It apepars that this patch was innocent, and we hope that "mm: avoid
waking kswapd for THP allocations when compaction is deferred or
contended" will fix the final kswapd-spinning cause.
Cc: Zdenek Kabelac <zkabelac@redhat.com>
Cc: Seth Jennings <sjenning@linux.vnet.ibm.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Jiri Slaby <jirislaby@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Robert Jennings <rcj@linux.vnet.ibm.com>
Cc: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With "mm: vmscan: scale number of pages reclaimed by reclaim/compaction
based on failures" reverted, Zdenek Kabelac reported the following
Hmm, so it's just took longer to hit the problem and observe
kswapd0 spinning on my CPU again - it's not as endless like before -
but still it easily eats minutes - it helps to turn off Firefox
or TB (memory hungry apps) so kswapd0 stops soon - and restart
those apps again. (And I still have like >1GB of cached memory)
kswapd0 R running task 0 30 2 0x00000000
Call Trace:
preempt_schedule+0x42/0x60
_raw_spin_unlock+0x55/0x60
put_super+0x31/0x40
drop_super+0x22/0x30
prune_super+0x149/0x1b0
shrink_slab+0xba/0x510
The sysrq+m indicates the system has no swap so it'll never reclaim
anonymous pages as part of reclaim/compaction. That is one part of the
problem but not the root cause as file-backed pages could also be
reclaimed.
The likely underlying problem is that kswapd is woken up or kept awake
for each THP allocation request in the page allocator slow path.
If compaction fails for the requesting process then compaction will be
deferred for a time and direct reclaim is avoided. However, if there
are a storm of THP requests that are simply rejected, it will still be
the the case that kswapd is awake for a prolonged period of time as
pgdat->kswapd_max_order is updated each time. This is noticed by the
main kswapd() loop and it will not call kswapd_try_to_sleep(). Instead
it will loopp, shrinking a small number of pages and calling
shrink_slab() on each iteration.
The temptation is to supply a patch that checks if kswapd was woken for
THP and if so ignore pgdat->kswapd_max_order but it'll be a hack and not
backed up by proper testing. As 3.7 is very close to release and this
is not a bug we should release with, a safer path is to revert "mm:
remove __GFP_NO_KSWAPD" for now and revisit it with the view to ironing
out the balance_pgdat() logic in general.
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Zdenek Kabelac <zkabelac@redhat.com>
Cc: Seth Jennings <sjenning@linux.vnet.ibm.com>
Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu>
Cc: Jiri Slaby <jirislaby@gmail.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Robert Jennings <rcj@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There was a general sentiment in a recent discussion (See
https://lkml.org/lkml/2012/9/18/258) that the __GFP flags should be
defined unconditionally. Currently, the only offender is GFP_NOTRACK,
which is conditional to KMEMCHECK.
Signed-off-by: Glauber Costa <glommer@parallels.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Mel Gorman <mgorman@suse.de>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When transparent huge pages were introduced, memory compaction and swap
storms were an issue, and the kernel had to be careful to not make THP
allocations cause pageout or compaction.
Now that we have working compaction deferral, kswapd is smart enough to
invoke compaction and the quadratic behaviour around isolate_free_pages
has been fixed, it should be safe to remove __GFP_NO_KSWAPD.
[minchan@kernel.org: Comment fix]
[mgorman@suse.de: Avoid direct reclaim for deferred compaction]
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Change the skb allocation API to indicate RX usage and use this to fall
back to the PFMEMALLOC reserve when needed. SKBs allocated from the
reserve are tagged in skb->pfmemalloc. If an SKB is allocated from the
reserve and the socket is later found to be unrelated to page reclaim, the
packet is dropped so that the memory remains available for page reclaim.
Network protocols are expected to recover from this packet loss.
[a.p.zijlstra@chello.nl: Ideas taken from various patches]
[davem@davemloft.net: Use static branches, coding style corrections]
[sebastian@breakpoint.cc: Avoid unnecessary cast, fix !CONFIG_NET build]
Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: David S. Miller <davem@davemloft.net>
Cc: Neil Brown <neilb@suse.de>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
__GFP_MEMALLOC will allow the allocation to disregard the watermarks, much
like PF_MEMALLOC. It allows one to pass along the memalloc state in
object related allocation flags as opposed to task related flags, such as
sk->sk_allocation. This removes the need for ALLOC_PFMEMALLOC as callers
using __GFP_MEMALLOC can get the ALLOC_NO_WATERMARK flag which is now
enough to identify allocations related to page reclaim.
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: David Miller <davem@davemloft.net>
Cc: Neil Brown <neilb@suse.de>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: Eric B Munson <emunson@mgebm.net>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This commit changes various functions that change pages and
pageblocks migrate type between MIGRATE_ISOLATE and
MIGRATE_MOVABLE in such a way as to allow to work with
MIGRATE_CMA migrate type.
Signed-off-by: Michal Nazarewicz <mina86@mina86.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Tested-by: Rob Clark <rob.clark@linaro.org>
Tested-by: Ohad Ben-Cohen <ohad@wizery.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Tested-by: Robert Nelson <robertcnelson@gmail.com>
Tested-by: Barry Song <Baohua.Song@csr.com>
The MIGRATE_CMA migration type has two main characteristics:
(i) only movable pages can be allocated from MIGRATE_CMA
pageblocks and (ii) page allocator will never change migration
type of MIGRATE_CMA pageblocks.
This guarantees (to some degree) that page in a MIGRATE_CMA page
block can always be migrated somewhere else (unless there's no
memory left in the system).
It is designed to be used for allocating big chunks (eg. 10MiB)
of physically contiguous memory. Once driver requests
contiguous memory, pages from MIGRATE_CMA pageblocks may be
migrated away to create a contiguous block.
To minimise number of migrations, MIGRATE_CMA migration type
is the last type tried when page allocator falls back to other
migration types when requested.
Signed-off-by: Michal Nazarewicz <mina86@mina86.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Tested-by: Rob Clark <rob.clark@linaro.org>
Tested-by: Ohad Ben-Cohen <ohad@wizery.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Tested-by: Robert Nelson <robertcnelson@gmail.com>
Tested-by: Barry Song <Baohua.Song@csr.com>
This commit adds the alloc_contig_range() function which tries
to allocate given range of pages. It tries to migrate all
already allocated pages that fall in the range thus freeing them.
Once all pages in the range are freed they are removed from the
buddy system thus allocated for the caller to use.
Signed-off-by: Michal Nazarewicz <mina86@mina86.com>
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Tested-by: Rob Clark <rob.clark@linaro.org>
Tested-by: Ohad Ben-Cohen <ohad@wizery.com>
Tested-by: Benjamin Gaignard <benjamin.gaignard@linaro.org>
Tested-by: Robert Nelson <robertcnelson@gmail.com>
Tested-by: Barry Song <Baohua.Song@csr.com>
The maximum number of dirty pages that exist in the system at any time is
determined by a number of pages considered dirtyable and a user-configured
percentage of those, or an absolute number in bytes.
This number of dirtyable pages is the sum of memory provided by all the
zones in the system minus their lowmem reserves and high watermarks, so
that the system can retain a healthy number of free pages without having
to reclaim dirty pages.
But there is a flaw in that we have a zoned page allocator which does not
care about the global state but rather the state of individual memory
zones. And right now there is nothing that prevents one zone from filling
up with dirty pages while other zones are spared, which frequently leads
to situations where kswapd, in order to restore the watermark of free
pages, does indeed have to write pages from that zone's LRU list. This
can interfere so badly with IO from the flusher threads that major
filesystems (btrfs, xfs, ext4) mostly ignore write requests from reclaim
already, taking away the VM's only possibility to keep such a zone
balanced, aside from hoping the flushers will soon clean pages from that
zone.
Enter per-zone dirty limits. They are to a zone's dirtyable memory what
the global limit is to the global amount of dirtyable memory, and try to
make sure that no single zone receives more than its fair share of the
globally allowed dirty pages in the first place. As the number of pages
considered dirtyable excludes the zones' lowmem reserves and high
watermarks, the maximum number of dirty pages in a zone is such that the
zone can always be balanced without requiring page cleaning.
