md: guard against possible bad array geometry in v1 metadata

Make sure the data doesn't start before the end of the superblock when the
superblock is at the start of the device.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

md: reduce CPU wastage on idle md array with a write-intent bitmap

On an md array with a write-intent bitmap, a thread wakes up every few seconds
and scans the bitmap looking for work to do.  If the array is idle, there will
be no work to do, but a lot of scanning is done to discover this.

So cache the fact that the bitmap is completely clean, and avoid scanning the
whole bitmap when the cache is known to be clean.

Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

md: fix deadlock in md/raid1 and md/raid10 when handling a read error

When handling a read error, we freeze the array to stop any other IO while
attempting to over-write with correct data.

This is done in the raid1d(raid10d) thread and must wait for all submitted IO
to complete (except for requests that failed and are sitting in the retry
queue - these are counted in ->nr_queue and will stay there during a freeze).

However write requests need attention from raid1d as bitmap updates might be
required.  This can cause a deadlock as raid1 is waiting for requests to
finish that themselves need attention from raid1d.

So we create a new function 'flush_pending_writes' to give that attention, and
call it in freeze_array to be sure that we aren't waiting on raid1d.

Thanks to "K.Tanaka" <k-tanaka@ce.jp.nec.com> for finding and reporting this
problem.

Cc: "K.Tanaka" <k-tanaka@ce.jp.nec.com>
Signed-off-by: Neil Brown <neilb@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

iommu: parisc: make the IOMMUs respect the segment boundary limits

Make PARISC's two IOMMU implementations not allocate a memory area spanning
LLD's segment boundary.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Kyle McMartin <kyle@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Cc: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

iommu: parisc: pass struct device to iommu_alloc_range

This adds struct device argument to sba_alloc_range and ccio_alloc_range, a
preparation for modifications to fix the IOMMU segment boundary problem.  This
change enables ccio_alloc_range to access to LLD's segment boundary limits.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Kyle McMartin <kyle@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Cc: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

iommu: export iommu_is_span_boundary helper function

iommu_is_span_boundary is used internally in the IOMMU helper
(lib/iommu-helper.c), a primitive function that judges whether a memory area
spans LLD's segment boundary or not.

It's difficult to convert some IOMMUs to use the IOMMU helper but
iommu_is_span_boundary is still useful for them.  So this patch exports it.

This is needed for the parisc iommu fixes.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Kyle McMartin <kyle@parisc-linux.org>
Cc: Matthew Wilcox <matthew@wil.cx>
Cc: Grant Grundler <grundler@parisc-linux.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hisax_fcpcipnp: move request_irq later in probe

After a quick glance at the code, we're getting the DEBUG_SHIRQ spurious
interrupt before we have the adapter template filled in.  Real interrupts
appear to be turned on by fcpci*_init(), so move request_irq until just before
that.

Signed-off-by: Kyle McMartin <kmcmartin@redhat.com>
Cc: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

eCryptfs: make ecryptfs_prepare_write decrypt the page

When the page is not up to date, ecryptfs_prepare_write() should be
acting much like ecryptfs_readpage(). This includes the painfully
obvious step of actually decrypting the page contents read from the
lower encrypted file.

Note that this patch resolves a bug in eCryptfs in 2.6.24 that one can
produce with these steps:

# mount -t ecryptfs /secret /secret
# echo "abc" > /secret/file.txt
# umount /secret
# mount -t ecryptfs /secret /secret
# echo "def" >> /secret/file.txt
# cat /secret/file.txt

Without this patch, the resulting data returned from cat is likely to
be something other than "abc\ndef\n".

(Thanks to Benedikt Driessen for reporting this.)

Signed-off-by: Michael Halcrow <mhalcrow@us.ibm.com>
Cc: Benedikt Driessen <bdriessen@escrypt.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris: correct syscall numbers in unistd.h for timerfd_settime and timerfd_gettime

Last commit for unistd was not correct, it only had a partial update of
syscall numbers for __NR_timerfd_settime and __NR_timerfd_gettime.  Also,
NR_syscalls was not incremented for the new syscalls.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cris: correct usage of __user for copy to and from user space in lib/usercopy and uaccess.h

Function __copy_user_zeroing in arch/lib/usercopy.c had the wrong parameter
set as __user, and in include/asm-cris/uaccess.h, it was not set at all for
some of the calling functions.

This will cut the number of warnings quite dramatically when using sparse.

While we're here, remove useless CVS log and correct confusing typo.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ACPI: thinkpad-acpi: fix hotkey_get_tablet_mode

I used the wrong return convention on hotkey_get_tablet_mode(), breaking a lot
of stuff.  Bad Henrique!

Fix it to return the status in the parameter-by-reference, and IO status on
the function return value.  Duh.

Signed-off-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br>
Cc: Zdenek Kabelac <zdenek.kabelac@gmail.com>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: Lukas Hejtmanek <xhejtman@ics.muni.cz>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/reiserfs/super.c: correct use of ! and &

In commit e6bafba5b4765a5a252f1b8d31cbf6d2459da337 ("wmi: (!x & y)
strikes again"), a bug was fixed that involved converting !x & y to !(x
& y).  The code below shows the same pattern, and thus should perhaps be
fixed in the same way.

This is not tested and clearly changes the semantics, so it is only
something to consider.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@ expression E1,E2; @@
(
  !E1 & !E2
|
- !E1 & E2
+ !(E1 & E2)
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Cc: Chris Mason <chris.mason@oracle.com>
Cc: Jeff Mahoney <jeffm@suse.com>
Cc: Jan Kara <jack@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/serial/m32r_sio.c: correct use of ! and &

In commit e6bafba5b4765a5a252f1b8d31cbf6d2459da337 ("wmi: (!x & y)
strikes again"), a bug was fixed that involved converting !x & y to !(x
& y).  The code below shows the same pattern, and thus should perhaps be
fixed in the same way.

