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>
Fixes a build failure reported by Alan Cox:
mm/hugetlb.c: In function `hugetlb_acct_memory': mm/hugetlb.c:1507:
error: implicit declaration of function `cpuset_mems_nr'
Also reverts Ingo's
commit e44d1b2998
Author: Ingo Molnar <mingo@elte.hu>
Date: Fri Jul 25 12:57:41 2008 +0200
mm/hugetlb.c: fix build failure with !CONFIG_SYSCTL
which fixed the build error but added some unused-static-function warnings.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
on !CONFIG_SYSCTL on x86 with latest -git i get:
mm/hugetlb.c: In function 'decrement_hugepage_resv_vma':
mm/hugetlb.c:83: error: 'reserve' undeclared (first use in this function)
mm/hugetlb.c:83: error: (Each undeclared identifier is reported only once
mm/hugetlb.c:83: error: for each function it appears in.)
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
With shared reservations (and now also with private reservations), we reserve
huge pages at mmap time. We also account for the mapping against fs quota to
prevent a reservation from being preempted by quota exhaustion.
When testing with the libhugetlbfs test suite, I found a problem with quota
accounting. FS quota for allocated pages is handled correctly but we are not
releasing quota for private pages that were reserved but never allocated. Do
this in hugetlb_vm_op_close() at the same time as unused page reservations are
released.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@saeurebad.de>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Hugh Dickins <hugh@veritas.com>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When removing a huge page from the hugepage pool for a fault the system checks
to see if the mapping requires additional pages to be reserved, and if it does
whether there are any unreserved pages remaining. If not, the allocation
fails without even attempting to get a page. In order to determine whether to
apply this check we call vma_has_private_reserves() which tells us if this vma
is MAP_PRIVATE and is the owner. This incorrectly triggers the remaining
reservation test for MAP_SHARED mappings which prevents allocation of the
final page in the pool even though it is reserved for this mapping.
In reality we only want to check this for MAP_PRIVATE mappings where the
process is not the original mapper. Replace vma_has_private_reserves() with
vma_has_reserves() which indicates whether further reserves are required, and
update the caller.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Adam Litke <agl@us.ibm.com>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Allow alloc_bootmem_huge_page() to be overridden by architectures that
can't always use bootmem. This requires huge_boot_pages to be available
for use by this function.
This is required for powerpc 16G pages, which have to be reserved prior to
boot-time. The location of these pages are indicated in the device tree.
Acked-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Jon Tollefson <kniht@linux.vnet.ibm.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Allow configurations with the default huge page size which is different to
the traditional HPAGE_SIZE size. The default huge page size is the one
represented in the legacy /proc ABIs, SHM, and which is defaulted to when
mounting hugetlbfs filesystems.
This is implemented with a new kernel option default_hugepagesz=, which
defaults to HPAGE_SIZE if not specified.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Straight forward extensions for huge pages located in the PUD instead of
PMDs.
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
- Reword sentence to clarify meaning with multiple options
- Add support for using GB prefixes for the page size
- Add extra printk to delayed > MAX_ORDER allocation code
Acked-by: Adam Litke <agl@us.ibm.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Make some infrastructure changes to allow boot-time allocation of
different hugepage page sizes.
- move all basic hstate initialisation into hugetlb_add_hstate
- create a new function hugetlb_hstate_alloc_pages() to do the
actual initial page allocations. Call this function early in
order to allocate giant pages from bootmem.
- Check for multiple hugepages= parameters
Acked-by: Adam Litke <agl@us.ibm.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Andrew Hastings <abh@cray.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is needed on x86-64 to handle GB pages in hugetlbfs, because it is
not practical to enlarge MAX_ORDER to 1GB.
Instead the 1GB pages are only allocated at boot using the bootmem
allocator using the hugepages=... option.
These 1G bootmem pages are never freed. In theory it would be possible to
implement that with some complications, but since it would be a one-way
street (>= MAX_ORDER pages cannot be allocated later) I decided not to
currently.
The >= MAX_ORDER code is not ifdef'ed per architecture. It is not very
big and the ifdef uglyness seemed not be worth it.
Known problems: /proc/meminfo and "free" do not display the memory
allocated for gb pages in "Total". This is a little confusing for the
user.
Acked-by: Andrew Hastings <abh@cray.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Need this as a separate function for a future patch.
No behaviour change.
Acked-by: Adam Litke <agl@us.ibm.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Provide new hugepages user APIs that are more suited to multiple hstates
in sysfs. There is a new directory, /sys/kernel/hugepages. Underneath
that directory there will be a directory per-supported hugepage size,
e.g.:
/sys/kernel/hugepages/hugepages-64kB
/sys/kernel/hugepages/hugepages-16384kB
/sys/kernel/hugepages/hugepages-16777216kB
corresponding to 64k, 16m and 16g respectively. Within each
hugepages-size directory there are a number of files, corresponding to the
tracked counters in the hstate, e.g.:
/sys/kernel/hugepages/hugepages-64/nr_hugepages
/sys/kernel/hugepages/hugepages-64/nr_overcommit_hugepages
/sys/kernel/hugepages/hugepages-64/free_hugepages
/sys/kernel/hugepages/hugepages-64/resv_hugepages
/sys/kernel/hugepages/hugepages-64/surplus_hugepages
Of these files, the first two are read-write and the latter three are
read-only. The size of the hugepage being manipulated is trivially
deducible from the enclosing directory and is always expressed in kB (to
match meminfo).
[dave@linux.vnet.ibm.com: fix build]
[nacc@us.ibm.com: hugetlb: hang off of /sys/kernel/mm rather than /sys/kernel]
[nacc@us.ibm.com: hugetlb: remove CONFIG_SYSFS dependency]
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add the ability to configure the hugetlb hstate used on a per mount basis.
- Add a new pagesize= option to the hugetlbfs mount that allows setting
the page size
- This option causes the mount code to find the hstate corresponding to the
specified size, and sets up a pointer to the hstate in the mount's
superblock.
- Change the hstate accessors to use this information rather than the
global_hstate they were using (requires a slight change in mm/memory.c
so we don't NULL deref in the error-unmap path -- see comments).
[np: take hstate out of hugetlbfs inode and vma->vm_private_data]
Acked-by: Adam Litke <agl@us.ibm.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Add basic support for more than one hstate in hugetlbfs. This is the key
to supporting multiple hugetlbfs page sizes at once.
- Rather than a single hstate, we now have an array, with an iterator
- default_hstate continues to be the struct hstate which we use by default
- Add functions for architectures to register new hstates
[akpm@linux-foundation.org: coding-style fixes]
Acked-by: Adam Litke <agl@us.ibm.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The goal of this patchset is to support multiple hugetlb page sizes. This
is achieved by introducing a new struct hstate structure, which
encapsulates the important hugetlb state and constants (eg. huge page
size, number of huge pages currently allocated, etc).
The hstate structure is then passed around the code which requires these
fields, they will do the right thing regardless of the exact hstate they
are operating on.
This patch adds the hstate structure, with a single global instance of it
(default_hstate), and does the basic work of converting hugetlb to use the
hstate.
Future patches will add more hstate structures to allow for different
hugetlbfs mounts to have different page sizes.
[akpm@linux-foundation.org: coding-style fixes]
Acked-by: Adam Litke <agl@us.ibm.com>
Acked-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Andi Kleen <ak@suse.de>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When a hugetlb mapping with a reservation is split, a new VMA is cloned
from the original. This new VMA is a direct copy of the original
including the reservation count. When this pair of VMAs are unmapped we
will incorrect double account the unused reservation and the overall
reservation count will be incorrect, in extreme cases it will wrap.
The problem occurs when we split an existing VMA say to unmap a page in
the middle. split_vma() will create a new VMA copying all fields from the
original. As we are storing our reservation count in vm_private_data this
is also copies, endowing the new VMA with a duplicate of the original
VMA's reservation. Neither of the new VMAs can exhaust these reservations
as they are too small, but when we unmap and close these VMAs we will
incorrect credit the remainder twice and resv_huge_pages will become out
of sync. This can lead to allocation failures on mappings with
reservations and even to resv_huge_pages wrapping which prevents all
subsequent hugepage allocations.
The simple fix would be to correctly apportion the remaining reservation
count when the split is made. However the only hook we have vm_ops->open
only has the new VMA we do not know the identity of the preceeding VMA.
Also even if we did have that VMA to hand we do not know how much of the
reservation was consumed each side of the split.
This patch therefore takes a different tack. We know that the whole of
any private mapping (which has a reservation) has a reservation over its
whole size. Any present pages represent consumed reservation. Therefore
if we track the instantiated pages we can calculate the remaining
reservation.
This patch reuses the existing regions code to track the regions for which
we have consumed reservation (ie. the instantiated pages), as each page
is faulted in we record the consumption of reservation for the new page.