As this is a placement decision in the page allocator and pages are
dirtied only after the allocation, this patch allows allocators to pass
__GFP_WRITE when they know in advance that the page will be written to and
become dirty soon. The page allocator will then attempt to allocate from
the first zone of the zonelist - which on NUMA is determined by the task's
NUMA memory policy - that has not exceeded its dirty limit.
At first glance, it would appear that the diversion to lower zones can
increase pressure on them, but this is not the case. With a full high
zone, allocations will be diverted to lower zones eventually, so it is
more of a shift in timing of the lower zone allocations. Workloads that
previously could fit their dirty pages completely in the higher zone may
be forced to allocate from lower zones, but the amount of pages that
"spill over" are limited themselves by the lower zones' dirty constraints,
and thus unlikely to become a problem.
For now, the problem of unfair dirty page distribution remains for NUMA
configurations where the zones allowed for allocation are in sum not big
enough to trigger the global dirty limits, wake up the flusher threads and
remedy the situation. Because of this, an allocation that could not
succeed on any of the considered zones is allowed to ignore the dirty
limits before going into direct reclaim or even failing the allocation,
until a future patch changes the global dirty throttling and flusher
thread activation so that they take individual zone states into account.
Test results
15M DMA + 3246M DMA32 + 504 Normal = 3765M memory
40% dirty ratio
16G USB thumb drive
10 runs of dd if=/dev/zero of=disk/zeroes bs=32k count=$((10 << 15))
seconds nr_vmscan_write
(stddev) min| median| max
xfs
vanilla: 549.747( 3.492) 0.000| 0.000| 0.000
patched: 550.996( 3.802) 0.000| 0.000| 0.000
fuse-ntfs
vanilla: 1183.094(53.178) 54349.000| 59341.000| 65163.000
patched: 558.049(17.914) 0.000| 0.000| 43.000
btrfs
vanilla: 573.679(14.015) 156657.000| 460178.000| 606926.000
patched: 563.365(11.368) 0.000| 0.000| 1362.000
ext4
vanilla: 561.197(15.782) 0.000|2725438.000|4143837.000
patched: 568.806(17.496) 0.000| 0.000| 0.000
Signed-off-by: Johannes Weiner <jweiner@redhat.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Tested-by: Wu Fengguang <fengguang.wu@intel.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Chris Mason <chris.mason@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Calling alloc_pages_exact_node() means the allocation only passes the
zonelist of a single node into the page allocator. If that node isn't
online, it's zonelist may never have been initialized causing a strange
oops that may not immediately be clear.
I recently debugged an issue where node 0 wasn't online and an allocator
was passing 0 to alloc_pages_exact_node() and it resulted in a NULL
pointer on zonelist->_zoneref. If CONFIG_DEBUG_VM is enabled, though, it
would be nice to catch this a bit earlier.
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Colin Cross reported;
Under the following conditions, __alloc_pages_slowpath can loop forever:
gfp_mask & __GFP_WAIT is true
gfp_mask & __GFP_FS is false
reclaim and compaction make no progress
order <= PAGE_ALLOC_COSTLY_ORDER
These conditions happen very often during suspend and resume,
when pm_restrict_gfp_mask() effectively converts all GFP_KERNEL
allocations into __GFP_WAIT.
The oom killer is not run because gfp_mask & __GFP_FS is false,
but should_alloc_retry will always return true when order is less
than PAGE_ALLOC_COSTLY_ORDER.
In his fix, he avoided retrying the allocation if reclaim made no progress
and __GFP_FS was not set. The problem is that this would result in
GFP_NOIO allocations failing that previously succeeded which would be very
unfortunate.
The big difference between GFP_NOIO and suspend converting GFP_KERNEL to
behave like GFP_NOIO is that normally flushers will be cleaning pages and
kswapd reclaims pages allowing GFP_NOIO to succeed after a short delay.
The same does not necessarily apply during suspend as the storage device
may be suspended.
This patch special cases the suspend case to fail the page allocation if
reclaim cannot make progress and adds some documentation on how
gfp_allowed_mask is currently used. Failing allocations like this may
cause suspend to abort but that is better than a livelock.
[mgorman@suse.de: Rework fix to be suspend specific]
[rientjes@google.com: Move suspended device check to should_alloc_retry]
Reported-by: Colin Cross <ccross@android.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
__GFP_OTHER_NODE is used for NUMA allocations on behalf of other nodes.
It's supposed to be passed through from the page allocator to
zone_statistics(), but it never gets there as gfp_allowed_mask is not
wide enough and masks out the flag early in the allocation path.
The result is an accounting glitch where successful NUMA allocations
by-agent are not properly attributed as local.
Increase __GFP_BITS_SHIFT so that it includes __GFP_OTHER_NODE.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Mel Gorman <mgorman@suse.de>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
VM_BUG_ON() if effectively a BUG_ON() undef #ifdef CONFIG_DEBUG_VM. That
is exactly what we have here now, and two different folks have suggested
doing it this way.
Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Running sparse on page_alloc.c today, it errors out:
include/linux/gfp.h:254:17: error: bad constant expression
include/linux/gfp.h:254:17: error: cannot size expression
which is a line in gfp_zone():
BUILD_BUG_ON((GFP_ZONE_BAD >> bit) & 1);
That's really unfortunate, because it ends up hiding all of the other
legitimate sparse messages like this:
mm/page_alloc.c:5315:59: warning: incorrect type in argument 1 (different base types)
mm/page_alloc.c:5315:59: expected unsigned long [unsigned] [usertype] size
mm/page_alloc.c:5315:59: got restricted gfp_t [usertype] <noident>
...
Having sparse be able to catch these very oopsable bugs is a lot more
important than keeping a BUILD_BUG_ON(). Kill the BUILD_BUG_ON().
Compiles on x86_64 with and without CONFIG_DEBUG_VM=y. defconfig boots
fine for me.
Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add a alloc_pages_exact_nid() that allocates on a specific node.
The naming is quite broken, but fixing that would need a larger renaming
action.
[akpm@linux-foundation.org: coding-style fixes]
[akpm@linux-foundation.org: tweak comment]
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add a new __GFP_OTHER_NODE flag to tell the low level numa statistics in
zone_statistics() that an allocation is on behalf of another thread. This
way the local and remote counters can be still correct, even when
background daemons like khugepaged are changing memory mappings.
This only affects the accounting, but I think it's worth doing that right
to avoid confusing users.
I first tried to just pass down the right node, but this required a lot of
changes to pass down this parameter and at least one addition of a 10th
argument to a 9 argument function. Using the flag is a lot less
intrusive.
Open: should be also used for migration?
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add a alloc_page_vma_node that allows passing the "local" node in. Used
in a followon patch.
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: 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>
Currently alloc_pages_vma() always uses the local node as policy node for
the LOCAL policy. Pass this node down as an argument instead.
No behaviour change from this patch, but will be needed for followons.
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: 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>
It's mostly a matter of replacing alloc_pages with alloc_pages_vma after
introducing alloc_pages_vma. khugepaged needs special handling as the
allocation has to happen inside collapse_huge_page where the vma is known
and an error has to be returned to the outer loop to sleep
alloc_sleep_millisecs in case of failure. But it retains the more
efficient logic of handling allocation failures in khugepaged in case of
CONFIG_NUMA=n.