This is not tested and clearly changes the semantics, so it is only
something to consider.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@ expression E1,E2; @@
(
  !E1 & !E2
|
- !E1 & E2
+ !(E1 & E2)
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Cc: Hirokazu Takata <takata@linux-m32r.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/isdn: correct use of ! and &

In commit e6bafba5b4765a5a252f1b8d31cbf6d2459da337 ("wmi: (!x & y)
strikes again"), a bug was fixed that involved converting !x & y to !(x
& y).  The code below shows the same pattern, and thus should perhaps be
fixed in the same way.

This is not tested and clearly changes the semantics, so it is only
something to consider.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@ expression E1,E2; @@
(
  !E1 & !E2
|
- !E1 & E2
+ !(E1 & E2)
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Cc: Karsten Keil <kkeil@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/char/isicom.c: correct use of ! and &

In commit e6bafba5b4765a5a252f1b8d31cbf6d2459da337 ("wmi: (!x & y)
strikes again"), a bug was fixed that involved converting !x & y to !(x
& y).  The code below shows the same pattern, and thus should perhaps be
fixed in the same way.

This is not tested and clearly changes the semantics, so it is only
something to consider.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@ expression E1,E2; @@
(
  !E1 & !E2
|
- !E1 & E2
+ !(E1 & E2)
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Cc: Jiri Slaby <jirislaby@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: fix oops on NULL lru list

While testing force_empty, during an exit_mmap, __mem_cgroup_remove_list
called from mem_cgroup_uncharge_page oopsed on a NULL pointer in the lru list.
 I couldn't see what racing tasks on other cpus were doing, but surmise that
another must have been in mem_cgroup_charge_common on the same page, between
its unlock_page_cgroup and spin_lock_irqsave near done (thanks to that kzalloc
which I'd almost changed to a kmalloc).

Normally such a race cannot happen, the ref_cnt prevents it, the final
uncharge cannot race with the initial charge.  But force_empty buggers the
ref_cnt, that's what it's all about; and thereafter forced pages are
vulnerable to races such as this (just think of a shared page also mapped into
an mm of another mem_cgroup than that just emptied).  And remain vulnerable
until they're freed indefinitely later.

This patch just fixes the oops by moving the unlock_page_cgroups down below
adding to and removing from the list (only possible given the previous patch);
and while we're at it, we might as well make it an invariant that
page->page_cgroup is always set while pc is on lru.

But this behaviour of force_empty seems highly unsatisfactory to me: why have
a ref_cnt if we always have to cope with it being violated (as in the earlier
page migration patch).  We may prefer force_empty to move pages to an orphan
mem_cgroup (could be the root, but better not), from which other cgroups could
recover them; we might need to reverse the locking again; but no time now for
such concerns.

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>

memcg: simplify force_empty and move_lists

As for force_empty, though this may not be the main topic here,
mem_cgroup_force_empty_list() can be implemented simpler.  It is possible to
make the function just call mem_cgroup_uncharge_page() instead of releasing
page_cgroups by itself.  The tip is to call get_page() before invoking
mem_cgroup_uncharge_page(), so the page won't be released during this
function.

Kamezawa-san points out that by the time mem_cgroup_uncharge_page() uncharges,
the page might have been reassigned to an lru of a different mem_cgroup, and
now be emptied from that; but Hugh claims that's okay, the end state is the
same as when it hasn't gone to another list.

And once force_empty stops taking lock_page_cgroup within mz->lru_lock,
mem_cgroup_move_lists() can be simplified to take mz->lru_lock directly while
holding page_cgroup lock (but still has to use try_lock_page_cgroup).

Signed-off-by: Hirokazu Takahashi <taka@valinux.co.jp>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: fix mem_cgroup_move_lists locking

Ever since the VM_BUG_ON(page_get_page_cgroup(page)) (now Bad page state) went
into page freeing, I've hit it from time to time in testing on some machines,
sometimes only after many days.  Recently found a machine which could usually
produce it within a few hours, which got me there at last.

The culprit is mem_cgroup_move_lists, whose locking is inadequate; and the
arrangement of structures was such that you got page_cgroups from the lru list
neatly put on to SLUB's freelist.  Kamezawa-san identified the same hole
independently.

The main problem was that it was missing the lock_page_cgroup it needs to
safely page_get_page_cgroup; but it's tricky to go beyond that too, and I
couldn't do it with SLAB_DESTROY_BY_RCU as I'd expected.  See the code for
comments on the constraints.

This patch immediately gets replaced by a simpler one from Hirokazu-san; but
is it just foolish pride that tells me to put this one on record, in case we
need to come back to it later?

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: css_put after remove_list

mem_cgroup_uncharge_page does css_put on the mem_cgroup before uncharging from
it, and before removing page_cgroup from one of its lru lists: isn't there a
danger that struct mem_cgroup memory could be freed and reused before
completing that, so corrupting something?  Never seen it, and for all I know
there may be other constraints which make it impossible; but let's be
defensive and reverse the ordering there.

mem_cgroup_force_empty_list is safe because there's an extra css_get around
all its works; but even so, change its ordering the same way round, to help
get in the habit of doing it like this.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: remove clear_page_cgroup and atomics

Remove clear_page_cgroup: it's an unhelpful helper, see for example how
mem_cgroup_uncharge_page had to unlock_page_cgroup just in order to call it
(serious races from that?  I'm not sure).

Once that's gone, you can see it's pointless for page_cgroup's ref_cnt to be
atomic: it's always manipulated under lock_page_cgroup, except where
force_empty unilaterally reset it to 0 (and how does uncharge's
atomic_dec_and_test protect against that?).