When we need to return unused reservations at unmap time we simply count
the consumed reservation region subtracting that from the whole of the
map. During a VMA split the newly opened VMA will point to the same
region map, as this map is offset oriented it remains valid for both of
the split VMAs. This map is referenced counted so that it is removed when
all VMAs which are part of the mmap are gone.
Thanks to Adam Litke and Mel Gorman for their review feedback.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Adam Litke <agl@us.ibm.com>
Cc: Johannes Weiner <hannes@saeurebad.de>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Michael Kerrisk <mtk.manpages@googlemail.com>
Cc: Jon Tollefson <kniht@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
By default all shared mappings and most private mappings now have
reservations associated with them. This improves semantics by providing
allocation guarentees to the mapper. However a small number of
applications may attempt to make very large sparse mappings, with these
strict reservations the system will never be able to honour the mapping.
This patch set brings MAP_NORESERVE support to hugetlb files. This allows
new mappings to be made to hugetlbfs files without an associated
reservation, for both shared and private mappings. This allows
applications which want to create very sparse mappings to opt-out of the
reservation system. Obviously as there is no reservation they are liable
to fault at runtime if the huge page pool becomes exhausted; buyer beware.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Adam Litke <agl@us.ibm.com>
Cc: Johannes Weiner <hannes@saeurebad.de>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Michael Kerrisk <mtk.manpages@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The following patch will require use of the reservation regions support.
Move this earlier in the file. No changes have been made to this code.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: Johannes Weiner <hannes@saeurebad.de>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Michael Kerrisk <mtk.manpages@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Create some new accessors for vma private data to cut down on and contain
the casts. Encapsulates the huge and small page offset calculations.
Also adds a couple of VM_BUG_ONs for consistency.
[akpm@linux-foundation.org: Make things static]
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Adam Litke <agl@us.ibm.com>
Cc: Johannes Weiner <hannes@saeurebad.de>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Michael Kerrisk <mtk.manpages@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After patch 2 in this series, a process that successfully calls mmap() for
a MAP_PRIVATE mapping will be guaranteed to successfully fault until a
process calls fork(). At that point, the next write fault from the parent
could fail due to COW if the child still has a reference.
We only reserve pages for the parent but a copy must be made to avoid
leaking data from the parent to the child after fork(). Reserves could be
taken for both parent and child at fork time to guarantee faults but if
the mapping is large it is highly likely we will not have sufficient pages
for the reservation, and it is common to fork only to exec() immediatly
after. A failure here would be very undesirable.
Note that the current behaviour of mainline with MAP_PRIVATE pages is
pretty bad. The following situation is allowed to occur today.
1. Process calls mmap(MAP_PRIVATE)
2. Process calls mlock() to fault all pages and makes sure it succeeds
3. Process forks()
4. Process writes to MAP_PRIVATE mapping while child still exists
5. If the COW fails at this point, the process gets SIGKILLed even though it
had taken care to ensure the pages existed
This patch improves the situation by guaranteeing the reliability of the
process that successfully calls mmap(). When the parent performs COW, it
will try to satisfy the allocation without using reserves. If that fails
the parent will steal the page leaving any children without a page.
Faults from the child after that point will result in failure. If the
child COW happens first, an attempt will be made to allocate the page
without reserves and the child will get SIGKILLed on failure.
To summarise the new behaviour:
1. If the original mapper performs COW on a private mapping with multiple
references, it will attempt to allocate a hugepage from the pool or
the buddy allocator without using the existing reserves. On fail, VMAs
mapping the same area are traversed and the page being COW'd is unmapped
where found. It will then steal the original page as the last mapper in
the normal way.
2. The VMAs the pages were unmapped from are flagged to note that pages
with data no longer exist. Future no-page faults on those VMAs will
terminate the process as otherwise it would appear that data was corrupted.
A warning is printed to the console that this situation occured.
2. If the child performs COW first, it will attempt to satisfy the COW
from the pool if there are enough pages or via the buddy allocator if
overcommit is allowed and the buddy allocator can satisfy the request. If
it fails, the child will be killed.
If the pool is large enough, existing applications will not notice that
the reserves were a factor. Existing applications depending on the
no-reserves been set are unlikely to exist as for much of the history of
hugetlbfs, pages were prefaulted at mmap(), allocating the pages at that
point or failing the mmap().
[npiggin@suse.de: fix CONFIG_HUGETLB=n build]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch reserves huge pages at mmap() time for MAP_PRIVATE mappings in
a similar manner to the reservations taken for MAP_SHARED mappings. The
reserve count is accounted both globally and on a per-VMA basis for
private mappings. This guarantees that a process that successfully calls
mmap() will successfully fault all pages in the future unless fork() is
called.
The characteristics of private mappings of hugetlbfs files behaviour after
this patch are;
1. The process calling mmap() is guaranteed to succeed all future faults until
it forks().
2. On fork(), the parent may die due to SIGKILL on writes to the private
mapping if enough pages are not available for the COW. For reasonably
reliable behaviour in the face of a small huge page pool, children of
hugepage-aware processes should not reference the mappings; such as
might occur when fork()ing to exec().
3. On fork(), the child VMAs inherit no reserves. Reads on pages already
faulted by the parent will succeed. Successful writes will depend on enough
huge pages being free in the pool.
4. Quotas of the hugetlbfs mount are checked at reserve time for the mapper
and at fault time otherwise.
Before this patch, all reads or writes in the child potentially needs page
allocations that can later lead to the death of the parent. This applies
to reads and writes of uninstantiated pages as well as COW. After the
patch it is only a write to an instantiated page that causes problems.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is a patchset to give reliable behaviour to a process that
successfully calls mmap(MAP_PRIVATE) on a hugetlbfs file. Currently, it
is possible for the process to be killed due to a small hugepage pool size
even if it calls mlock().
MAP_SHARED mappings on hugetlbfs reserve huge pages at mmap() time. This
guarantees all future faults against the mapping will succeed. This
allows local allocations at first use improving NUMA locality whilst
retaining reliability.
MAP_PRIVATE mappings do not reserve pages. This can result in an
application being SIGKILLed later if a huge page is not available at fault
time. This makes huge pages usage very ill-advised in some cases as the
unexpected application failure cannot be detected and handled as it is
immediately fatal. Although an application may force instantiation of the
pages using mlock(), this may lead to poor memory placement and the
process may still be killed when performing COW.
This patchset introduces a reliability guarantee for the process which
creates a private mapping, i.e. the process that calls mmap() on a
hugetlbfs file successfully. The first patch of the set is purely
mechanical code move to make later diffs easier to read. The second patch
will guarantee faults up until the process calls fork(). After patch two,
as long as the child keeps the mappings, the parent is no longer
guaranteed to be reliable. Patch 3 guarantees that the parent will always
successfully COW by unmapping the pages from the child in the event there
are insufficient pages in the hugepage pool in allocate a new page, be it
via a static or dynamic pool.
Existing hugepage-aware applications are unlikely to be affected by this
change. For much of hugetlbfs's history, pages were pre-faulted at mmap()
time or mmap() failed which acts in a reserve-like manner. If the pool is
sized correctly already so that parent and child can fault reliably, the
application will not even notice the reserves. It's only when the pool is
too small for the application to function perfectly reliably that the
reserves come into play.
Credit goes to Andy Whitcroft for cleaning up a number of mistakes during
review before the patches were released.
This patch:
A later patch in this set needs to call hugetlb_acct_memory() before it is
defined. This patch moves the function without modification. This makes
later diffs easier to read.
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
It's confusing that set_max_huge_pages() contained two different
variables named "ret", and although the code works correctly this should
be fixed.
The inner of the two variables can simply be removed.
Spotted by sparse.
Signed-off-by: Adrian Bunk <bunk@kernel.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>
Add __GFP_REPEAT to hugepage allocations. Do so to not necessitate userspace
putting pressure on the VM by repeated echo's into /proc/sys/vm/nr_hugepages
to grow the pool. With the previous patch to allow for large-order
__GFP_REPEAT attempts to loop for a bit (as opposed to indefinitely), this
increases the likelihood of getting hugepages when the system experiences (or
recently experienced) load.
Mel tested the patchset on an x86_32 laptop. With the patches, it was easier
to use the proc interface to grow the hugepage pool. The following is the
output of a script that grows the pool as much as possible running on
2.6.25-rc9.