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Lately I've been working to make KVM use hugepages transparently without
the usual restrictions of hugetlbfs. Some of the restrictions I'd like to
see removed:
1) hugepages have to be swappable or the guest physical memory remains
locked in RAM and can't be paged out to swap
2) if a hugepage allocation fails, regular pages should be allocated
instead and mixed in the same vma without any failure and without
userland noticing
3) if some task quits and more hugepages become available in the
buddy, guest physical memory backed by regular pages should be
relocated on hugepages automatically in regions under
madvise(MADV_HUGEPAGE) (ideally event driven by waking up the
kernel deamon if the order=HPAGE_PMD_SHIFT-PAGE_SHIFT list becomes
not null)
4) avoidance of reservation and maximization of use of hugepages whenever
possible. Reservation (needed to avoid runtime fatal faliures) may be ok for
1 machine with 1 database with 1 database cache with 1 database cache size
known at boot time. It's definitely not feasible with a virtualization
hypervisor usage like RHEV-H that runs an unknown number of virtual machines
with an unknown size of each virtual machine with an unknown amount of
pagecache that could be potentially useful in the host for guest not using
O_DIRECT (aka cache=off).
hugepages in the virtualization hypervisor (and also in the guest!) are
much more important than in a regular host not using virtualization,
becasue with NPT/EPT they decrease the tlb-miss cacheline accesses from 24
to 19 in case only the hypervisor uses transparent hugepages, and they
decrease the tlb-miss cacheline accesses from 19 to 15 in case both the
linux hypervisor and the linux guest both uses this patch (though the
guest will limit the addition speedup to anonymous regions only for
now...). Even more important is that the tlb miss handler is much slower
on a NPT/EPT guest than for a regular shadow paging or no-virtualization
scenario. So maximizing the amount of virtual memory cached by the TLB
pays off significantly more with NPT/EPT than without (even if there would
be no significant speedup in the tlb-miss runtime).
The first (and more tedious) part of this work requires allowing the VM to
handle anonymous hugepages mixed with regular pages transparently on
regular anonymous vmas. This is what this patch tries to achieve in the
least intrusive possible way. We want hugepages and hugetlb to be used in
a way so that all applications can benefit without changes (as usual we
leverage the KVM virtualization design: by improving the Linux VM at
large, KVM gets the performance boost too).
The most important design choice is: always fallback to 4k allocation if
the hugepage allocation fails! This is the _very_ opposite of some large
pagecache patches that failed with -EIO back then if a 64k (or similar)
allocation failed...
Second important decision (to reduce the impact of the feature on the
existing pagetable handling code) is that at any time we can split an
hugepage into 512 regular pages and it has to be done with an operation
that can't fail. This way the reliability of the swapping isn't decreased
(no need to allocate memory when we are short on memory to swap) and it's
trivial to plug a split_huge_page* one-liner where needed without
polluting the VM. Over time we can teach mprotect, mremap and friends to
handle pmd_trans_huge natively without calling split_huge_page*. The fact
it can't fail isn't just for swap: if split_huge_page would return -ENOMEM
(instead of the current void) we'd need to rollback the mprotect from the
middle of it (ideally including undoing the split_vma) which would be a
big change and in the very wrong direction (it'd likely be simpler not to
call split_huge_page at all and to teach mprotect and friends to handle
hugepages instead of rolling them back from the middle). In short the
very value of split_huge_page is that it can't fail.
The collapsing and madvise(MADV_HUGEPAGE) part will remain separated and
incremental and it'll just be an "harmless" addition later if this initial
part is agreed upon. It also should be noted that locking-wise replacing
regular pages with hugepages is going to be very easy if compared to what
I'm doing below in split_huge_page, as it will only happen when
page_count(page) matches page_mapcount(page) if we can take the PG_lock
and mmap_sem in write mode. collapse_huge_page will be a "best effort"
that (unlike split_huge_page) can fail at the minimal sign of trouble and
we can try again later. collapse_huge_page will be similar to how KSM
works and the madvise(MADV_HUGEPAGE) will work similar to
madvise(MADV_MERGEABLE).
The default I like is that transparent hugepages are used at page fault
time. This can be changed with
/sys/kernel/mm/transparent_hugepage/enabled. The control knob can be set
to three values "always", "madvise", "never" which mean respectively that
hugepages are always used, or only inside madvise(MADV_HUGEPAGE) regions,
or never used. /sys/kernel/mm/transparent_hugepage/defrag instead
controls if the hugepage allocation should defrag memory aggressively
"always", only inside "madvise" regions, or "never".
The pmd_trans_splitting/pmd_trans_huge locking is very solid. The
put_page (from get_user_page users that can't use mmu notifier like
O_DIRECT) that runs against a __split_huge_page_refcount instead was a
pain to serialize in a way that would result always in a coherent page
count for both tail and head. I think my locking solution with a
compound_lock taken only after the page_first is valid and is still a
PageHead should be safe but it surely needs review from SMP race point of
view. In short there is no current existing way to serialize the O_DIRECT
final put_page against split_huge_page_refcount so I had to invent a new
one (O_DIRECT loses knowledge on the mapping status by the time gup_fast
returns so...). And I didn't want to impact all gup/gup_fast users for
now, maybe if we change the gup interface substantially we can avoid this
locking, I admit I didn't think too much about it because changing the gup
unpinning interface would be invasive.
If we ignored O_DIRECT we could stick to the existing compound refcounting
code, by simply adding a get_user_pages_fast_flags(foll_flags) where KVM
(and any other mmu notifier user) would call it without FOLL_GET (and if
FOLL_GET isn't set we'd just BUG_ON if nobody registered itself in the
current task mmu notifier list yet). But O_DIRECT is fundamental for
decent performance of virtualized I/O on fast storage so we can't avoid it
to solve the race of put_page against split_huge_page_refcount to achieve
a complete hugepage feature for KVM.
Swap and oom works fine (well just like with regular pages ;). MMU
notifier is handled transparently too, with the exception of the young bit
on the pmd, that didn't have a range check but I think KVM will be fine
because the whole point of hugepages is that EPT/NPT will also use a huge
pmd when they notice gup returns pages with PageCompound set, so they
won't care of a range and there's just the pmd young bit to check in that
case.
NOTE: in some cases if the L2 cache is small, this may slowdown and waste
memory during COWs because 4M of memory are accessed in a single fault
instead of 8k (the payoff is that after COW the program can run faster).
So we might want to switch the copy_huge_page (and clear_huge_page too) to
not temporal stores. I also extensively researched ways to avoid this
cache trashing with a full prefault logic that would cow in 8k/16k/32k/64k
up to 1M (I can send those patches that fully implemented prefault) but I
concluded they're not worth it and they add an huge additional complexity
and they remove all tlb benefits until the full hugepage has been faulted
in, to save a little bit of memory and some cache during app startup, but
they still don't improve substantially the cache-trashing during startup
if the prefault happens in >4k chunks. One reason is that those 4k pte
entries copied are still mapped on a perfectly cache-colored hugepage, so
the trashing is the worst one can generate in those copies (cow of 4k page
copies aren't so well colored so they trashes less, but again this results
in software running faster after the page fault). Those prefault patches
allowed things like a pte where post-cow pages were local 4k regular anon
pages and the not-yet-cowed pte entries were pointing in the middle of
some hugepage mapped read-only. If it doesn't payoff substantially with
todays hardware it will payoff even less in the future with larger l2
caches, and the prefault logic would blot the VM a lot. If one is
emebdded transparent_hugepage can be disabled during boot with sysfs or
with the boot commandline parameter transparent_hugepage=0 (or
transparent_hugepage=2 to restrict hugepages inside madvise regions) that
will ensure not a single hugepage is allocated at boot time. It is simple
enough to just disable transparent hugepage globally and let transparent
hugepages be allocated selectively by applications in the MADV_HUGEPAGE
region (both at page fault time, and if enabled with the
collapse_huge_page too through the kernel daemon).
This patch supports only hugepages mapped in the pmd, archs that have
smaller hugepages will not fit in this patch alone. Also some archs like
power have certain tlb limits that prevents mixing different page size in
the same regions so they will not fit in this framework that requires
"graceful fallback" to basic PAGE_SIZE in case of physical memory
fragmentation. hugetlbfs remains a perfect fit for those because its
software limits happen to match the hardware limits. hugetlbfs also
remains a perfect fit for hugepage sizes like 1GByte that cannot be hoped
to be found not fragmented after a certain system uptime and that would be
very expensive to defragment with relocation, so requiring reservation.
hugetlbfs is the "reservation way", the point of transparent hugepages is
not to have any reservation at all and maximizing the use of cache and
hugepages at all times automatically.