Simplify this page_cgroup locking: if you've got the lock and the pc is
attached, then the ref_cnt must be positive: VM_BUG_ONs to check that, and to
check that pc->page matches page (we're on the way to finding why sometimes it
doesn't, but this patch doesn't fix that).

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: memcontrol uninlined and static

More cleanup to memcontrol.c, this time changing some of the code generated.
Let the compiler decide what to inline (except for page_cgroup_locked which is
only used when CONFIG_DEBUG_VM): the __always_inline on lock_page_cgroup etc.
was quite a waste since bit_spin_lock etc.  are inlines in a header file; made
mem_cgroup_force_empty and mem_cgroup_write_strategy static.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: memcontrol whitespace cleanups

Sorry, before getting down to more important changes, I'd like to do some
cleanup in memcontrol.c.  This patch doesn't change the code generated, but
cleans up whitespace, moves up a double declaration, removes an unused enum,
removes void returns, removes misleading comments, that kind of thing.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: remove mem_cgroup_uncharge

Nothing uses mem_cgroup_uncharge apart from mem_cgroup_uncharge_page, (a
trivial wrapper around it) and mem_cgroup_end_migration (which does the same
as mem_cgroup_uncharge_page).  And it often ends up having to lock just to let
its caller unlock.  Remove it (but leave the silly locking until a later
patch).

Moved mem_cgroup_cache_charge next to mem_cgroup_charge in memcontrol.h.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: mem_cgroup_charge never NULL

My memcgroup patch to fix hang with shmem/tmpfs added NULL page handling to
mem_cgroup_charge_common.  It seemed convenient at the time, but hard to
justify now: there's a perfectly appropriate swappage to charge and uncharge
instead, this is not on any hot path through shmem_getpage, and no performance
hit was observed from the slight extra overhead.

So revert that NULL page handling from mem_cgroup_charge_common; and make it
clearer by bringing page_cgroup_assign_new_page_cgroup into its body - that
was a helper I found more of a hindrance to understanding.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: bad page if page_cgroup when free

Replace free_hot_cold_page's VM_BUG_ON(page_get_page_cgroup(page)) by a "Bad
page state" and clear: most users don't have CONFIG_DEBUG_VM on, and if it
were set here, it'd likely cause corruption when the page is reused.

Don't use page_assign_page_cgroup to clear it: that should be private to
memcontrol.c, and always called with the lock taken; and memmap_init_zone
doesn't need it either - like page->mapping and other pointers throughout the
kernel, Linux assumes pointers in zeroed structures are NULL pointers.

Instead use page_reset_bad_cgroup, added to memcontrol.h for this only.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: fix VM_BUG_ON from page migration

Page migration gave me free_hot_cold_page's VM_BUG_ON page->page_cgroup.
remove_migration_pte was calling mem_cgroup_charge on the new page whenever it
found a swap pte, before it had determined it to be a migration entry.  That
left a surplus reference count on the page_cgroup, so it was still attached
when the page was later freed.

Move that mem_cgroup_charge down to where we're sure it's a migration entry.
We were already under i_mmap_lock or anon_vma->lock, so its GFP_KERNEL was
already inappropriate: change that to GFP_ATOMIC.

It's essential that remove_migration_pte removes all the migration entries,
other crashes follow if not.  So proceed even when the charge fails: normally
it cannot, but after a mem_cgroup_force_empty it might - comment in the code.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: when do_swap's do_wp_page fails

Don't uncharge when do_swap_page's call to do_wp_page fails: the page which
was charged for is there in the pagetable, and will be correctly uncharged
when that area is unmapped - it was only its COWing which failed.

And while we're here, remove earlier XXX comment: yes, OR in do_wp_page's
return value (maybe VM_FAULT_WRITE) with do_swap_page's there; but if it
fails, mask out success bits, which might confuse some arches e.g.  sparc.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: page_cache_release not __free_page

There's nothing wrong with mem_cgroup_charge failure in do_wp_page and
do_anonymous page using __free_page, but it does look odd when nearby code
uses page_cache_release: use that instead (while turning a blind eye to
ancient inconsistencies of page_cache_release versus put_page).

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: move_lists on page not page_cgroup

Each caller of mem_cgroup_move_lists is having to use page_get_page_cgroup:
it's more convenient if it acts upon the page itself not the page_cgroup; and
in a later patch this becomes important to handle within memcontrol.c.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: mm_match_cgroup not vm_match_cgroup

vm_match_cgroup is a perverse name for a macro to match mm with cgroup: rename
it mm_match_cgroup, matching mm_init_cgroup and mm_free_cgroup.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hirokazu Takahashi <taka@valinux.co.jp>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

markers: add an if(0) to __mark_check_format()

Wrap __mark_check_format() into an if(0) to make sure that parameters such as

trace_mark(mm_page_alloc, "order %u pfn %lu", order, page?page_to_pfn(page):0);

(where page_to_pfn() has side-effects) won't generate code because of the
__mark_check_format().

Thanks to Jan Kiszka for reporting this.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Cc: Jan Kiszka <jan.kiszka@siemens.com>
Cc: "Frank Ch. Eigler" <fche@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

markers: don't risk NULL deref in marker

get_marker() may return NULL, so test for it.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Jesper Juhl <jesper.juhl@gmail.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

.gitignore: ignore emacs backup and temporary files.

Signed-off-by: Chris Dearman <chris@mips.com>
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>

alpha: remove unused DEBUG_FORCEDAC define in IOMMU

This just removes unused DEBUG_FORCEDAC define in the IOMMU code.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

alpha: make IOMMU respect the segment boundary limits

This patch makes the IOMMU code not allocate a memory area spanning LLD's
segment boundary.

is_span_boundary() judges whether a memory area spans LLD's segment boundary.
If iommu_arena_find_pages() finds such a area, it tries to find the next
available memory area.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

alpha: IOMMU had better access to the free space bitmap at only one place

iommu_arena_find_pages duplicates the code to access to the bitmap for free
space management.  This patch convert the IOMMU code to have only one place to
access the bitmap, in the popular way that other IOMMUs (e.g.  POWER and
SPARC) do.