Allocating hugepages test
-------------------------
Disabling OOM Killer for current test process
Starting page count: 0
Attempt 1: 57 pages Progress made with 57 pages
Attempt 2: 73 pages Progress made with 16 pages
Attempt 3: 74 pages Progress made with 1 pages
Attempt 4: 75 pages Progress made with 1 pages
Attempt 5: 77 pages Progress made with 2 pages
77 pages was the most it allocated but it took 5 attempts from userspace
to get it. With the 3 patches in this series applied,
Allocating hugepages test
-------------------------
Disabling OOM Killer for current test process
Starting page count: 0
Attempt 1: 75 pages Progress made with 75 pages
Attempt 2: 76 pages Progress made with 1 pages
Attempt 3: 79 pages Progress made with 3 pages
And 79 pages was the most it got. Your patches were able to allocate the
bulk of possible pages on the first attempt.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Tested-by: Mel Gorman <mel@csn.ul.ie>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Huge ptes have a special type on s390 and cannot be handled with the standard
pte functions in certain cases, e.g. because of a different location of the
invalid bit. This patch adds some new architecture- specific functions to
hugetlb common code, as a prerequisite for the s390 large page support.
This won't affect other architectures in functionality, but I need to add some
new dummy inline functions to the headers.
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A cow break on a hugetlbfs page with page_count > 1 will set a new pte with
set_huge_pte_at(), w/o any tlb flush operation. The old pte will remain in
the tlb and subsequent write access to the page will result in a page fault
loop, for as long as it may take until the tlb is flushed from somewhere else.
This patch introduces an architecture-specific huge_ptep_clear_flush()
function, which is called before the the set_huge_pte_at() in hugetlb_cow().
ATTENTION: This is just a nop on all architectures for now, the s390
implementation will come with our large page patch later. Other architectures
should define their own huge_ptep_clear_flush() if needed.
Acked-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After further discussion with Christoph Lameter, it has become clear that my
earlier attempts to clean up the mempolicy reference counting were a bit of
overkill in some areas, resulting in superflous ref/unref in what are usually
fast paths. In other areas, further inspection reveals that I botched the
unref for interleave policies.
A separate patch, suitable for upstream/stable trees, fixes up the known
errors in the previous attempt to fix reference counting.
This patch reworks the memory policy referencing counting and, one hopes,
simplifies the code. Maybe I'll get it right this time.
See the update to the numa_memory_policy.txt document for a discussion of
memory policy reference counting that motivates this patch.
Summary:
Lookup of mempolicy, based on (vma, address) need only add a reference for
shared policy, and we need only unref the policy when finished for shared
policies. So, this patch backs out all of the unneeded extra reference
counting added by my previous attempt. It then unrefs only shared policies
when we're finished with them, using the mpol_cond_put() [conditional put]
helper function introduced by this patch.
Note that shmem_swapin() calls read_swap_cache_async() with a dummy vma
containing just the policy. read_swap_cache_async() can call alloc_page_vma()
multiple times, so we can't let alloc_page_vma() unref the shared policy in
this case. To avoid this, we make a copy of any non-null shared policy and
remove the MPOL_F_SHARED flag from the copy. This copy occurs before reading
a page [or multiple pages] from swap, so the overhead should not be an issue
here.
I introduced a new static inline function "mpol_cond_copy()" to copy the
shared policy to an on-stack policy and remove the flags that would require a
conditional free. The current implementation of mpol_cond_copy() assumes that
the struct mempolicy contains no pointers to dynamically allocated structures
that must be duplicated or reference counted during copy.
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is a change that was requested some time ago by Mel Gorman. Makes sense
to me, so here it is.
Note: I retain the name "mpol_free_shared_policy()" because it actually does
free the shared_policy, which is NOT a reference counted object. However, ...
The mempolicy object[s] referenced by the shared_policy are reference counted,
so mpol_put() is used to release the reference held by the shared_policy. The
mempolicy might not be freed at this time, because some task attached to the
shared object associated with the shared policy may be in the process of
allocating a page based on the mempolicy. In that case, the task performing
the allocation will hold a reference on the mempolicy, obtained via
mpol_shared_policy_lookup(). The mempolicy will be freed when all tasks
holding such a reference have called mpol_put() for the mempolicy.
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Allocating huge pages directly from the buddy allocator is not guaranteed to
succeed. Success depends on several factors (such as the amount of physical
memory available and the level of fragmentation). With the addition of
dynamic hugetlb pool resizing, allocations can occur much more frequently.
For these reasons it is desirable to keep track of huge page allocation
successes and failures.
Add two new vmstat entries to track huge page allocations that succeed and
fail. The presence of the two entries is contingent upon CONFIG_HUGETLB_PAGE
being enabled.
[akpm@linux-foundation.org: reduced ifdeffery]
Signed-off-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Eric Munson <ebmunson@us.ibm.com>
Tested-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: Andy Whitcroft <apw@shadowen.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>
To reduce hugetlb_lock acquisitions and releases when freeing excess surplus
pages, scan the page list in two parts. First, transfer the needed pages to
the hugetlb pool. Then drop the lock and free the remaining pages back to the
buddy allocator.
In the common case there are zero excess pages and no lock operations are
required.
Thanks Mel Gorman for this improvement.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: David Gibson <david@gibson.dropbear.id.au>
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>
Filtering zonelists requires very frequent use of zone_idx(). This is costly
as it involves a lookup of another structure and a substraction operation. As
the zone_idx is often required, it should be quickly accessible. The node idx
could also be stored here if it was found that accessing zone->node is
significant which may be the case on workloads where nodemasks are heavily
used.
This patch introduces a struct zoneref to store a zone pointer and a zone
index. The zonelist then consists of an array of these struct zonerefs which
are looked up as necessary. Helpers are given for accessing the zone index as
well as the node index.
[kamezawa.hiroyu@jp.fujitsu.com: Suggested struct zoneref instead of embedding information in pointers]
[hugh@veritas.com: mm-have-zonelist: fix memcg ooms]
[hugh@veritas.com: just return do_try_to_free_pages]
[hugh@veritas.com: do_try_to_free_pages gfp_mask redundant]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Christoph Lameter <clameter@sgi.com>
Acked-by: David Rientjes <rientjes@google.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: Nick Piggin <nickpiggin@yahoo.com.au>
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>
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>
Running the counters testcase from libhugetlbfs results in on 2.6.25-rc5
and 2.6.25-rc5-mm1:
BUG: soft lockup - CPU#3 stuck for 61s! [counters:10531]
NIP: c0000000000d1f3c LR: c0000000000d1f2c CTR: c0000000001b5088
REGS: c000005db12cb360 TRAP: 0901 Not tainted (2.6.25-rc5-autokern1)
MSR: 8000000000009032 <EE,ME,IR,DR> CR: 48008448 XER: 20000000
TASK = c000005dbf3d6000[10531] 'counters' THREAD: c000005db12c8000 CPU: 3
GPR00: 0000000000000004 c000005db12cb5e0 c000000000879228 0000000000000004
GPR04: 0000000000000010 0000000000000000 0000000000200200 0000000000100100
GPR08: c0000000008aba10 000000000000ffff 0000000000000004 0000000000000000
GPR12: 0000000028000442 c000000000770080
NIP [c0000000000d1f3c] .return_unused_surplus_pages+0x84/0x18c
LR [c0000000000d1f2c] .return_unused_surplus_pages+0x74/0x18c
Call Trace:
[c000005db12cb5e0] [c000005db12cb670] 0xc000005db12cb670 (unreliable)
[c000005db12cb670] [c0000000000d24c4] .hugetlb_acct_memory+0x2e0/0x354
[c000005db12cb740] [c0000000001b5048] .truncate_hugepages+0x1d4/0x214
[c000005db12cb890] [c0000000001b50a4] .hugetlbfs_delete_inode+0x1c/0x3c
[c000005db12cb920] [c000000000103fd8] .generic_delete_inode+0xf8/0x1c0
[c000005db12cb9b0] [c0000000001b5100] .hugetlbfs_drop_inode+0x3c/0x24c
[c000005db12cba50] [c00000000010287c] .iput+0xdc/0xf8
[c000005db12cbad0] [c0000000000fee54] .dentry_iput+0x12c/0x194
[c000005db12cbb60] [c0000000000ff050] .d_kill+0x6c/0xa4
[c000005db12cbbf0] [c0000000000ffb74] .dput+0x18c/0x1b0
[c000005db12cbc70] [c0000000000e9e98] .__fput+0x1a4/0x1e8
[c000005db12cbd10] [c0000000000e61ec] .filp_close+0xb8/0xe0
[c000005db12cbda0] [c0000000000e62d0] .sys_close+0xbc/0x134
[c000005db12cbe30] [c00000000000872c] syscall_exit+0x0/0x40
Instruction dump:
ebbe8038 38800010 e8bf0002 3bbd0008 7fa3eb78 38a50001 7ca507b4 4818df25
60000000 38800010 38a00000 7c601b78 <7fa3eb78> 2f800010 409d0008 38000010
This was tracked down to a potential livelock in
return_unused_surplus_hugepages(). In the case where we have surplus
pages on some node, but no free pages on the same node, we may never
break out of the loop. To avoid this livelock, terminate the search if
we iterate a number of times equal to the number of online nodes without
freeing a page.