Some performance result:
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largep
ages3
memset page fault 1566023
memset tlb miss 453854
memset second tlb miss 453321
random access tlb miss 41635
random access second tlb miss 41658
vmx andrea # LD_PRELOAD=/usr/lib64/libhugetlbfs.so HUGETLB_MORECORE=yes HUGETLB_PATH=/mnt/huge/ ./largepages3
memset page fault 1566471
memset tlb miss 453375
memset second tlb miss 453320
random access tlb miss 41636
random access second tlb miss 41637
vmx andrea # ./largepages3
memset page fault 1566642
memset tlb miss 453417
memset second tlb miss 453313
random access tlb miss 41630
random access second tlb miss 41647
vmx andrea # ./largepages3
memset page fault 1566872
memset tlb miss 453418
memset second tlb miss 453315
random access tlb miss 41618
random access second tlb miss 41659
vmx andrea # echo 0 > /proc/sys/vm/transparent_hugepage
vmx andrea # ./largepages3
memset page fault 2182476
memset tlb miss 460305
memset second tlb miss 460179
random access tlb miss 44483
random access second tlb miss 44186
vmx andrea # ./largepages3
memset page fault 2182791
memset tlb miss 460742
memset second tlb miss 459962
random access tlb miss 43981
random access second tlb miss 43988
============
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#define SIZE (3UL*1024*1024*1024)
int main()
{
char *p = malloc(SIZE), *p2;
struct timeval before, after;
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset page fault %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
memset(p, 0, SIZE);
gettimeofday(&after, NULL);
printf("memset second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
gettimeofday(&before, NULL);
for (p2 = p; p2 < p+SIZE; p2 += 4096)
*p2 = 0;
gettimeofday(&after, NULL);
printf("random access second tlb miss %Lu\n",
(after.tv_sec-before.tv_sec)*1000000UL +
after.tv_usec-before.tv_usec);
return 0;
}
============
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Transparent hugepage allocations must be allowed not to invoke kswapd or
any other kind of indirect reclaim (especially when the defrag sysfs is
control disabled). It's unacceptable to swap out anonymous pages
(potentially anonymous transparent hugepages) in order to create new
transparent hugepages. This is true for the MADV_HUGEPAGE areas too
(swapping out a kvm virtual machine and so having it suffer an unbearable
slowdown, so another one with guest physical memory marked MADV_HUGEPAGE
can run 30% faster if it is running memory intensive workloads, makes no
sense). If a transparent hugepage allocation fails the slowdown is minor
and there is total fallback, so kswapd should never be asked to swapout
memory to allow the high order allocation to succeed.
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There is a problem that swap pages allocated before the creation of
a hibernation image can be released and used for storing the contents
of different memory pages while the image is being saved. Since the
kernel stored in the image doesn't know of that, it causes memory
corruption to occur after resume from hibernation, especially on
systems with relatively small RAM that need to swap often.
This issue can be addressed by keeping the GFP_IOFS bits clear
in gfp_allowed_mask during the entire hibernation, including the
saving of the image, until the system is finally turned off or
the hibernation is aborted. Unfortunately, for this purpose
it's necessary to rework the way in which the hibernate and
suspend code manipulates gfp_allowed_mask.
This change is based on an earlier patch from Hugh Dickins.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Reported-by: Ondrej Zary <linux@rainbow-software.org>
Acked-by: Hugh Dickins <hughd@google.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: stable@kernel.org
Introduce ___GFP_* masks in order for gfp_t to not be mixed with plain
integers which causes a lot of warnings like the following:
warning: restricted gfp_t degrades to integer
Signed-off-by: Namhyung Kim <namhyung@gmail.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>
Add parenthesis in a define. This doesn't change functionality.
checkpatch errors:
1) white space fixes
2) add spaces after comas
Signed-off-by: matt mooney <mfm@muteddisk.com>
Cc: Dan Carpenter <error27@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fix minor spelling errors in a few comments; no code changes.
Signed-off-by: matt mooney <mfm@muteddisk.com>
Cc: Dan Carpenter <error27@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
__GFP_NOFAIL was deprecated in dab48dab, so add a comment that no new
users should be added.
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
There are quite a few GFP_KERNEL memory allocations made during
suspend/hibernation and resume that may cause the system to hang, because
the I/O operations they depend on cannot be completed due to the
underlying devices being suspended.
Avoid this problem by clearing the __GFP_IO and __GFP_FS bits in
gfp_allowed_mask before suspend/hibernation and restoring the original
values of these bits in gfp_allowed_mask durig the subsequent resume.
[akpm@linux-foundation.org: fix CONFIG_PM=n linkage]
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Reported-by: Maxim Levitsky <maximlevitsky@gmail.com>
Cc: Sebastian Ott <sebott@linux.vnet.ibm.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
free_hot_page() is just a wrapper around free_hot_cold_page() with
parameter 'cold = 0'. After adding a clear comment for
free_hot_cold_page(), it is reasonable to remove a level of call.
[akpm@linux-foundation.org: fix build]
Signed-off-by: Li Hong <lihong.hi@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Rik van Riel <riel@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Larry Woodman <lwoodman@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Li Ming Chun <macli@brc.ubc.ca>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Americo Wang <xiyou.wangcong@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
gcc permitting variable length arrays makes the current construct used for
BUILD_BUG_ON() useless, as that doesn't produce any diagnostic if the
controlling expression isn't really constant. Instead, this patch makes
it so that a bit field gets used here. Consequently, those uses where the
condition isn't really constant now also need fixing.
Note that in the gfp.h, kmemcheck.h, and virtio_config.h cases
MAYBE_BUILD_BUG_ON() really just serves documentation purposes - even if
the expression is compile time constant (__builtin_constant_p() yields
true), the array is still deemed of variable length by gcc, and hence the
whole expression doesn't have the intended effect.
[akpm@linux-foundation.org: make arch/sparc/include/asm/vio.h compile]
[akpm@linux-foundation.org: more nonsensical assertions in tpm.c..]
Signed-off-by: Jan Beulich <jbeulich@novell.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Rajiv Andrade <srajiv@linux.vnet.ibm.com>
Cc: Mimi Zohar <zohar@us.ibm.com>
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The function free_cold_page() has no callers so delete it.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The page allocator also needs the masking of gfp flags during boot,
so this moves it out of slab/slub and uses it with the page allocator
as well.
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
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
This simplifies the code in gfp_zone() and also keeps the ability of the
compiler to use constant folding to get rid of gfp_zone processing.
The lookup of the zone is done using a bitfield stored in an integer. So
the code in gfp_zone is a simple extraction of bits from a constant
bitfield. The compiler is generating a load of a constant into a register
and then performs a shift and mask operation to get the zone from a gfp_t.
No cachelines are touched and no branches have to be predicted by the
compiler.
We are doing some macro tricks here to convince the compiler to always do
the constant folding if possible.
Signed-off-by: Christoph Lameter <cl@linux-foundation.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Mel Gorman <mel@csn.ul.ie>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently, the following scenario appears to be possible in theory:
* Tasks are frozen for hibernation or suspend.
* Free pages are almost exhausted.
* Certain piece of code in the suspend code path attempts to allocate
some memory using GFP_KERNEL and allocation order less than or
equal to PAGE_ALLOC_COSTLY_ORDER.
* __alloc_pages_internal() cannot find a free page so it invokes the
OOM killer.
* The OOM killer attempts to kill a task, but the task is frozen, so
it doesn't die immediately.
* __alloc_pages_internal() jumps to 'restart', unsuccessfully tries
to find a free page and invokes the OOM killer.
* No progress can be made.
Although it is now hard to trigger during hibernation due to the memory
shrinking carried out by the hibernation code, it is theoretically
possible to trigger during suspend after the memory shrinking has been
removed from that code path. Moreover, since memory allocations are
going to be used for the hibernation memory shrinking, it will be even
more likely to happen during hibernation.
To prevent it from happening, introduce the oom_killer_disabled switch
that will cause __alloc_pages_internal() to fail in the situations in
which the OOM killer would have been called and make the freezer set
this switch after tasks have been successfully frozen.
[akpm@linux-foundation.org: be nicer to the namespace]
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Cc: Fengguang Wu <fengguang.wu@gmail.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Pavel Machek <pavel@ucw.cz>
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>
Callers of alloc_pages_node() can optionally specify -1 as a node to mean
"allocate from the current node". However, a number of the callers in
fast paths know for a fact their node is valid. To avoid a comparison and
branch, this patch adds alloc_pages_exact_node() that only checks the nid
with VM_BUG_ON(). Callers that know their node is valid are then
converted.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Acked-by: Paul Mundt <lethal@linux-sh.org> [for the SLOB NUMA bits]
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
No user of the allocator API should be passing in an order >= MAX_ORDER
but we check for it on each and every allocation. Delete this check and
make it a VM_BUG_ON check further down the call path.