This patch is preparation for modifications to fix the IOMMU segment boundary
problem.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

alpha: convert IOMMU to use ALIGN()

This patch is preparation for modifications to fix the IOMMU segment boundary
problem.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include falloc.h in header-y

Include falloc.h in header-y; it defines a flag for the fallocate sysctl.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRIS: Import string.c (memcpy) from newlib: fixes compile error with gcc 4

Adrian Bunk reported another compile error with a SVN head GCC:

...
  CC      arch/cris/arch-v10/lib/string.o
/home/bunk/linux/kernel-2.6/git/linux-2.6/arch/cris/arch-v10/lib/string.c:138:
error: lvalue required as increment operand
/home/bunk/linux/kernel-2.6/git/linux-2.6/arch/cris/arch-v10/lib/string.c:138:
error: lvalue required as increment operand
/home/bunk/linux/kernel-2.6/git/linux-2.6/arch/cris/arch-v10/lib/string.c:139:
error: lvalue required as increment operand
...

This is due to the use of the construct:

*((long*)dst)++ = lc;

Which isn't legal since casts don't return an lvalue.

The solution is to import the implementation from newlib,
which is continually autotested together with GCC mainline,
and uses the construct:

*(long *) dst = lc; dst += 4;

Since this is an import of a file from newlib, I'm not touching
the formatting or correcting any checkpatch errors.

As for the earlier fix for memset.c, even if the two files for
CRIS v10 and CRIS v32 are identical at the moment, it might
be possible to tweak the CRIS v32 version.
Thus, I'm not yet folding them into the same file, at least not
until we've done some research on it.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <starvik@axis.com>
Cc: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ipwireless: fix potential tty == NULL dereference

The Coverity checker spotted the following inconsequent NULL checking in
drivers/char/pcmcia/ipwireless/network.c:ipwireless_network_packet_received()

if (tty && channel_idx == IPW_CHANNEL_RAS
&& (network->ras_control_lines &
IPW_CONTROL_LINE_DCD) != 0
&& ipwireless_tty_is_modem(tty)) {
...
else
ipwireless_tty_received(tty, data, length);

Cc: Adrian Bunk <bunk@kernel.org>
Signed-off-by: David Sterba <dsterba@suse.cz>
Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Cc: "John W. Linville" <linville@tuxdriver.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sm501: add support for the SM502 programmable PLL

SM502 has a programmable PLL which can provide the panel pixel clock instead
of the 288MHz and 336MHz PLLs.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Ville Syrjala <syrjala@sci.fi>
Cc: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sm501: remove a duplicated table

misc_div is a subset of px_div so eliminate the smaller table.

Signed-off-by: Ville Syrjala <syrjala@sci.fi>
Cc: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sm501fb: fix timing limits

Vertical sync height register can only hold 6 bits.  Fix the hsync start test
to use > instead of >=.  Also add a few clarifying comments.

Signed-off-by: Ville Syrjala <syrjala@sci.fi>
Acked-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sm501fb: set transp.offset to 0 in 8bpp and 16bpp modes

Even though it may not be strictly necessary transp.offset should probably be
0 when alpha channel is not available.

Signed-off-by: Ville Syrjala <syrjala@sci.fi>
Acked-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sm501fb: RGB offsets are reversed in 16bpp modes

The RGB offsets were reversed in 16bpp modes.  Simply trying to reverse the
offsets when endianness differs is clearly the wrong thing to do but that is
an issue for another patch.

Signed-off-by: Ville Syrjala <syrjala@sci.fi>
Acked-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sm501fb: direct color visual does not work

The sm501fb palette code clearly does not handle direct color so change the
driver to use true color visual for 16bpp.

Signed-off-by: Ville Syrjala <syrjala@sci.fi>
Acked-by: Magnus Damm <damm@igel.co.jp>
Acked-by: Ben Dooks <ben-linux@fluff.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ndelay(): switch to C function to avoid 64-bit division

We should be able to do ndelay(some_u64), but that can cause a call to
__divdi3() to be emitted because the ndelay() macros does a divide.

Fix it by switching to static inline which will force the u64 arg to be
treated as an unsigned long.  udelay() takes an unsigned long arg.

[bunk@kernel.org: reported m68k build breakage]
Cc: Adrian Bunk <bunk@kernel.org>
Cc: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Cc: Martin Michlmayr <tbm@cyrius.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ds1wm: report bus reset error

The patch replaces dev_dbg() by dev_err(), so the user could actually see the
error, instead of wondering why w1 doesn't work.  The root cause of the bus
reset error isn't yet debugged though, but this sometimes happens on iPaq
H5555.

And while I'm at it, some cosmetic cleanups also made (few lines were using
spaces instead of tabs).

Signed-off-by: Anton Vorontsov <cbouatmailru@gmail.com>
Acked-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ds1wm: should check for IS_ERR(clk) instead of NULL

On the error condition clk_get() returns ERR_PTR(..), so checking for NULL
doesn't work.  ds1wm module causes a kernel oops when ds1wm clock isn't
registered.

This patch converts NULL check to IS_ERR(), plus uses PTR_ERR()
for the return code.

Signed-off-by: Anton Vorontsov <cbouatmailru@gmail.com>
Acked-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

powerpc: mpc5200: fix build error on mpc52xx_psc_spi device driver

Commit id 94f389485e27641348c1951ab8d65157122a8939 (Separate MPC52xx PSC FIOF
regsiters from the rest of PSC) split the PSC fifo registers away from the
core PSC regs.  Doing so broke the mpc52xx_psc_spi driver.