Thanks to Andy Whitcroft and Adam Litke for helping with debugging and
the patch.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Currently we show the surplus hugetlb pool state in /proc/meminfo, but
not in the per-node meminfo files, even though we track the information
on a per-node basis. Printing it there can help track down dynamic pool
bugs including the one in the follow-on patch.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Free pages in the hugetlb pool are free and as such have a reference count of
zero. Regular allocations into the pool from the buddy are "freed" into the
pool which results in their page_count dropping to zero. However, surplus
pages can be directly utilized by the caller without first being freed to the
pool. Therefore, a call to put_page_testzero() is in order so that such a
page will be handed to the caller with a correct count.
This has not affected end users because the bad page count is reset before the
page is handed off. However, under CONFIG_DEBUG_VM this triggers a BUG when
the page count is validated.
Thanks go to Mel for first spotting this issue and providing an initial fix.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Adam Litke noticed that currently we grow the hugepage pool independent of any
cpuset the running process may be in, but when shrinking the pool, the cpuset
is checked. This leads to inconsistency when shrinking the pool in a
restricted cpuset -- an administrator may have been able to grow the pool on a
node restricted by a containing cpuset, but they cannot shrink it there.
There are two options: either prevent growing of the pool outside of the
cpuset or allow shrinking outside of the cpuset. >From previous discussions
on linux-mm, /proc/sys/vm/nr_hugepages is an administrative interface that
should not be restricted by cpusets. So allow shrinking the pool by removing
pages from nodes outside of current's cpuset.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: William Irwin <wli@holomorphy.com>
Cc: Lee Schermerhorn <Lee.Schermerhonr@hp.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
A hugetlb reservation may be inadequately backed in the event of racing
allocations and frees when utilizing surplus huge pages. Consider the
following series of events in processes A and B:
A) Allocates some surplus pages to satisfy a reservation
B) Frees some huge pages
A) A notices the extra free pages and drops hugetlb_lock to free some of
its surplus pages back to the buddy allocator.
B) Allocates some huge pages
A) Reacquires hugetlb_lock and returns from gather_surplus_huge_pages()
Avoid this by commiting the reservation after pages have been allocated but
before dropping the lock to free excess pages. For parity, release the
reservation in return_unused_surplus_pages().
This patch also corrects the cpuset_mems_nr() error path in
hugetlb_acct_memory(). If the cpuset check fails, uncommit the
reservation, but also be sure to return any surplus huge pages that may
have been allocated to back the failed reservation.
Thanks to Andy Whitcroft for discovering this.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When we free a page via free_huge_page and we detect that we are in surplus
the page will be returned to the buddy. After this we no longer own the page.
However at the end free_huge_page we clear out our mapping pointer from
page private. Even where the page is not a surplus we free the page to
the hugepage pool, drop the pool locks and then clear page private. In
either case the page may have been reallocated. BAD.
Make sure we clear out page private before we free the page.
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
proc_doulongvec_minmax() calls copy_to_user()/copy_from_user(), so we can't
hold hugetlb_lock over the call. Use a dummy variable to store the sysctl
result, like in hugetlb_sysctl_handler(), then grab the lock to update
nr_overcommit_huge_pages.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Reported-by: Miles Lane <miles.lane@gmail.com>
Cc: Adam Litke <agl@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When I replaced hugetlb_dynamic_pool with nr_overcommit_hugepages I used
proc_doulongvec_minmax() directly. However, hugetlb.c's locking rules
require that all counter modifications occur under the hugetlb_lock. Add a
callback into the hugetlb code similar to the one for nr_hugepages. Grab
the lock around the manipulation of nr_overcommit_hugepages in
proc_doulongvec_minmax().
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
After running SetPageUptodate, preceeding stores to the page contents to
actually bring it uptodate may not be ordered with the store to set the
page uptodate.
Therefore, another CPU which checks PageUptodate is true, then reads the
page contents can get stale data.
Fix this by having an smp_wmb before SetPageUptodate, and smp_rmb after
PageUptodate.
Many places that test PageUptodate, do so with the page locked, and this
would be enough to ensure memory ordering in those places if
SetPageUptodate were only called while the page is locked. Unfortunately
that is not always the case for some filesystems, but it could be an idea
for the future.
Also bring the handling of anonymous page uptodateness in line with that of
file backed page management, by marking anon pages as uptodate when they
_are_ uptodate, rather than when our implementation requires that they be
marked as such. Doing allows us to get rid of the smp_wmb's in the page
copying functions, which were especially added for anonymous pages for an
analogous memory ordering problem. Both file and anonymous pages are
handled with the same barriers.
FAQ:
Q. Why not do this in flush_dcache_page?
A. Firstly, flush_dcache_page handles only one side (the smb side) of the
ordering protocol; we'd still need smp_rmb somewhere. Secondly, hiding away
memory barriers in a completely unrelated function is nasty; at least in the
PageUptodate macros, they are located together with (half) the operations
involved in the ordering. Thirdly, the smp_wmb is only required when first
bringing the page uptodate, wheras flush_dcache_page should be called each time
it is written to through the kernel mapping. It is logically the wrong place to
put it.
Q. Why does this increase my text size / reduce my performance / etc.
A. Because it is adding the necessary instructions to eliminate the data-race.
Q. Can it be improved?
A. Yes, eg. if you were to create a rule that all SetPageUptodate operations
run under the page lock, we could avoid the smp_rmb places where PageUptodate
is queried under the page lock. Requires audit of all filesystems and at least
some would need reworking. That's great you're interested, I'm eagerly awaiting
your patches.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The shared page table code for hugetlb memory on x86 and x86_64
is causing a leak. When a user of hugepages exits using this code
the system leaks some of the hugepages.
-------------------------------------------------------
Part of /proc/meminfo just before database startup:
HugePages_Total: 5500
HugePages_Free: 5500
HugePages_Rsvd: 0
Hugepagesize: 2048 kB
Just before shutdown:
HugePages_Total: 5500
HugePages_Free: 4475
HugePages_Rsvd: 0
Hugepagesize: 2048 kB
After shutdown:
HugePages_Total: 5500
HugePages_Free: 4988
HugePages_Rsvd:
0 Hugepagesize: 2048 kB
----------------------------------------------------------
The problem occurs durring a fork, in copy_hugetlb_page_range(). It
locates the dst_pte using huge_pte_alloc(). Since huge_pte_alloc() calls
huge_pmd_share() it will share the pmd page if can, yet the main loop in
copy_hugetlb_page_range() does a get_page() on every hugepage. This is a
violation of the shared hugepmd pagetable protocol and creates additional
referenced to the hugepages causing a leak when the unmap of the VMA
occurs. We can skip the entire replication of the ptes when the hugepage
pagetables are shared. The attached patch skips copying the ptes and the
get_page() calls if the hugetlbpage pagetable is shared.
[akpm@linux-foundation.org: coding-style cleanups]
Signed-off-by: Larry Woodman <lwoodman@redhat.com>
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
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>
In the error path of both shared and private hugetlb page allocation,
the file system quota is never undone, leading to fs quota leak. Fix
them up.
[akpm@linux-foundation.org: cleanup, micro-optimise]
Signed-off-by: Ken Chen <kenchen@google.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
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>
This reverts commit 54f9f80d65 ("hugetlb:
Add hugetlb_dynamic_pool sysctl")
Given the new sysctl nr_overcommit_hugepages, the boolean dynamic pool
sysctl is not needed, as its semantics can be expressed by 0 in the
overcommit sysctl (no dynamic pool) and non-0 in the overcommit sysctl
(pool enabled).
(Needed in 2.6.24 since it reverts a post-2.6.23 userspace-visible change)
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
hugetlb: introduce nr_overcommit_hugepages sysctl
While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I
became convinced that having a boolean sysctl was insufficient:
1) To support per-node control of hugepages, I have previously submitted
patches to add a sysfs attribute related to nr_hugepages. However, with
a boolean global value and per-mount quota enforcement constraining the
dynamic pool, adding corresponding control of the dynamic pool on a
per-node basis seems inconsistent to me.
2) Administration of the hugetlb dynamic pool with multiple hugetlbfs
mount points is, arguably, more arduous than it needs to be. Each quota
would need to be set separately, and the sum would need to be monitored.
To ease the administration, and to help make the way for per-node
control of the static & dynamic hugepage pool, I added a separate
sysctl, nr_overcommit_hugepages. This value serves as a high watermark
for the overall hugepage pool, while nr_hugepages serves as a low
watermark. The boolean sysctl can then be removed, as the condition
nr_overcommit_hugepages > 0
indicates the same administrative setting as
hugetlb_dynamic_pool == 1
Quotas still serve as local enforcement of the size of the pool on a
per-mount basis.