[akpm@linux-foundation.org: s/VM_BUG_ON/WARN_ON_ONCE/]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The start of a large patch series to clean up and optimise the page
allocator.
The performance improvements are in a wide range depending on the exact
machine but the results I've seen so fair are approximately;
kernbench: 0 to 0.12% (elapsed time)
0.49% to 3.20% (sys time)
aim9: -4% to 30% (for page_test and brk_test)
tbench: -1% to 4%
hackbench: -2.5% to 3.45% (mostly within the noise though)
netperf-udp -1.34% to 4.06% (varies between machines a bit)
netperf-tcp -0.44% to 5.22% (varies between machines a bit)
I haven't sysbench figures at hand, but previously they were within the
-0.5% to 2% range.
On netperf, the client and server were bound to opposite number CPUs to
maximise the problems with cache line bouncing of the struct pages so I
expect different people to report different results for netperf depending
on their exact machine and how they ran the test (different machines, same
cpus client/server, shared cache but two threads client/server, different
socket client/server etc).
I also measured the vmlinux sizes for a single x86-based config with
CONFIG_DEBUG_INFO enabled but not CONFIG_DEBUG_VM. The core of the
.config is based on the Debian Lenny kernel config so I expect it to be
reasonably typical.
This patch:
__alloc_pages_internal is the core page allocator function but essentially
it is an alias of __alloc_pages_nodemask. Naming a publicly available and
exported function "internal" is also a big ugly. This patch renames
__alloc_pages_internal() to __alloc_pages_nodemask() and deletes the old
nodemask function.
Warning - This patch renames an exported symbol. No kernel driver is
affected by external drivers calling __alloc_pages_internal() should
change the call to __alloc_pages_nodemask() without any alteration of
parameters.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Christoph Lameter <cl@linux-foundation.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This adds support for tracking the initializedness of memory that
was allocated with the page allocator. Highmem requests are not
tracked.
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
[build fix for !CONFIG_KMEMCHECK]
Signed-off-by: Ingo Molnar <mingo@elte.hu>
[rebased for mainline inclusion]
Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
With kmemcheck enabled, the slab allocator needs to do this:
1. Tell kmemcheck to allocate the shadow memory which stores the status of
each byte in the allocation proper, e.g. whether it is initialized or
uninitialized.
2. Tell kmemcheck which parts of memory that should be marked uninitialized.
There are actually a few more states, such as "not yet allocated" and
"recently freed".
If a slab cache is set up using the SLAB_NOTRACK flag, it will never return
memory that can take page faults because of kmemcheck.
If a slab cache is NOT set up using the SLAB_NOTRACK flag, callers can still
request memory with the __GFP_NOTRACK flag. This does not prevent the page
faults from occuring, however, but marks the object in question as being
initialized so that no warnings will ever be produced for this object.
In addition to (and in contrast to) __GFP_NOTRACK, the
__GFP_NOTRACK_FALSE_POSITIVE flag indicates that the allocation should
not be tracked _because_ it would produce a false positive. Their values
are identical, but need not be so in the future (for example, we could now
enable/disable false positives with a config option).
Parts of this patch were contributed by Pekka Enberg but merged for
atomicity.
Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
[rebased for mainline inclusion]
Signed-off-by: Vegard Nossum <vegard.nossum@gmail.com>
As explained by Benjamin Herrenschmidt:
Oh and btw, your patch alone doesn't fix powerpc, because it's missing
a whole bunch of GFP_KERNEL's in the arch code... You would have to
grep the entire kernel for things that check slab_is_available() and
even then you'll be missing some.
For example, slab_is_available() didn't always exist, and so in the
early days on powerpc, we used a mem_init_done global that is set form
mem_init() (not perfect but works in practice). And we still have code
using that to do the test.
Therefore, mask out __GFP_WAIT, __GFP_IO, and __GFP_FS in the slab allocators
in early boot code to avoid enabling interrupts.
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
Impact: cleanup, potential bugfix
Not sure what changed to expose this, but clearly that numa_node_id()
doesn't belong in mmzone.h (the inline in gfp.h is probably overkill, too).
In file included from include/linux/topology.h:34,
from arch/x86/mm/numa.c:2:
/home/rusty/patches-cpumask/linux-2.6/arch/x86/include/asm/topology.h:64:1: warning: "numa_node_id" redefined
In file included from include/linux/topology.h:32,
from arch/x86/mm/numa.c:2:
include/linux/mmzone.h:770:1: warning: this is the location of the previous definition
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Mike Travis <travis@sgi.com>
LKML-Reference: <200903132343.37661.rusty@rustcorp.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
GFP_HIGHUSER_PAGECACHE is just an alias for GFP_HIGHUSER_MOVABLE, making
that harder to track down: remove it, and its out-of-work brothers
GFP_NOFS_PAGECACHE and GFP_USER_PAGECACHE.
Since we're making that improvement to hotremove_migrate_alloc(), I think
we can now also remove one of the "o"s from its comment.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
alloc_pages_exact() is similar to alloc_pages(), except that it allocates
the minimum number of pages to fulfill the request. This is useful if you
want to allocate a very large buffer that is slightly larger than an even
power-of-two number of pages. In that case, alloc_pages() will waste a
lot of memory.
I have a video driver that wants to allocate a 5MB buffer. alloc_pages()
wiill waste 3MB of physically-contiguous memory.
Signed-off-by: Timur Tabi <timur@freescale.com>
Cc: Andi Kleen <andi@firstfloor.org>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Two zonelist patch series rewrote __page_alloc() largely. Now, it is just
a wrapper function. Inlining them will save a function call.
[akpm@linux-foundation.org: export __alloc_pages_internal]
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The definition and use of __GFP_REPEAT, __GFP_NOFAIL and __GFP_NORETRY in the
core VM have somewhat differing comments as to their actual semantics.
Annoyingly, the flags definition has inline and header comments, which might
be interpreted as not being equivalent. Just add references to the header
comments in the inline ones so they don't go out of sync in the future. In
their use in __alloc_pages() clarify that the current implementation treats
low-order allocations and __GFP_REPEAT allocations as distinct cases.
To clarify, the flags' semantics are:
__GFP_NORETRY means try no harder than one run through __alloc_pages
__GFP_REPEAT means __GFP_NOFAIL
__GFP_NOFAIL means repeat forever
order <= PAGE_ALLOC_COSTLY_ORDER means __GFP_NOFAIL
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This hack, "base = MAX_NR_ZONES", at __GFP_THISNODE was used for old
zonliests.
Now, new zonelist[] have a list for __GFP_THISNODE and this hack is incorrect.
Should be removed.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The MPOL_BIND policy creates a zonelist that is used for allocations
controlled by that mempolicy. As the per-node zonelist is already being
filtered based on a zone id, this patch adds a version of __alloc_pages() that
takes a nodemask for further filtering. This eliminates the need for
MPOL_BIND to create a custom zonelist.
A positive benefit of this is that allocations using MPOL_BIND now use the
local node's distance-ordered zonelist instead of a custom node-id-ordered
zonelist. I.e., pages will be allocated from the closest allowed node with
available memory.
[Lee.Schermerhorn@hp.com: Mempolicy: update stale documentation and comments]
[Lee.Schermerhorn@hp.com: Mempolicy: make dequeue_huge_page_vma() obey MPOL_BIND nodemask]
[Lee.Schermerhorn@hp.com: Mempolicy: make dequeue_huge_page_vma() obey MPOL_BIND nodemask rework]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently a node has two sets of zonelists, one for each zone type in the
system and a second set for GFP_THISNODE allocations. Based on the zones
allowed by a gfp mask, one of these zonelists is selected. All of these
zonelists consume memory and occupy cache lines.
This patch replaces the multiple zonelists per-node with two zonelists. The
first contains all populated zones in the system, ordered by distance, for
fallback allocations when the target/preferred node has no free pages. The
second contains all populated zones in the node suitable for GFP_THISNODE
allocations.