This patch teaches the mpc52xx_psc_spi driver about the new PSC fifo
register definitions.

Signed-off-by: Grant Likely <grant.likely@secretlab.ca>
Cc: David Brownell <david-b@pacbell.net>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Kumar Gala <galak@gate.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pktcdvd: reduce stack consumption

On my system, pkt_open() consumes 584 bytes because the compiler decides to
inline lots of functions that would not normally be part of long call chains.
The following patch fixes that problem on my system.

Signed-off-by: Peter Osterlund <petero2@telia.com>
Cc: Nix <nix@esperi.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

add noinline_for_stack

People are adding `noinline' in various places to prevent excess stack
consumption due to gcc inlining.  But once this is done, it is quite unobvious
why the `noinline' is present in the code.  We can comment each and every
site, or we can use noinline_for_stack.

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tridentfb: resource management fixes in probe function

Correct error paths in probe function.

The probe function enables mmio mode so it important to disable the mmio
mode before exiting the probe function.  Otherwise, the console is left in
unusable state (garbled fonts at least, lock up at worst).

[akpm@linux-foundation.org: cleanups]
Signed-off-by: Krzysztof Helt <krzysztof.h1@wp.pl>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: rename to Memory Resource Controller

Rename Memory Controller to Memory Resource Controller.  Reflect the same
changes in the CONFIG definition for the Memory Resource Controller.  Group
together the config options for Resource Counters and Memory Resource
Controller.

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Kprobes: move kprobe examples to samples/

Move kprobes examples from Documentation/kprobes.txt to under samples/.
Patch originally by Randy Dunlap.

o Updated the patch to apply on 2.6.25-rc3
o Modified examples code to build on multiple architectures. Currently,
  the kprobe and jprobe examples code works for x86 and powerpc
o Cleaned up unneeded #includes
o Cleaned up Kconfig per Sam Ravnborg's suggestions to fix build break
  on archs that don't have kretprobes
o Implemented suggestions by Mathieu Desnoyers on CONFIG_KRETPROBES
o Included Andrew Morton's cleanup based on x86-git
o Modified kretprobe_example to act as a arch-agnostic module to
  determine routine execution times:
Use 'modprobe kretprobe_example func=<func_name>' to determine
execution time of func_name in nanoseconds.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Kprobes: indicate kretprobe support in Kconfig

Add CONFIG_HAVE_KRETPROBES to the arch/<arch>/Kconfig file for relevant
architectures with kprobes support.  This facilitates easy handling of
in-kernel modules (like samples/kprobes/kretprobe_example.c) that depend on
kretprobes being present in the kernel.

Thanks to Sam Ravnborg for helping make the patch more lean.

Per Mathieu's suggestion, added CONFIG_KRETPROBES and fixed up dependencies.

Signed-off-by: Ananth N Mavinakayanahalli <ananth@in.ibm.com>
Acked-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Acked-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

VT notifier fix for VT switch

VT notifier callbacks need to be aware of console switches.  This is already
partially done from console_callback(), but at that time fg_console, cursor
positions, etc.  are not yet updated and hence screen readers fetch the old
values.

This adds an update notify after all of the values are updated in
redraw_screen(vc, 1).

Signed-off-by: Samuel Thibault <samuel.thibault@ens-lyon.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

alloc_percpu() fails to allocate percpu data

Some oprofile results obtained while using tbench on a 2x2 cpu machine were
very surprising.

For example, loopback_xmit() function was using high number of cpu cycles
to perform the statistic updates, supposed to be real cheap since they use
percpu data

        pcpu_lstats = netdev_priv(dev);
        lb_stats = per_cpu_ptr(pcpu_lstats, smp_processor_id());
        lb_stats->packets++;  /* HERE : serious contention */
        lb_stats->bytes += skb->len;

struct pcpu_lstats is a small structure containing two longs.  It appears
that on my 32bits platform, alloc_percpu(8) allocates a single cache line,
instead of giving to each cpu a separate cache line.

Using the following patch gave me impressive boost in various benchmarks
( 6 % in tbench)
(all percpu_counters hit this bug too)

Long term fix (ie >= 2.6.26) would be to let each CPU allocate their own
block of memory, so that we dont need to roudup sizes to L1_CACHE_BYTES, or
merging the SGI stuff of course...

Note : SLUB vs SLAB is important here to *show* the improvement, since they
dont have the same minimum allocation sizes (8 bytes vs 32 bytes).  This
could very well explain regressions some guys reported when they switched
to SLUB.

Signed-off-by: Eric Dumazet <dada1@cosmosbay.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

vfs: fix NULL pointer dereference in fsync_buffers_list()

Fix NULL pointer dereference in fsync_buffers_list() introduced by recent fix
of races in private_list handling.  Since bh->b_assoc_map has been cleared in
__remove_assoc_queue() we should really use original value stored in the
'mapping' variable.

Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

zlc_setup(): handle jiffies wraparound

jiffies subtraction may cause an overflow problem.  It should be using
time_after().

[akpm@linux-foundation.org: include jiffies.h]
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
Cc: Paul Jackson <pj@sgi.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Fix "Malformed early option 'loglevel'"

Keith Mannthey said:

  The parameter hotadd_percent is setup right but there is a "Malformed
  early option 'numa'" message.

Rusty Russell said:

  This happens when the function registered with early_param() returns
  non-zero.  __setup() functions return 1 if OK, module_param() and
  early_param() return 0 or a -ve error code.