A few caveats:
1) There is a race whereby the global surplus huge page counter is
incremented before a hugepage has allocated. Another process could then
try grow the pool, and fail to convert a surplus huge page to a normal
huge page and instead allocate a fresh huge page. I believe this is
benign, as no memory is leaked (the actual pages are still tracked
correctly) and the counters won't go out of sync.
2) Shrinking the static pool while a surplus is in effect will allow the
number of surplus huge pages to exceed the overcommit value. As long as
this condition holds, however, no more surplus huge pages will be
allowed on the system until one of the two sysctls are increased
sufficiently, or the surplus huge pages go out of use and are freed.
Successfully tested on x86_64 with the current libhugetlbfs snapshot,
modified to use the new sysctl.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The follow_hugetlb_page() fix I posted (merged as git commit
5b23dbe817) missed one case. If the pte is
present, but not writable and write access is requested by the caller to
get_user_pages(), the code will do the wrong thing. Rather than calling
hugetlb_fault to make the pte writable, it notes the presence of the pte
and continues.
This simple one-liner makes sure we also fault on the pte for this case.
Please apply.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Acked-by: Dave Kleikamp <shaggy@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
For administrative purpose, we want to query actual block usage for
hugetlbfs file via fstat. Currently, hugetlbfs always return 0. Fix that
up since kernel already has all the information to track it properly.
Signed-off-by: Ken Chen <kenchen@google.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
return_unused_surplus_pages() can become static.
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When a MAP_SHARED mmap of a hugetlbfs file succeeds, huge pages are reserved
to guarantee no problems will occur later when instantiating pages. If quotas
are in force, page instantiation could fail due to a race with another process
or an oversized (but approved) shared mapping.
To prevent these scenarios, debit the quota for the full reservation amount up
front and credit the unused quota when the reservation is released.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
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>
Add a second parameter 'delta' to hugetlb_get_quota and hugetlb_put_quota to
allow bulk updating of the sbinfo->free_blocks counter. This will be used by
the next patch in the series.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
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>
Now that quota is credited by free_huge_page(), calls to hugetlb_get_quota()
seem out of place. The alloc/free API is unbalanced because we handle the
hugetlb_put_quota() but expect the caller to open-code hugetlb_get_quota().
Move the get inside alloc_huge_page to clean up this disparity.
This patch has been kept apart from the previous patch because of the somewhat
dodgy ERR_PTR() use herein. Moving the quota logic means that
alloc_huge_page() has two failure modes. Quota failure must result in a
SIGBUS while a standard allocation failure is OOM. Unfortunately, ERR_PTR()
doesn't like the small positive errnos we have in VM_FAULT_* so they must be
negated before they are used.
Does anyone take issue with the way I am using PTR_ERR. If so, what are your
thoughts on how to clean this up (without needing an if,else if,else block at
each alloc_huge_page() callsite)?
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
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>
The hugetlbfs quota management system was never taught to handle MAP_PRIVATE
mappings when that support was added. Currently, quota is debited at page
instantiation and credited at file truncation. This approach works correctly
for shared pages but is incomplete for private pages. In addition to
hugetlb_no_page(), private pages can be instantiated by hugetlb_cow(); but
this function does not respect quotas.
Private huge pages are treated very much like normal, anonymous pages. They
are not "backed" by the hugetlbfs file and are not stored in the mapping's
radix tree. This means that private pages are invisible to
truncate_hugepages() so that function will not credit the quota.
This patch (based on a prototype provided by Ken Chen) moves quota crediting
for all pages into free_huge_page(). page->private is used to store a pointer
to the mapping to which this page belongs. This is used to credit quota on
the appropriate hugetlbfs instance.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
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>
Hugetlbfs implements a quota system which can limit the amount of memory that
can be used by the filesystem. Before allocating a new huge page for a file,
the quota is checked and debited. The quota is then credited when truncating
the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED
mappings. Before detailing the problems and my proposed solutions, we should
agree on a definition of quotas that properly addresses both private and
shared pages. Since the purpose of quotas is to limit total memory
consumption on a per-filesystem basis, I argue that all pages allocated by the
fs (private and shared) should be charged against quota.
Private Mappings
================
The current code will debit quota for private pages sometimes, but will never
credit it. At a minimum, this causes a leak in the quota accounting which
renders the accounting essentially useless as it is. Shared pages have a one
to one mapping with a hugetlbfs file and are easy to account by debiting on
allocation and crediting on truncate. Private pages are anonymous in nature
and have a many to one relationship with their hugetlbfs files (due to copy on
write). Because private pages are not indexed by the mapping's radix tree,
thier quota cannot be credited at file truncation time. Crediting must be
done when the page is unmapped and freed.
Shared Pages
============
I discovered an issue concerning the interaction between the MAP_SHARED
reservation system and quotas. Since quota is not checked until page
instantiation, an over-quota mmap/reservation will initially succeed. When
instantiating the first over-quota page, the program will receive SIGBUS.
This is inconsistent since the reservation is supposed to be a guarantee. The
solution is to debit the full amount of quota at reservation time and credit
the unused portion when the reservation is released.
This patch series brings quotas back in line by making the following
modifications:
* Private pages
- Debit quota in alloc_huge_page()
- Credit quota in free_huge_page()
* Shared pages
- Debit quota for entire reservation at mmap time
- Credit quota for instantiated pages in free_huge_page()
- Credit quota for unused reservation at munmap time
This patch:
The shared page reservation and dynamic pool resizing features have made the
allocation of private vs. shared huge pages quite different. By splitting
out the private/shared-specific portions of the process into their own
functions, readability is greatly improved. alloc_huge_page now calls the
proper helper and performs common operations.
[akpm@linux-foundation.org: coding-style cleanups]
Signed-off-by: Adam Litke <agl@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: Andy Whitcroft <apw@shadowen.org>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
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>
When calling get_user_pages(), a write flag is passed in by the caller to
indicate if write access is required on the faulted-in pages. Currently,
follow_hugetlb_page() ignores this flag and always faults pages for
read-only access. This can cause data corruption because a device driver
that calls get_user_pages() with write set will not expect COW faults to
occur on the returned pages.
This patch passes the write flag down to follow_hugetlb_page() and makes
sure hugetlb_fault() is called with the right write_access parameter.
[ezk@cs.sunysb.edu: build fix]
Signed-off-by: Adam Litke <agl@us.ibm.com>
Reviewed-by: Ken Chen <kenchen@google.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Erez Zadok <ezk@cs.sunysb.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Get rid of sparse related warnings from places that use integer as NULL
pointer.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org>
Cc: Andi Kleen <ak@suse.de>
Cc: Jeff Garzik <jeff@garzik.org>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Ian Kent <raven@themaw.net>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Davide Libenzi <davidel@xmailserver.org>
Cc: Stephen Smalley <sds@tycho.nsa.gov>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When gather_surplus_pages() fails to allocate enough huge pages to satisfy
the requested reservation, it frees what it did allocate back to the buddy
allocator. put_page() should be called instead of update_and_free_page()
to ensure that pool counters are updated as appropriate and the page's
refcount is decremented.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Acked-by: Dave Hansen <haveblue@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Ken Chen <kenchen@google.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>
Anton found a problem with the hugetlb pool allocation when some nodes have
no memory (http://marc.info/?l=linux-mm&m=118133042025995&w=2). Lee worked
on versions that tried to fix it, but none were accepted. Christoph has
created a set of patches which allow for GFP_THISNODE allocations to fail
if the node has no memory.
Currently, alloc_fresh_huge_page() returns NULL when it is not able to
allocate a huge page on the current node, as specified by its custom
interleave variable. The callers of this function, though, assume that a
failure in alloc_fresh_huge_page() indicates no hugepages can be allocated
on the system period. This might not be the case, for instance, if we have
an uneven NUMA system, and we happen to try to allocate a hugepage on a
node with less memory and fail, while there is still plenty of free memory
on the other nodes.
To correct this, make alloc_fresh_huge_page() search through all online
nodes before deciding no hugepages can be allocated. Add a helper function
for actually allocating the hugepage. Use a new global nid iterator to
control which nid to allocate on.
Note: we expect particular semantics for __GFP_THISNODE, which are now
enforced even for memoryless nodes. That is, there is should be no
fallback to other nodes. Therefore, we rely on the nid passed into
alloc_pages_node() to be the nid the page comes from. If this is
incorrect, accounting will break.
Tested on x86 !NUMA, x86 NUMA, x86_64 NUMA and ppc64 NUMA (with 2
memoryless nodes).