An iterator macro is introduced called for_each_zone_zonelist() that interates
through each zone allowed by the GFP flags in the selected zonelist.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Introduce a node_zonelist() helper function. It is used to lookup the
appropriate zonelist given a node and a GFP mask. The patch on its own is a
cleanup but it helps clarify parts of the two-zonelist-per-node patchset. If
necessary, it can be merged with the next patch in this set without problems.
Reviewed-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Migrate flags must be set on slab creation as agreed upon when the antifrag
logic was reviewed. Otherwise some slabs of a slabcache will end up in the
unmovable and others in the reclaimable section depending on which flag was
active when a new slab page was allocated.
This likely slid in somehow when antifrag was merged. Remove it.
The buffer_heads are always allocated with __GFP_RECLAIMABLE because the
SLAB_RECLAIM_ACCOUNT option is set. The set_migrateflags() never had any
effect there.
Radix tree allocations are not directly reclaimable but they are allocated
with __GFP_RECLAIMABLE set on each allocation. We now set
SLAB_RECLAIM_ACCOUNT on radix tree slab creation making sure that radix
tree slabs are consistently placed in the reclaimable section. Radix tree
slabs will also be accounted as such.
There is then no user left of set_migratepages. So remove it.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
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>
FASTCALL() is always expanded to empty, remove it.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
- Add comments explaing how drain_pages() works.
- Eliminate useless functions
- Rename drain_all_local_pages to drain_all_pages(). It does drain
all pages not only those of the local processor.
- Eliminate useless interrupt off / on sequences. drain_pages()
disables interrupts on its own. The execution thread is
pinned to processor by the caller. So there is no need to
disable interrupts.
- Put drain_all_pages() declaration in gfp.h and remove the
declarations from suspend.h and from mm/memory_hotplug.c
- Make software suspend call drain_all_pages(). The draining
of processor local pages is may not the right approach if
software suspend wants to support SMP. If they call drain_all_pages
then we can make drain_pages() static.
[akpm@linux-foundation.org: fix build]
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Daniel Walker <dwalker@mvista.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch provides fragmentation avoidance statistics via /proc/pagetypeinfo.
The information is collected only on request so there is no runtime overhead.
The statistics are in three parts:
The first part prints information on the size of blocks that pages are
being grouped on and looks like
Page block order: 10
Pages per block: 1024
The second part is a more detailed version of /proc/buddyinfo and looks like
Free pages count per migrate type at order 0 1 2 3 4 5 6 7 8 9 10
Node 0, zone DMA, type Unmovable 0 0 0 0 0 0 0 0 0 0 0
Node 0, zone DMA, type Reclaimable 1 0 0 0 0 0 0 0 0 0 0
Node 0, zone DMA, type Movable 0 0 0 0 0 0 0 0 0 0 0
Node 0, zone DMA, type Reserve 0 4 4 0 0 0 0 1 0 1 0
Node 0, zone Normal, type Unmovable 111 8 4 4 2 3 1 0 0 0 0
Node 0, zone Normal, type Reclaimable 293 89 8 0 0 0 0 0 0 0 0
Node 0, zone Normal, type Movable 1 6 13 9 7 6 3 0 0 0 0
Node 0, zone Normal, type Reserve 0 0 0 0 0 0 0 0 0 0 4
The third part looks like
Number of blocks type Unmovable Reclaimable Movable Reserve
Node 0, zone DMA 0 1 2 1
Node 0, zone Normal 3 17 94 4
To walk the zones within a node with interrupts disabled, walk_zones_in_node()
is introduced and shared between /proc/buddyinfo, /proc/zoneinfo and
/proc/pagetypeinfo to reduce code duplication. It seems specific to what
vmstat.c requires but could be broken out as a general utility function in
mmzone.c if there were other other potential users.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Acked-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 patch marks a number of allocations that are either short-lived such as
network buffers or are reclaimable such as inode allocations. When something
like updatedb is called, long-lived and unmovable kernel allocations tend to
be spread throughout the address space which increases fragmentation.
This patch groups these allocations together as much as possible by adding a
new MIGRATE_TYPE. The MIGRATE_RECLAIMABLE type is for allocations that can be
reclaimed on demand, but not moved. i.e. they can be migrated by deleting
them and re-reading the information from elsewhere.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The function of GFP_LEVEL_MASK seems to be unclear. In order to clear up
the mystery we get rid of it and replace GFP_LEVEL_MASK with 3 sets of GFP
flags:
GFP_RECLAIM_MASK Flags used to control page allocator reclaim behavior.
GFP_CONSTRAINT_MASK Flags used to limit where allocations can occur.
GFP_SLAB_BUG_MASK Flags that the slab allocator BUG()s on.
These replace the uses of GFP_LEVEL mask in the slab allocators and in
vmalloc.c.
The use of the flags not included in these sets may occur as a result of a
slab allocation standing in for a page allocation when constructing scatter
gather lists. Extraneous flags are cleared and not passed through to the
page allocator. __GFP_MOVABLE/RECLAIMABLE, __GFP_COLD and __GFP_COMP will
now be ignored if passed to a slab allocator.
Change the allocation of allocator meta data in SLAB and vmalloc to not
pass through flags listed in GFP_CONSTRAINT_MASK. SLAB already removes the
__GFP_THISNODE flag for such allocations. Generalize that to also cover
vmalloc. The use of GFP_CONSTRAINT_MASK also includes __GFP_HARDWALL.
The impact of allocator metadata placement on access latency to the
cachelines of the object itself is minimal since metadata is only
referenced on alloc and free. The attempt is still made to place the meta
data optimally but we consistently allow fallback both in SLAB and vmalloc
(SLUB does not need to allocate metadata like that).
Allocator metadata may serve multiple in kernel users and thus should not
be subject to the limitations arising from a single allocation context.
[akpm@linux-foundation.org: fix fallback_alloc()]
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>
GFP_THISNODE checks that the zone selected is within the pgdat (node) of the
first zone of a nodelist. That only works if the node has memory. A
memoryless node will have its first node on another pgdat (node).
GFP_THISNODE currently will return simply memory on the first pgdat. Thus it
is returning memory on other nodes. GFP_THISNODE should fail if there is no
local memory on a node.
Add a new set of zonelists for each node that only contain the nodes that
belong to the zones itself so that no fallback is possible.
Then modify gfp_type to pickup the right zone based on the presence of
__GFP_THISNODE.
Drop the existing GFP_THISNODE checks from the page_allocators hot path.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Tested-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Bob Picco <bob.picco@hp.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@skynet.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The following 8 patches against 2.6.20-mm2 create a zone called ZONE_MOVABLE
that is only usable by allocations that specify both __GFP_HIGHMEM and
__GFP_MOVABLE. This has the effect of keeping all non-movable pages within a
single memory partition while allowing movable allocations to be satisfied
from either partition. The patches may be applied with the list-based
anti-fragmentation patches that groups pages together based on mobility.
The size of the zone is determined by a kernelcore= parameter specified at
boot-time. This specifies how much memory is usable by non-movable
allocations and the remainder is used for ZONE_MOVABLE. Any range of pages
within ZONE_MOVABLE can be released by migrating the pages or by reclaiming.
When selecting a zone to take pages from for ZONE_MOVABLE, there are two
things to consider. First, only memory from the highest populated zone is
used for ZONE_MOVABLE. On the x86, this is probably going to be ZONE_HIGHMEM
but it would be ZONE_DMA on ppc64 or possibly ZONE_DMA32 on x86_64. Second,
the amount of memory usable by the kernel will be spread evenly throughout
NUMA nodes where possible. If the nodes are not of equal size, the amount of
memory usable by the kernel on some nodes may be greater than others.
By default, the zone is not as useful for hugetlb allocations because they are
pinned and non-migratable (currently at least). A sysctl is provided that
allows huge pages to be allocated from that zone. This means that the huge
page pool can be resized to the size of ZONE_MOVABLE during the lifetime of
the system assuming that pages are not mlocked. Despite huge pages being
non-movable, we do not introduce additional external fragmentation of note as
huge pages are always the largest contiguous block we care about.