For instance:

Linux version 2.6.25-rc3-t (raa@steel) (gcc version 4.1.3 20070929 (prerelease) (Ubuntu 4.1.2-16ubuntu2)) #22 SMP PREEMPT Tue Feb 26
BIOS-provided physical RAM map:
 BIOS-e820: 0000000000000000 - 00000000000a0000 (usable)
 BIOS-e820: 00000000000f0000 - 0000000000100000 (reserved)
 BIOS-e820: 0000000000100000 - 000000003fff0000 (usable)
 BIOS-e820: 000000003fff0000 - 000000003fff3000 (ACPI NVS)
 BIOS-e820: 000000003fff3000 - 0000000040000000 (ACPI data)
 BIOS-e820: 00000000fec00000 - 0000000100000000 (reserved)
Malformed early option 'loglevel'
127MB HIGHMEM available.
896MB LOWMEM available.

Command line:

BOOT_IMAGE=2.6.25-t ro root=809 ro console=ttyS0,57600n8 console=tty0 loglevel=5

Acked-by: Yinghai Lu <yhlu.kernel@gmai.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Keith Mannthey <kmannth@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

core dump: user_regset writeback

This makes the user_regset-based core dump code call user_regset writeback
hooks when available.  This is necessary groundwork to allow IA64 to set
CORE_DUMP_USE_REGSET.

Cc: Shaohua Li <shaohua.li@intel.com>
Signed-off-by: Roland McGrath <roland@redhat.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>

Add memory resource controller maintainers

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Control Groups: add Paul Menage as maintainer

Control Groups: Add Paul Menage as maintainer

Signed-off-by: Paul Menage <menage@google.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: YAMAMOTO Takashi <yamamoto@valinux.co.jp>
Cc: Hugh Dickins <hugh@veritas.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gpio: <linux/gpio.h> and "no GPIO support here" stubs

Add a <linux/gpio.h> defining fail/warn stubs for GPIO calls on platforms that
don't support the GPIO programming interface.  That includes the arch-specific
implementation glue otherwise.

This facilitates a new model for GPIO usage: drivers that can use GPIOs if
they're available, but don't require them.  One example of such a driver is
NAND driver for various FreeScale chips.  On platforms update with GPIO
support, they can be used instead of a worst-case delay to verify that the
BUSY signal is off.

(Also includes a couple minor unrelated doc updates.)

Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

specialix.c: fix possible double-unlock

Noticed by sparse, trivial to see:
drivers/char/specialix.c:2112:3: warning: context imbalance in 'sx_throttle' - unexpected unlock

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

rtc: add support for the S-35390A RTC chip

This adds basic get/set time support for the Seiko Instruments S-35390A.
This chip communicates using I2C and is used on the QNAP TS-109/TS-209 NAS
devices.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Byron Bradley <byron.bbradley@gmail.com>
Acked-by: Jean Delvare <khali@linux-fr.org>
Acked-by: David Brownell <david-b@pacbell.net>
Tested-by: Tim Ellis <tim@ngndg.com>
Cc: Alessandro Zummo <a.zummo@towertech.it>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory Resource Controller use strstrip while parsing arguments

The memory controller has a requirement that while writing values, we need
to use echo -n. This patch fixes the problem and makes the UI more consistent.

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

CRIS v10: Include mm.h instead of vmstat.h in kernel/time.c

Commit 2f569afd9ced9ebec9a6eb3dbf6f83429be0a7b4
(CONFIG_HIGHPTE vs. sub-page page tables) introduced use of
inc_zone_page_state and dec_zone_page_state in include/linux/mm.h.

Those are defined in include/linux/vmstat.h, but after it includes
mm.h, making it impossible to include vmstat.h since inc_zone_page_state
and dec_zone_page_state then would be undefined.

arch/cris/arch-v10/kernel/time.c does just this, which makes the
CRIS v10 build break with the following error:

...
  CC      arch/cris/arch-v10/kernel/time.o
In file included from include/linux/vmstat.h:7,
                 from arch/cris/arch-v10/kernel/time.c:17:
include/linux/mm.h: In function 'pgtable_page_ctor':
include/linux/mm.h:902: error: implicit declaration of function 'inc_zone_page_state'
include/linux/mm.h: In function 'pgtable_page_dtor':
include/linux/mm.h:908: error: implicit declaration of function 'dec_zone_page_state'
make[2]: *** [arch/cris/arch-v10/kernel/time.o] Error 1
make[1]: *** [arch/cris/arch-v10/kernel] Error 2
make: *** [sub-make] Error 2
...

By changing kernel/time.c to include linux/mm.h, the build succeeds.

Signed-off-by: Jesper Nilsson <jesper.nilsson@axis.com>
Cc: Mikael Starvik <mikael.starvik@axis.com>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

serial: add PNP ID GVC0303 for Archtek 3334BRV ISA modem

Thomas Lehmann <thomas.lehmann@alumni.tu-berlin.de> verified that this
entry works.

Signed-off-by: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

update checkpatch.pl to version 0.15

This version brings a number of minor fixes updating the type detector and
the unary tracker.  It also brings a few small fixes for false positives.
It also reverts the --file warning.  Of note:

 - limit CVS checks to added lines
 - improved type detections
 - fixes to the unary tracker

Andy Whitcroft (13):
      Version: 0.15
      EXPORT_SYMBOL checks need to accept array variables
      export checks must match DECLARE_foo and LIST_HEAD
      possible types: cleanup debugging missing line
      values: track values through preprocessor conditional paths
      typeof is actually a type
      possible types: detect definitions which cross lines
      values: include line numbers on value debug information
      values: ensure we find correctly record pending brackets
      values: simplify the brace history stack
      CVS keyword checks should only apply to added lines
      loosen spacing for comments
      allow braces for single statement blocks with multiline conditionals

Harvey Harrison (1):
      checkpatch: remove fastcall

Ingo Molnar (1):
      checkpatch.pl: revert wrong --file message

Uwe Kleine-Koenig (1):
      fix typo "goot" -> "good"

Signed-off-by: Andy Whitcroft <apw@shadowen.org>
Cc: Randy Dunlap <rdunlap@xenotime.net>
Cc: Joel Schopp <jschopp@austin.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cgroup: fix default notify_on_release setting

The documentation says the default value of notify_on_release of a child
cgroup is inherited from its parent, which is reasonable, but the
implementation just sets the flag disabled.

Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

x86: a P4 is a P6 not an i486

P4 has been coming out as CPU_FAMILY=4 instead of 6: fix MPENTIUM4 typo.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  scsi: missing add of padded bytes to io completion byte count

Merge branch 'upstream-linus' of git://git./linux/kernel/git/mfasheh/ocfs2

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2:
  [PATCH] fs/ocfs2/aops.c: Correct use of ! and &
  [2.6 patch] ocfs2: make dlm_do_assert_master() static
  [2.6 patch] make ocfs2_downconvert_thread() static
  [2.6 patch] fs/ocfs2/: possible cleanups
  [PATCH] ocfs2: le*_add_cpu conversion
  ocfs2: Fix writeout in ocfs2_data_convert_worker()
  ocfs2: Enable localalloc for local mounts

Merge branch 'fixes' of git://git./linux/kernel/git/djbw/async_tx

* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/djbw/async_tx:
  ioat: fix 'ack' handling, driver must ensure that 'ack' is zero
  dmaengine: fix sparse warning
  fsldma: do not cleanup descriptors in hardirq context
  dmaengine: add driver for Freescale MPC85xx DMA controller

scsi: missing add of padded bytes to io completion byte count

Original patch from Tejun Heo <htejun@gmail.com> but should use ->extra_len
and not ->data_len, as we would then overshoot the original request size.

Signed-off-by: Jens Axboe <jens.axboe@oracle.com>

Merge branch 'for-linus' of git://git./linux/kernel/git/mingo/linux-2.6-sched-devel

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mingo/linux-2.6-sched-devel:
  sched: revert load_balance_monitor() changes

Merge branch 'for-linus' of git://git./linux/kernel/git/x86/linux-2.6-x86

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-x86:
  x86/xen: fix DomU boot problem
  x86: not set node to cpu_to_node if the node is not online
  x86, i387: fix ptrace leakage using init_fpu()

Merge branch 'for-linus' of git://git./linux/kernel/git/avi/kvm

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/avi/kvm:
  x86: disable KVM for Voyager and friends
  KVM: VMX: Avoid rearranging switched guest msrs while they are loaded
  KVM: MMU: Fix race when instantiating a shadow pte
  KVM: Route irq 0 to vcpu 0 exclusively
  KVM: Avoid infinite-frequency local apic timer
  KVM: make MMU_DEBUG compile again
  KVM: move alloc_apic_access_page() outside of non-preemptable region
  KVM: SVM: fix Windows XP 64 bit installation crash
  KVM: remove the usage of the mmap_sem for the protection of the memory slots.
  KVM: emulate access to MSR_IA32_MCG_CTL
  KVM: Make the supported cpuid list a host property rather than a vm property
  KVM: Fix kvm_arch_vcpu_ioctl_set_sregs so that set_cr0 works properly
  KVM: SVM: set NM intercept when enabling CR0.TS in the guest
  KVM: SVM: Fix lazy FPU switching

ioat: fix 'ack' handling, driver must ensure that 'ack' is zero

Initialize 'ack' to zero in case the descriptor has been recycled.

Prevents "kernel BUG at crypto/async_tx/async_xor.c:185!"

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Shannon Nelson <shannon.nelson@intel.com>
Cc: stable@kernel.org

dmaengine: fix sparse warning

include/linux/dmaengine.h:364:2: warning: returning void-valued expression

Signed-off-by: Dan Williams <dan.j.williams@intel.com>

fsldma: do not cleanup descriptors in hardirq context

"Cleaning" descriptors involves calling pending callbacks and clients
assume that their callback will only ever happen in softirq context.
Delay cleanup to the tasklet.

Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Acked-by: Zhang Wei <wei.zhang@freescale.com>

dmaengine: add driver for Freescale MPC85xx DMA controller

The driver implements DMA engine API for Freescale MPC85xx DMA controller,
which could be used by devices in the silicon.  The driver supports the
Basic mode of Freescale MPC85xx DMA controller.  The MPC85xx processors
supported include MPC8540/60, MPC8555, MPC8548, MPC8641 and so on.

The MPC83xx(MPC8349, MPC8360) are also supported.

[kamalesh@linux.vnet.ibm.com: build fix]
[dan.j.williams@intel.com: merge mm fixes, rebase on async_tx-2.6.25]
Signed-off-by: Zhang Wei <wei.zhang@freescale.com>
Signed-off-by: Ebony Zhu <ebony.zhu@freescale.com>
Acked-by: Kumar Gala <galak@gate.crashing.org>
Cc: Shannon Nelson <shannon.nelson@intel.com>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>

sched: revert load_balance_monitor() changes

The following commits cause a number of regressions:

  commit 58e2d4ca581167c2a079f4ee02be2f0bc52e8729
  Author: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
  Date:   Fri Jan 25 21:08:00 2008 +0100
  sched: group scheduling, change how cpu load is calculated

  commit 6b2d7700266b9402e12824e11e0099ae6a4a6a79
  Author: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
  Date:   Fri Jan 25 21:08:00 2008 +0100
  sched: group scheduler, fix fairness of cpu bandwidth allocation for task groups

Namely:
 - very frequent wakeups on SMP, reported by PowerTop users.
 - cacheline trashing on (large) SMP
 - some latencies larger than 500ms

While there is a mergeable patch to fix the latter, the former issues
are not fixable in a manner suitable for .25 (we're at -rc3 now).

Hence we revert them and try again in v2.6.26.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
CC: Srivatsa Vaddagiri <vatsa@linux.vnet.ibm.com>
Tested-by: Alexey Zaytsev <alexey.zaytsev@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86/xen: fix DomU boot problem

Construct Xen guest e820 map with a hole between 640K-1M.