Before on the ppc64 box:
Trying to clear the hugetlb pool
Done. 0 free
Trying to resize the pool to 100
Node 0 HugePages_Free: 25
Node 1 HugePages_Free: 75
Node 2 HugePages_Free: 0
Node 3 HugePages_Free: 0
Done. Initially 100 free
Trying to resize the pool to 200
Node 0 HugePages_Free: 50
Node 1 HugePages_Free: 150
Node 2 HugePages_Free: 0
Node 3 HugePages_Free: 0
Done. 200 free
After:
Trying to clear the hugetlb pool
Done. 0 free
Trying to resize the pool to 100
Node 0 HugePages_Free: 50
Node 1 HugePages_Free: 50
Node 2 HugePages_Free: 0
Node 3 HugePages_Free: 0
Done. Initially 100 free
Trying to resize the pool to 200
Node 0 HugePages_Free: 100
Node 1 HugePages_Free: 100
Node 2 HugePages_Free: 0
Node 3 HugePages_Free: 0
Done. 200 free
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Acked-by: Christoph Lameter <clameter@sgi.com>
Cc: Adam Litke <agl@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Cc: Ken Chen <kenchen@google.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.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>
When shrinking the size of the hugetlb pool via the nr_hugepages sysctl, we
are careful to keep enough pages around to satisfy reservations. But the
calculation is flawed for the following scenario:
Action Pool Counters (Total, Free, Resv)
====== =============
Set pool to 1 page 1 1 0
Map 1 page MAP_PRIVATE 1 1 0
Touch the page to fault it in 1 0 0
Set pool to 3 pages 3 2 0
Map 2 pages MAP_SHARED 3 2 2
Set pool to 2 pages 2 1 2 <-- Mistake, should be 3 2 2
Touch the 2 shared pages 2 0 1 <-- Program crashes here
The last touch above will terminate the process due to lack of huge pages.
This patch corrects the calculation so that it factors in pages being used
for private mappings. Andrew, this is a standalone fix suitable for
mainline. It is also now corrected in my latest dynamic pool resizing
patchset which I will send out soon.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Acked-by: Ken Chen <kenchen@google.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Badari Pulavarty <pbadari@us.ibm.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>
The maximum size of the huge page pool can be controlled using the overall
size of the hugetlb filesystem (via its 'size' mount option). However in the
common case the this will not be set as the pool is traditionally fixed in
size at boot time. In order to maintain the expected semantics, we need to
prevent the pool expanding by default.
This patch introduces a new sysctl controlling dynamic pool resizing. When
this is enabled the pool will expand beyond its base size up to the size of
the hugetlb filesystem. It is disabled by default.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Dave McCracken <dave.mccracken@oracle.com>
Cc: William Irwin <bill.irwin@oracle.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Ken Chen <kenchen@google.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Shared mappings require special handling because the huge pages needed to
fully populate the VMA must be reserved at mmap time. If not enough pages are
available when making the reservation, allocate all of the shortfall at once
from the buddy allocator and add the pages directly to the hugetlb pool. If
they cannot be allocated, then fail the mapping. The page surplus is
accounted for in the same way as for private mappings; faulted surplus pages
will be freed at unmap time. Reserved, surplus pages that have not been used
must be freed separately when their reservation has been released.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Dave McCracken <dave.mccracken@oracle.com>
Cc: William Irwin <bill.irwin@oracle.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Ken Chen <kenchen@google.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Because we overcommit hugepages for MAP_PRIVATE mappings, it is possible that
the hugetlb pool will be exhausted or completely reserved when a hugepage is
needed to satisfy a page fault. Before killing the process in this situation,
try to allocate a hugepage directly from the buddy allocator.
The explicitly configured pool size becomes a low watermark. When dynamically
grown, the allocated huge pages are accounted as a surplus over the watermark.
As huge pages are freed on a node, surplus pages are released to the buddy
allocator so that the pool will shrink back to the watermark.
Surplus accounting also allows for friendlier explicit pool resizing. When
shrinking a pool that is fully in-use, increase the surplus so pages will be
returned to the buddy allocator as soon as they are freed. When growing a
pool that has a surplus, consume the surplus first and then allocate new
pages.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Dave McCracken <dave.mccracken@oracle.com>
Cc: William Irwin <bill.irwin@oracle.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Ken Chen <kenchen@google.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Dynamic huge page pool resizing.
In most real-world scenarios, configuring the size of the hugetlb pool
correctly is a difficult task. If too few pages are allocated to the pool,
applications using MAP_SHARED may fail to mmap() a hugepage region and
applications using MAP_PRIVATE may receive SIGBUS. Isolating too much memory
in the hugetlb pool means it is not available for other uses, especially those
programs not using huge pages.
The obvious answer is to let the hugetlb pool grow and shrink in response to
the runtime demand for huge pages. The work Mel Gorman has been doing to
establish a memory zone for movable memory allocations makes dynamically
resizing the hugetlb pool reliable within the limits of that zone. This patch
series implements dynamic pool resizing for private and shared mappings while
being careful to maintain existing semantics. Please reply with your comments
and feedback; even just to say whether it would be a useful feature to you.
Thanks.
How it works
============
Upon depletion of the hugetlb pool, rather than reporting an error immediately,
first try and allocate the needed huge pages directly from the buddy allocator.
Care must be taken to avoid unbounded growth of the hugetlb pool, so the
hugetlb filesystem quota is used to limit overall pool size.
The real work begins when we decide there is a shortage of huge pages. What
happens next depends on whether the pages are for a private or shared mapping.
Private mappings are straightforward. At fault time, if alloc_huge_page()
fails, we allocate a page from the buddy allocator and increment the source
node's surplus_huge_pages counter. When free_huge_page() is called for a page
on a node with a surplus, the page is freed directly to the buddy allocator
instead of the hugetlb pool.
Because shared mappings require all of the pages to be reserved up front, some
additional work must be done at mmap() to support them. We determine the
reservation shortage and allocate the required number of pages all at once.
These pages are then added to the hugetlb pool and marked reserved. Where that
is not possible the mmap() will fail. As with private mappings, the
appropriate surplus counters are updated. Since reserved huge pages won't
necessarily be used by the process, we can't be sure that free_huge_page() will
always be called to return surplus pages to the buddy allocator. To prevent
the huge page pool from bloating, we must free unused surplus pages when their
reservation has ended.
Controlling it
==============
With the entire patch series applied, pool resizing is off by default so unless
specific action is taken, the semantics are unchanged.
To take advantage of the flexibility afforded by this patch series one must
tolerate a change in semantics. To control hugetlb pool growth, the following
techniques can be employed:
* A sysctl tunable to enable/disable the feature entirely
* The size= mount option for hugetlbfs filesystems to limit pool size
Performance
===========
When contiguous memory is readily available, it is expected that the cost of
dynamicly resizing the pool will be small. This series has been performance
tested with 'stream' to measure this cost.
Stream (http://www.cs.virginia.edu/stream/) was linked with libhugetlbfs to
enable remapping of the text and data/bss segments into huge pages.
Stream with small array
-----------------------
Baseline: nr_hugepages = 0, No libhugetlbfs segment remapping
Preallocated: nr_hugepages = 5, Text and data/bss remapping
Dynamic: nr_hugepages = 0, Text and data/bss remapping
Rate (MB/s)
Function Baseline Preallocated Dynamic
Copy: 4695.6266 5942.8371 5982.2287
Scale: 4451.5776 5017.1419 5658.7843
Add: 5815.8849 7927.7827 8119.3552
Triad: 5949.4144 8527.6492 8110.6903
Stream with large array
-----------------------
Baseline: nr_hugepages = 0, No libhugetlbfs segment remapping
Preallocated: nr_hugepages = 67, Text and data/bss remapping
Dynamic: nr_hugepages = 0, Text and data/bss remapping
Rate (MB/s)
Function Baseline Preallocated Dynamic
Copy: 2227.8281 2544.2732 2546.4947
Scale: 2136.3208 2430.7294 2421.2074
Add: 2773.1449 4004.0021 3999.4331
Triad: 2748.4502 3777.0109 3773.4970
* All numbers are averages taken from 10 consecutive runs with a maximum
standard deviation of 1.3 percent noted.
This patch:
Simply move update_and_free_page() so that it can be reused later in this
patch series. The implementation is not changed.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Acked-by: Andy Whitcroft <apw@shadowen.org>
Acked-by: Dave McCracken <dave.mccracken@oracle.com>
Acked-by: William Irwin <bill.irwin@oracle.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Ken Chen <kenchen@google.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Current ia64 kernel flushes icache by lazy_mmu_prot_update() *after*
set_pte(). This is too late. This patch removes lazy_mmu_prot_update and
add modfied set_pte() for flushing if necessary.
This patch flush icache of a page when
new pte has exec bit.
&& new pte has present bit
&& new pte is user's page.