Credit goes to Andy Whitcroft for catching a large variety of problems during
review of the patches.
This patch creates an additional zone, ZONE_MOVABLE. This zone is only usable
by allocations which specify both __GFP_HIGHMEM and __GFP_MOVABLE. Hot-added
memory continues to be placed in their existing destination as there is no
mechanism to redirect them to a specific zone.
[y-goto@jp.fujitsu.com: Fix section mismatch of memory hotplug related code]
[akpm@linux-foundation.org: various fixes]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It is often known at allocation time whether a page may be migrated or not.
This patch adds a flag called __GFP_MOVABLE and a new mask called
GFP_HIGH_MOVABLE. Allocations using the __GFP_MOVABLE can be either migrated
using the page migration mechanism or reclaimed by syncing with backing
storage and discarding.
An API function very similar to alloc_zeroed_user_highpage() is added for
__GFP_MOVABLE allocations called alloc_zeroed_user_highpage_movable(). The
flags used by alloc_zeroed_user_highpage() are not changed because it would
change the semantics of an existing API. After this patch is applied there
are no in-kernel users of alloc_zeroed_user_highpage() so it probably should
be marked deprecated if this patch is merged.
Note that this patch includes a minor cleanup to the use of __GFP_ZERO in
shmem.c to keep all flag modifications to inode->mapping in the
shmem_dir_alloc() helper function. This clean-up suggestion is courtesy of
Hugh Dickens.
Additional credit goes to Christoph Lameter and Linus Torvalds for shaping the
concept. Credit to Hugh Dickens for catching issues with shmem swap vector
and ramfs allocations.
[akpm@linux-foundation.org: build fix]
[hugh@veritas.com: __GFP_ZERO cleanup]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently the slab allocators contain callbacks into the page allocator to
perform the draining of pagesets on remote nodes. This requires SLUB to have
a whole subsystem in order to be compatible with SLAB. Moving node draining
out of the slab allocators avoids a section of code in SLUB.
Move the node draining so that is is done when the vm statistics are updated.
At that point we are already touching all the cachelines with the pagesets of
a processor.
Add a expire counter there. If we have to update per zone or global vm
statistics then assume that the pageset will require subsequent draining.
The expire counter will be decremented on each vm stats update pass until it
reaches zero. Then we will drain one batch from the pageset. The draining
will cause vm counter updates which will then cause another expiration until
the pcp is empty. So we will drain a batch every 3 seconds.
Note that remote node draining is a somewhat esoteric feature that is required
on large NUMA systems because otherwise significant portions of system memory
can become trapped in pcp queues. The number of pcp is determined by the
number of processors and nodes in a system. A system with 4 processors and 2
nodes has 8 pcps which is okay. But a system with 1024 processors and 512
nodes has 512k pcps with a high potential for large amount of memory being
caught in them.
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>
There is no user remaining and I have never seen any use of that flag.
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>
Make ZONE_DMA optional in core code.
- ifdef all code for ZONE_DMA and related definitions following the example
for ZONE_DMA32 and ZONE_HIGHMEM.
- Without ZONE_DMA, ZONE_HIGHMEM and ZONE_DMA32 we get to a ZONES_SHIFT of
0.
- Modify the VM statistics to work correctly without a DMA zone.
- Modify slab to not create DMA slabs if there is no ZONE_DMA.
[akpm@osdl.org: cleanup]
[jdike@addtoit.com: build fix]
[apw@shadowen.org: Simplify calculation of the number of bits we need for ZONES_SHIFT]
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: Andi Kleen <ak@suse.de>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Matthew Wilcox <willy@debian.org>
Cc: James Bottomley <James.Bottomley@steeleye.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
... operations with it are OK as is, but flags & ~0 will have no idea that
this ~0 is meant to be ~gfp_t.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add an arch_alloc_page to match arch_free_page.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
GFP_THISNODE must be set to 0 in the non numa case otherwise we disable retry
and warnings for failing allocations in the SMP and UP case.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
In many places we will need to use the same combination of flags. Specify
a single GFP_THISNODE definition for ease of use in gfp.h.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add a new gfp flag __GFP_THISNODE to avoid fallback to other nodes. This
flag is essential if a kernel component requires memory to be located on a
certain node. It will be needed for alloc_pages_node() to force allocation
on the indicated node and for alloc_pages() to force allocation on the
current node.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I wonder why we need this bitmask indexing into zone->node_zonelists[]?
We always start with the highest zone and then include all lower zones
if we build zonelists.
Are there really cases where we need allocation from ZONE_DMA or
ZONE_HIGHMEM but not ZONE_NORMAL? It seems that the current implementation
of highest_zone() makes that already impossible.
If we go linear on the index then gfp_zone() == highest_zone() and a lot
of definitions fall by the wayside.
We can now revert back to the use of gfp_zone() in mempolicy.c ;-)
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There is a check in zonelist_policy that compares pieces of the bitmap
obtained from a gfp mask via GFP_ZONETYPES with a zone number in function
zonelist_policy().
The bitmap is an ORed mask of __GFP_DMA, __GFP_DMA32 and __GFP_HIGHMEM.
The policy_zone is a zone number with the possible values of ZONE_DMA,
ZONE_DMA32, ZONE_HIGHMEM and ZONE_NORMAL. These are two different domains
of values.
For some reason seemed to work before the zone reduction patchset (It
definitely works on SGI boxes since we just have one zone and the check
cannot fail).
With the zone reduction patchset this check definitely fails on systems
with two zones if the system actually has memory in both zones.
This is because ZONE_NORMAL is selected using no __GFP flag at
all and thus gfp_zone(gfpmask) == 0. ZONE_DMA is selected when __GFP_DMA
is set. __GFP_DMA is 0x01. So gfp_zone(gfpmask) == 1.
policy_zone is set to ZONE_NORMAL (==1) if ZONE_NORMAL and ZONE_DMA are
populated.
For ZONE_NORMAL gfp_zone(<no _GFP_DMA>) yields 0 which is <
policy_zone(ZONE_NORMAL) and so policy is not applied to regular memory
allocations!
Instead gfp_zone(__GFP_DMA) == 1 which results in policy being applied
to DMA allocations!
What we realy want in that place is to establish the highest allowable
zone for a given gfp_mask. If the highest zone is higher or equal to the
policy_zone then memory policies need to be applied. We have such
a highest_zone() function in page_alloc.c.
So move the highest_zone() function from mm/page_alloc.c into
include/linux/gfp.h. On the way we simplify the function and use the new
zone_type that was also introduced with the zone reduction patchset plus we
also specify the right type for the gfp flags parameter.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Make ZONE_HIGHMEM optional
- ifdef out code and definitions related to CONFIG_HIGHMEM
- __GFP_HIGHMEM falls back to normal allocations if there is no
ZONE_HIGHMEM
- GFP_ZONEMASK becomes 0x01 if there is no DMA32 and no HIGHMEM
zone.
[jdike@addtoit.com: build fix]
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Signed-off-by: Christoph Lameter <clameter@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Make ZONE_DMA32 optional
- Add #ifdefs around ZONE_DMA32 specific code and definitions.
- Add CONFIG_ZONE_DMA32 config option and use that for x86_64
that alone needs this zone.
- Remove the use of CONFIG_DMA_IS_DMA32 and CONFIG_DMA_IS_NORMAL
for ia64 and fix up the way per node ZVCs are calculated.
- Fall back to prior GFP_ZONEMASK of 0x03 if there is no
DMA32 zone.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Introduce GFP_NOWAIT, as an alias for GFP_ATOMIC & ~__GFP_HIGH.
This also changes XFS, which is the only in-tree user of this idiom that I
could find. The XFS piece is compile-tested only.
Signed-off-by: Jeff Dike <jdike@addtoit.com>
Acked-by: Nathan Scott <nathans@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The cache reaper currently tries to free all alien caches and all remote
per cpu pages in each pass of cache_reap. For a machines with large number
of nodes (such as Altix) this may lead to sporadic delays of around ~10ms.
Interrupts are disabled while reclaiming creating unacceptable delays.
This patch changes that behavior by adding a per cpu reap_node variable.
