with the usual performance and it is much more likely that
race conditions can be reproduced.
I. Resiliency
If basic sanity checks are on then SLUB is capable of detecting
common error conditions and recover as best as possible to allow the
system to continue.
J. Tracing
Tracing can be enabled via the slab_debug=T,<slabcache> option
during boot. SLUB will then protocol all actions on that slabcache
and dump the object contents on free.
K. On demand DMA cache creation.
Generally DMA caches are not needed. If a kmalloc is used with
__GFP_DMA then just create this single slabcache that is needed.
For systems that have no ZONE_DMA requirement the support is
completely eliminated.
L. Performance increase
Some benchmarks have shown speed improvements on kernbench in the
range of 5-10%. The locking overhead of slub is based on the
underlying base allocation size. If we can reliably allocate
larger order pages then it is possible to increase slub
performance much further. The anti-fragmentation patches may
enable further performance increases.
Tested on:
i386 UP + SMP, x86_64 UP + SMP + NUMA emulation, IA64 NUMA + Simulator
SLUB Boot options
slub_nomerge Disable merging of slabs
slub_min_order=x Require a minimum order for slab caches. This
increases the managed chunk size and therefore
reduces meta data and locking overhead.
slub_min_objects=x Mininum objects per slab. Default is 8.
slub_max_order=x Avoid generating slabs larger than order specified.
slub_debug Enable all diagnostics for all caches
slub_debug=<options> Enable selective options for all caches
slub_debug=<o>,<cache> Enable selective options for a certain set of
caches
Available Debug options
F Double Free checking, sanity and resiliency
R Red zoning
P Object / padding poisoning
U Track last free / alloc
T Trace all allocs / frees (only use for individual slabs).
To use SLUB: Apply this patch and then select SLUB as the default slab
allocator.
[hugh@veritas.com: fix an oops-causing locking error]
[akpm@linux-foundation.org: various stupid cleanups and small fixes]
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>