&& (old *ptep is not present
|| new pte's pfn is not same to old *ptep's ptn)
&& new pte's page has no Pg_arch_1 bit.
Pg_arch_1 is set when a page is cache consistent.
I think this condition checks are much easier to understand than considering
"Where sync_icache_dcache() should be inserted ?".
pte_user() for ia64 was removed by http://lkml.org/lkml/2007/6/12/67 as
clean-up. So, I added it again.
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Acked-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The virtual address space argument of clear_user_highpage is supposed to be
the virtual address where the page being cleared will eventually be mapped.
This allows architectures with virtually indexed caches a few clever
tricks. That sort of trick falls over in painful ways if the virtual
address argument is wrong.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch proposes fixes to the reference counting of memory policy in the
page allocation paths and in show_numa_map(). Extracted from my "Memory
Policy Cleanups and Enhancements" series as stand-alone.
Shared policy lookup [shmem] has always added a reference to the policy,
but this was never unrefed after page allocation or after formatting the
numa map data.
Default system policy should not require additional ref counting, nor
should the current task's task policy. However, show_numa_map() calls
get_vma_policy() to examine what may be [likely is] another task's policy.
The latter case needs protection against freeing of the policy.
This patch adds a reference count to a mempolicy returned by
get_vma_policy() when the policy is a vma policy or another task's
mempolicy. Again, shared policy is already reference counted on lookup. A
matching "unref" [__mpol_free()] is performed in alloc_page_vma() for
shared and vma policies, and in show_numa_map() for shared and another
task's mempolicy. We can call __mpol_free() directly, saving an admittedly
inexpensive inline NULL test, because we know we have a non-NULL policy.
Handling policy ref counts for hugepages is a bit trickier.
huge_zonelist() returns a zone list that might come from a shared or vma
'BIND policy. In this case, we should hold the reference until after the
huge page allocation in dequeue_hugepage(). The patch modifies
huge_zonelist() to return a pointer to the mempolicy if it needs to be
unref'd after allocation.
Kernel Build [16cpu, 32GB, ia64] - average of 10 runs:
w/o patch w/ refcount patch
Avg Std Devn Avg Std Devn
Real: 100.59 0.38 100.63 0.43
User: 1209.60 0.37 1209.91 0.31
System: 81.52 0.42 81.64 0.34
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Acked-by: Andi Kleen <ak@suse.de>
Cc: Christoph Lameter <clameter@sgi.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>
It seems a simple mistake was made when converting follow_hugetlb_page()
over to the VM_FAULT flags bitmasks (in "mm: fault feedback #2", commit
83c54070ee).
By using the wrong bitmask, hugetlb_fault() failures are not being
recognized. This results in an infinite loop whenever follow_hugetlb_page
is involved in a failed fault.
Signed-off-by: Adam Litke <agl@us.ibm.com>
Acked-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
dequeue_huge_page() has a serious memory leak upon hugetlb page
allocation. The for loop continues on allocating hugetlb pages out of
all allowable zone, where this function is supposedly only dequeue one
and only one pages.
Fixed it by breaking out of the for loop once a hugetlb page is found.
Signed-off-by: Ken Chen <kenchen@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use appropriate accessor function to set compound page destructor
function.
Cc: William Irwin <wli@holomorphy.com>
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The fix to that race in alloc_fresh_huge_page() which could give an illegal
node ID did not need nid_lock at all: the fix was to replace static int nid
by static int prev_nid and do the work on local int nid. nid_lock did make
sure that racers strictly roundrobin the nodes, but that's not something we
need to enforce strictly. Kill nid_lock.
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>
mm/hugetlb.c: In function `dequeue_huge_page':
mm/hugetlb.c:72: warning: 'nid' might be used uninitialized in this function
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Adam Litke <agl@us.ibm.com>
Cc: David Gibson <hermes@gibson.dropbear.id.au>
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>
This patch completes Linus's wish that the fault return codes be made into
bit flags, which I agree makes everything nicer. This requires requires
all handle_mm_fault callers to be modified (possibly the modifications
should go further and do things like fault accounting in handle_mm_fault --
however that would be for another patch).
[akpm@linux-foundation.org: fix alpha build]
[akpm@linux-foundation.org: fix s390 build]
[akpm@linux-foundation.org: fix sparc build]
[akpm@linux-foundation.org: fix sparc64 build]
[akpm@linux-foundation.org: fix ia64 build]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Ian Molton <spyro@f2s.com>
Cc: Bryan Wu <bryan.wu@analog.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Yoshinori Sato <ysato@users.sourceforge.jp>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Cc: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Roman Zippel <zippel@linux-m68k.org>
Cc: Greg Ungerer <gerg@uclinux.org>
Cc: Matthew Wilcox <willy@debian.org>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Heiko Carstens <heiko.carstens@de.ibm.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp>
Cc: Richard Curnow <rc@rc0.org.uk>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Dike <jdike@addtoit.com>
Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it>
Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp>
Cc: Chris Zankel <chris@zankel.net>
Acked-by: Kyle McMartin <kyle@mcmartin.ca>
Acked-by: Haavard Skinnemoen <hskinnemoen@atmel.com>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Acked-by: Andi Kleen <ak@muc.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ Still apparently needs some ARM and PPC loving - Linus ]
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Change ->fault prototype. We now return an int, which contains
VM_FAULT_xxx code in the low byte, and FAULT_RET_xxx code in the next byte.
FAULT_RET_ code tells the VM whether a page was found, whether it has been
locked, and potentially other things. This is not quite the way he wanted
it yet, but that's changed in the next patch (which requires changes to
arch code).
This means we no longer set VM_CAN_INVALIDATE in the vma in order to say
that a page is locked which requires filemap_nopage to go away (because we
can no longer remain backward compatible without that flag), but we were
going to do that anyway.
struct fault_data is renamed to struct vm_fault as Linus asked. address
is now a void __user * that we should firmly encourage drivers not to use
without really good reason.
The page is now returned via a page pointer in the vm_fault struct.
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Robert P. J. Day <rpjday@mindspring.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Huge pages are not movable so are not allocated from ZONE_MOVABLE. However,
as ZONE_MOVABLE will always have pages that can be migrated or reclaimed, it
can be used to satisfy hugepage allocations even when the system has been
running a long time. This allows an administrator to resize the hugepage pool
at runtime depending on the size of ZONE_MOVABLE.
This patch adds a new sysctl called hugepages_treat_as_movable. When a
non-zero value is written to it, future allocations for the huge page pool
will use ZONE_MOVABLE. 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.
[akpm@linux-foundation.org: various fixes]
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>
That static `nid' index needs locking. Without it we can end up calling
alloc_pages_node() with an illegal node ID and the kernel crashes.
Acked-by: gurudas pai <gurudas.pai@oracle.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
nid is initialized to numa_node_id() but will either be overwritten in
the loop or not used in the conditional. So remove the initialization.
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Some changes done a while ago to avoid pounding on ptep_set_access_flags and
update_mmu_cache in some race situations break sun4c which requires
update_mmu_cache() to always be called on minor faults.
This patch reworks ptep_set_access_flags() semantics, implementations and
callers so that it's now responsible for returning whether an update is
necessary or not (basically whether the PTE actually changed). This allow
fixing the sparc implementation to always return 1 on sun4c.
[akpm@linux-foundation.org: fixes, cleanups]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: David Miller <davem@davemloft.net>
Cc: Mark Fortescue <mark@mtfhpc.demon.co.uk>
Acked-by: William Lee Irwin III <wli@holomorphy.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When cpuset is configured, it breaks the strict hugetlb page reservation as
the accounting is done on a global variable. Such reservation is
completely rubbish in the presence of cpuset because the reservation is not
checked against page availability for the current cpuset. Application can
still potentially OOM'ed by kernel with lack of free htlb page in cpuset
that the task is in. Attempt to enforce strict accounting with cpuset is
almost impossible (or too ugly) because cpuset is too fluid that task or
memory node can be dynamically moved between cpusets.
The change of semantics for shared hugetlb mapping with cpuset is
undesirable. However, in order to preserve some of the semantics, we fall
back to check against current free page availability as a best attempt and
hopefully to minimize the impact of changing semantics that cpuset has on
hugetlb.
Signed-off-by: Ken Chen <kenchen@google.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: Christoph Lameter <clameter@engr.sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The internal hugetlb resv_huge_pages variable can permanently leak nonzero
value in the error path of hugetlb page fault handler when hugetlb page is
used in combination of cpuset. The leaked count can permanently trap N
number of hugetlb pages in unusable "reserved" state.
Steps to reproduce the bug:
(1) create two cpuset, user1 and user2
(2) reserve 50 htlb pages in cpuset user1
(3) attempt to shmget/shmat 50 htlb page inside cpuset user2
(4) kernel oom the user process in step 3
(5) ipcrm the shm segment
At this point resv_huge_pages will have a count of 49, even though
there are no active hugetlbfs file nor hugetlb shared memory segment
in the system. The leak is permanent and there is no recovery method
other than system reboot. The leaked count will hold up all future use
of that many htlb pages in all cpusets.