Instead of attempting to free all caches, we free only one alien cache and
the per cpu pages from one remote node. That reduces the time spend in
cache_reap. However, doing so will lengthen the time it takes to
completely drain all remote per cpu pagesets and all alien caches. The
time needed will grow with the number of nodes in the system. All caches
are drained when they overflow their respective capacity. So the drawback
here is only that a bit of memory may be wasted for awhile longer.
Details:
1. Rename drain_remote_pages to drain_node_pages to allow the specification
of the node to drain of pcp pages.
2. Add additional functions init_reap_node, next_reap_node for NUMA
that manage a per cpu reap_node counter.
3. Add a reap_alien function that reaps only from the current reap_node.
For us this seems to be a critical issue. Holdoffs of an average of ~7ms
cause some HPC benchmarks to slow down significantly. F.e. NAS parallel
slows down dramatically. NAS parallel has a 12-16 seconds runtime w/o rotor
compared to 5.8 secs with the rotor patches. It gets down to 5.05 secs with
the additional interrupt holdoff reductions.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Following kmalloc_node.
Needed for another patch to return -1 for unknown nodes in x86-64.
Cc: Christoph Lameter <clameter@engr.sgi.com>
Cc: kiran@scalex86.org
Signed-off-by: Andi Kleen <ak@suse.de>
[ Changed 0 to numa_node_id() on suggestion by Christoph Lameter ]
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Clarify in comments that GFP_ATOMIC means both "don't sleep" and "use
emergency pools", hence both ALLOC_HARDER and ALLOC_HIGH.
Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There was some confusion about the different zone usage, this should fix
up the resulting mess in the GFP zonemask handling.
The different zone usage is still confusing (it's very easy to mix up
the individual zone numbers with the GFP zone _list_ numbers), so we
might want to clean up some of this in the future, but in the meantime
this should fix the actual problems.
Acked-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add a new 4GB GFP_DMA32 zone between the GFP_DMA and GFP_NORMAL zones.
As a bit of historical background: when the x86-64 port
was originally designed we had some discussion if we should
use a 16MB DMA zone like i386 or a 4GB DMA zone like IA64 or
both. Both was ruled out at this point because it was in early
2.4 when VM is still quite shakey and had bad troubles even
dealing with one DMA zone. We settled on the 16MB DMA zone mainly
because we worried about older soundcards and the floppy.
But this has always caused problems since then because
device drivers had trouble getting enough DMA able memory. These days
the VM works much better and the wide use of NUMA has proven
it can deal with many zones successfully.
So this patch adds both zones.
This helps drivers who need a lot of memory below 4GB because
their hardware is not accessing more (graphic drivers - proprietary
and free ones, video frame buffer drivers, sound drivers etc.).
Previously they could only use IOMMU+16MB GFP_DMA, which
was not enough memory.
Another common problem is that hardware who has full memory
addressing for >4GB misses it for some control structures in memory
(like transmit rings or other metadata). They tended to allocate memory
in the 16MB GFP_DMA or the IOMMU/swiotlb then using pci_alloc_consistent,
but that can tie up a lot of precious 16MB GFPDMA/IOMMU/swiotlb memory
(even on AMD systems the IOMMU tends to be quite small) especially if you have
many devices. With the new zone pci_alloc_consistent can just put
this stuff into memory below 4GB which works better.
One argument was still if the zone should be 4GB or 2GB. The main
motivation for 2GB would be an unnamed not so unpopular hardware
raid controller (mostly found in older machines from a particular four letter
company) who has a strange 2GB restriction in firmware. But
that one works ok with swiotlb/IOMMU anyways, so it doesn't really
need GFP_DMA32. I chose 4GB to be compatible with IA64 and because
it seems to be the most common restriction.
The new zone is so far added only for x86-64.
For other architectures who don't set up this
new zone nothing changes. Architectures can set a compatibility
define in Kconfig CONFIG_DMA_IS_DMA32 that will define GFP_DMA32
as GFP_DMA. Otherwise it's a nop because on 32bit architectures
it's normally not needed because GFP_NORMAL (=0) is DMA able
enough.
One problem is still that GFP_DMA means different things on different
architectures. e.g. some drivers used to have #ifdef ia64 use GFP_DMA
(trusting it to be 4GB) #elif __x86_64__ (use other hacks like
the swiotlb because 16MB is not enough) ... . This was quite
ugly and is now obsolete.
These should be now converted to use GFP_DMA32 unconditionally. I haven't done
this yet. Or best only use pci_alloc_consistent/dma_alloc_coherent
which will use GFP_DMA32 transparently.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Remove last remnant of the defunct early reclaim page logic, the no longer
used __GFP_NORECLAIM flag bit.
Signed-off-by: Paul Jackson <pj@sgi.com>
Acked-by: Martin Hicks <mort@bork.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Beginning of gfp_t annotations:
- -Wbitwise added to CHECKFLAGS
- old __bitwise renamed to __bitwise__
- __bitwise defined to either __bitwise__ or nothing, depending on
__CHECK_ENDIAN__ being defined
- gfp_t switched from __nocast to __bitwise__
- force cast to gfp_t added to __GFP_... constants
- new helper - gfp_zone(); extracts zone bits out of gfp_t value and casts
the result to int
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
- added typedef unsigned int __nocast gfp_t;
- replaced __nocast uses for gfp flags with gfp_t - it gives exactly
the same warnings as far as sparse is concerned, doesn't change
generated code (from gcc point of view we replaced unsigned int with
typedef) and documents what's going on far better.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Add another GFP flag: __GFP_HARDWALL.
A subsequent "cpuset_zone_allowed" patch will use this flag to mark GFP_USER
allocations, and distinguish them from GFP_KERNEL allocations.
Allocations (such as GFP_USER) marked GFP_HARDWALL are constrainted to the
current tasks cpuset. Other allocations (such as GFP_KERNEL) can steal from
the possibly larger nearest mem_exclusive cpuset ancestor, if memory is tight
on every node in the current cpuset.
This patch collides with Mel Gorman's patch to reduce fragmentation in the
standard buddy allocator, which adds two GFP flags. This was discussed on
linux-mm in July. Most likely, one of his flags for user reclaimable memory
can be the same as my __GFP_HARDWALL flag, under some generic name meaning its
user address space memory.
Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
The pageset array can potentially acquire a huge amount of memory on large
NUMA systems. F.e. on a system with 512 processors and 256 nodes there
will be 256*512 pagesets. If each pageset only holds 5 pages then we are
talking about 655360 pages.With a 16K page size on IA64 this results in
potentially 10 Gigabytes of memory being trapped in pagesets. The typical
cases are much less for smaller systems but there is still the potential of
memory being trapped in off node pagesets. Off node memory may be rarely
used if local memory is available and so we may potentially have memory in
seldom used pagesets without this patch.
The slab allocator flushes its per cpu caches every 2 seconds. The
following patch flushes the off node pageset caches in the same way by
tying into the slab flush.
The patch also changes /proc/zoneinfo to include the number of pages
currently in each pageset.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
When using the early zone reclaim, it was noticed that allocating new pages
that should be spread across the whole system caused eviction of local pages.
This adds a new GFP flag to prevent early reclaim from happening during
certain allocation attempts. The example that is implemented here is for page
cache pages. We want page cache pages to be spread across the whole system,
and we don't want page cache pages to evict other pages to get local memory.
Signed-off-by: Martin Hicks <mort@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Mempools have 2 problems.
The first is that mempool_alloc can possibly get stuck in __alloc_pages
when they should opt to fail, and take an element from their reserved pool.
The second is that it will happily eat emergency PF_MEMALLOC reserves
instead of going to their reserved pools.
Fix the first by passing __GFP_NORETRY in the allocation calls in
mempool_alloc. Fix the second by introducing a __GFP_MEMPOOL flag which
directs the page allocator not to allocate from the reserve pool.
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Initial git repository build. I'm not bothering with the full history,
even though we have it. We can create a separate "historical" git
archive of that later if we want to, and in the meantime it's about
3.2GB when imported into git - space that would just make the early
git days unnecessarily complicated, when we don't have a lot of good
infrastructure for it.
Let it rip!
Andrew Morton