The culprit is that the error path of alloc_huge_page() did not
properly undo the change it made to resv_huge_page, causing
inconsistent state.
Signed-off-by: Ken Chen <kenchen@google.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Adam Litke <agl@us.ibm.com>
Cc: Martin Bligh <mbligh@google.com>
Acked-by: David Gibson <dwg@au1.ibm.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
__unmap_hugepage_range() is buggy that it does not preserve dirty state of
huge_pte when unmapping hugepage range. It causes data corruption in the
event of dop_caches being used by sys admin. For example, an application
creates a hugetlb file, modify pages, then unmap it. While leaving the
hugetlb file alive, comes along sys admin doing a "echo 3 >
/proc/sys/vm/drop_caches".
drop_pagecache_sb() will happily free all pages that aren't marked dirty if
there are no active mapping. Later when application remaps the hugetlb
file back and all data are gone, triggering catastrophic flip over on
application.
Not only that, the internal resv_huge_pages count will also get all messed
up. Fix it up by marking page dirty appropriately.
Signed-off-by: Ken Chen <kenchen@google.com>
Cc: "Nish Aravamudan" <nish.aravamudan@gmail.com>
Cc: Adam Litke <agl@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: <stable@kernel.org>
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
To allow a more effective copy_user_highpage() on certain architectures,
a vma argument is added to the function and cow_user_page() allowing
the implementation of these functions to check for the VM_EXEC bit.
The main part of this patch was originally written by Ralf Baechle;
Atushi Nemoto did the the debugging.
Signed-off-by: Atsushi Nemoto <anemo@mba.ocn.ne.jp>
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Elaborate the API for calling cpuset_zone_allowed(), so that users have to
explicitly choose between the two variants:
cpuset_zone_allowed_hardwall()
cpuset_zone_allowed_softwall()
Until now, whether or not you got the hardwall flavor depended solely on
whether or not you or'd in the __GFP_HARDWALL gfp flag to the gfp_mask
argument.
If you didn't specify __GFP_HARDWALL, you implicitly got the softwall
version.
Unfortunately, this meant that users would end up with the softwall version
without thinking about it. Since only the softwall version might sleep,
this led to bugs with possible sleeping in interrupt context on more than
one occassion.
The hardwall version requires that the current tasks mems_allowed allows
the node of the specified zone (or that you're in interrupt or that
__GFP_THISNODE is set or that you're on a one cpuset system.)
The softwall version, depending on the gfp_mask, might allow a node if it
was allowed in the nearest enclusing cpuset marked mem_exclusive (which
requires taking the cpuset lock 'callback_mutex' to evaluate.)
This patch removes the cpuset_zone_allowed() call, and forces the caller to
explicitly choose between the hardwall and the softwall case.
If the caller wants the gfp_mask to determine this choice, they should (1)
be sure they can sleep or that __GFP_HARDWALL is set, and (2) invoke the
cpuset_zone_allowed_softwall() routine.
This adds another 100 or 200 bytes to the kernel text space, due to the few
lines of nearly duplicate code at the top of both cpuset_zone_allowed_*
routines. It should save a few instructions executed for the calls that
turned into calls of cpuset_zone_allowed_hardwall, thanks to not having to
set (before the call) then check (within the call) the __GFP_HARDWALL flag.
For the most critical call, from get_page_from_freelist(), the same
instructions are executed as before -- the old cpuset_zone_allowed()
routine it used to call is the same code as the
cpuset_zone_allowed_softwall() routine that it calls now.
Not a perfect win, but seems worth it, to reduce this chance of hitting a
sleeping with irq off complaint again.
Signed-off-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Currently we we use the lru head link of the second page of a compound page
to hold its destructor. This was ok when it was purely an internal
implmentation detail. However, hugetlbfs overrides this destructor
violating the layering. Abstract this out as explicit calls, also
introduce a type for the callback function allowing them to be type
checked. For each callback we pre-declare the function, causing a type
error on definition rather than on use elsewhere.
[akpm@osdl.org: cleanups]
Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Imprecise RSS accounting is an irritating ill effect with pt sharing. After
consulted with several VM experts, I have tried various methods to solve that
problem: (1) iterate through all mm_structs that share the PT and increment
count; (2) keep RSS count in page table structure and then sum them up at
reporting time. None of the above methods yield any satisfactory
implementation.
Since process RSS accounting is pure information only, I propose we don't
count them at all for hugetlb page. rlimit has such field, though there is
absolutely no enforcement on limiting that resource. One other method is to
account all RSS at hugetlb mmap time regardless they are faulted or not. I
opt for the simplicity of no accounting at all.
Hugetlb page are special, they are reserved up front in global reservation
pool and is not reclaimable. From physical memory resource point of view, it
is already consumed regardless whether there are users using them.
If the concern is that RSS can be used to control resource allocation, we
already can specify hugetlb fs size limit and sysadmin can enforce that at
mount time. Combined with the two points mentioned above, I fail to see if
there is anything got affected because of this patch.
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Cc: Dave McCracken <dmccr@us.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Adam Litke <agl@us.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Following up with the work on shared page table done by Dave McCracken. This
set of patch target shared page table for hugetlb memory only.
The shared page table is particular useful in the situation of large number of
independent processes sharing large shared memory segments. In the normal
page case, the amount of memory saved from process' page table is quite
significant. For hugetlb, the saving on page table memory is not the primary
objective (as hugetlb itself already cuts down page table overhead
significantly), instead, the purpose of using shared page table on hugetlb is
to allow faster TLB refill and smaller cache pollution upon TLB miss.
With PT sharing, pte entries are shared among hundreds of processes, the cache
consumption used by all the page table is smaller and in return, application
gets much higher cache hit ratio. One other effect is that cache hit ratio
with hardware page walker hitting on pte in cache will be higher and this
helps to reduce tlb miss latency. These two effects contribute to higher
application performance.
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Acked-by: Hugh Dickins <hugh@veritas.com>
Cc: Dave McCracken <dmccr@us.ibm.com>
Cc: William Lee Irwin III <wli@holomorphy.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: Adam Litke <agl@us.ibm.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
If you truncated an mmap'ed hugetlbfs file, then faulted on the truncated
area, /proc/meminfo's HugePages_Rsvd wrapped hugely "negative". Reinstate my
preliminary i_size check before attempting to allocate the page (though this
only fixes the most obvious case: more work will be needed here).
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Adam Litke <agl@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: "Chen, Kenneth W" <kenneth.w.chen@intel.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
commit fe1668ae5b causes kernel to oops with
libhugetlbfs test suite. The problem is that hugetlb pages can be shared
by multiple mappings. Multiple threads can fight over page->lru in the
unmap path and bad things happen. We now serialize __unmap_hugepage_range
to void concurrent linked list manipulation. Such serialization is also
needed for shared page table page on hugetlb area. This patch will fixed
the bug and also serve as a prepatch for shared page table.
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Spotted by Hugh that hugetlb page is free'ed back to global pool before
performing any TLB flush in unmap_hugepage_range(). This potentially allow
threads to abuse free-alloc race condition.
The generic tlb gather code is unsuitable to use by hugetlb, I just open
coded a page gathering list and delayed put_page until tlb flush is
performed.
Cc: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Acked-by: William Irwin <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
There are many places where we need to determine the node of a zone.
Currently we use a difficult to read sequence of pointer dereferencing.
Put that into an inline function and use throughout VM. Maybe we can find
a way to optimize the lookup in the future.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
I found two location in hugetlb.c where we chase pointer instead of using
page_to_nid(). Page_to_nid is more effective and can get the node directly
from page flags.
Signed-off-by: Christoph Lameter <clameter@sgi.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Current hugetlb strict accounting for shared mapping always assume mapping
starts at zero file offset and reserves pages between zero and size of the
file. This assumption often reserves (or lock down) a lot more pages then
necessary if application maps at none zero file offset. libhugetlbfs is
one example that requires proper reservation on shared mapping starts at
none zero offset.
This patch extends the reservation and hugetlb strict accounting to support
any arbitrary pair of (offset, len), resulting a much more robust and
accurate scheme. More importantly, it won't lock down any hugetlb pages
outside file mapping.
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Acked-by: Adam Litke <agl@us.ibm.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: William Lee Irwin III <wli@holomorphy.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
With strict page reservation, I think kernel should enforce number of free
hugetlb page don't fall below reserved count. Currently it is possible in
the sysctl path. Add proper check in sysctl to disallow that.
Signed-off-by: Ken Chen <kenneth.w.chen@intel.com>
Cc: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