It's pure luck that Xen kernels have gotten away with it in the past.

The patch below seems like the right thing to do. It certainly boots in
a domU without the DMI problem (without any of the other related patches
such as Alexander's).

Signed-off-by: Ian Campbell <ijc@hellion.org.uk>
Cc: H. Peter Anvin <hpa@zytor.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Tested-by: Mark McLoughlin <markmc@redhat.com>
Acked-by: Mark McLoughlin <markmc@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Thomas Gleixner <tglx@linutronix.de>

x86: not set node to cpu_to_node if the node is not online

resolve boot problem reported by Mel Gorman:

   http://lkml.org/lkml/2008/2/13/404

init_cpu_to_node will use cpu->apic (from MADT or mptable) and
apic->node(from SRAT or AMD config space with k8_bus_64.c) to have
cpu->node mapping, and later identify_cpu will overwrite them
again...(with nearby_node...)

this patch checks if the node is online, otherwise it will not
update cpu_node map. so keep cpu_node map to online node before
identify_cpu..., to prevent possible error.

Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Thomas Gleixner <tglx@linutronix.de>

x86, i387: fix ptrace leakage using init_fpu()

This bug got introduced by the recent i387 merge:

  commit 4421011120b2304e5c248ae4165a2704588aedf1
  Author: Roland McGrath <roland@redhat.com>
  Date:   Wed Jan 30 13:31:50 2008 +0100

      x86: x86 i387 user_regset

Current usage of unlazy_fpu() in ptrace specific routines is wrong.
unlazy_fpu() will not init fpu if the task never used math. So the
ptrace calls can expose the parent tasks FPU data in some cases.

Replace it with the init_fpu() which will init the math state, if the
task never used math before.

Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
Acked-by: Thomas Gleixner <tglx@linutronix.de>

Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block

* 'for-linus' of git://git.kernel.dk/linux-2.6-block:
  block: fix blkdev_issue_flush() not detecting and passing EOPNOTSUPP back
  block: fix shadowed variable warning in blk-map.c
  block: remove extern on function definition
  cciss: remove READ_AHEAD define and use block layer defaults
  make cdrom.c:check_for_audio_disc() static
  block/genhd.c: proper externs
  unexport blk_rq_map_user_iov
  unexport blk_{get,put}_queue
  block/genhd.c: cleanups
  proper prototype for blk_dev_init()
  block/blk-tag.c should #include "blk.h"
  Fix DMA access of block device in 64-bit kernel on some non-x86 systems with 4GB or upper 4GB memory
  block: separate out padding from alignment
  block: restore the meaning of rq->data_len to the true data length
  resubmit: cciss: procfs updates to display info about many
  splice: only return -EAGAIN if there's hope of more data
  block: fix kernel-docbook parameters and files

m68k{,nommu}: Wire up new timerfd syscalls

m68k{,nommu}: Wire up the new timerfd syscalls, which were introduced in
commit 4d672e7ac79b5ec5cdc90e450823441e20464691 ("timerfd: new timerfd API").

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Acked-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

m68knommu: fix fec driver interrupt races

The FEC driver has a common interrupt handler for all interrupt event
types. It is raised on a number of distinct interrupt vectors.
This handler can't be re-entered while processing an interrupt, so
make sure all requested vectors are flagged as IRQF_DISABLED.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

m68knommu: declare do_IRQ()

Need a declaration of do_IRQ for the 68328 interrupt handling code.
It is common to all m68knommu targets, so a common declaration makes
sense.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

m68knommu: remove duplicate hw_tick() code

Remove duplicate hw_tick() function from 68328 timers code.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

m68knommu: update defconfig

Update the m68knommu defconfig.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Build fix for drivers/s390/char/defkeymap.c

Commit 5ce2087ed0eb424e0889bdc9102727f65d2ecdde (Fix default compose
table initialization) left a trailing quote.

  CC      drivers/s390/char/defkeymap.o
drivers/s390/char/defkeymap.c:155: error: missing terminating ' character
drivers/s390/char/defkeymap.c:156: error: syntax error before ';' token
make[3]: *** [drivers/s390/char/defkeymap.o] Error 1

Fix that.

Signed-off-by: Tony Breeds <tony@bakeyournoodle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge git://git./linux/kernel/git/sfrench/cifs-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
  [CIFS] remove unused variable
  [CIFS] consolidate duplicate code in posix/unix inode handling
  [CIFS] fix build break when proc disabled
  [CIFS] factoring out common code in get_inode_info functions
  [CIFS] fix prepath conversion when server supports posix paths
  [CIFS] Only convert / when server does not support posix paths
  [CIFS] Fix mixed case name in structure dfs_info3_param
  [CIFS] fixup prefixpaths which contain multiple path components
  [CIFS] fix typo
  [CIFS] patch to fix incorrect encoding of number of aces on set mode
  [CIFS] Fix typo in quota operations
  [CIFS] clean up some hard to read ifdefs
  [CIFS] reduce checkpatch warnings
  [CIFS] fix warning in cifs_spnego.c

freezer vs stopped or traced

This changes the "freezer" code used by suspend/hibernate in its treatment
of tasks in TASK_STOPPED (job control stop) and TASK_TRACED (ptrace) states.

As I understand it, the intent of the "freezer" is to hold all tasks
from doing anything significant.  For this purpose, TASK_STOPPED and
TASK_TRACED are "frozen enough".  It's possible the tasks might resume
from ptrace calls (if the tracer were unfrozen) or from signals
(including ones that could come via timer interrupts, etc).  But this
doesn't matter as long as they quickly block again while "freezing" is
in effect.  Some minor adjustments to the signal.c code make sure that
try_to_freeze() very shortly follows all wakeups from both kinds of
stop.  This lets the freezer code safely leave stopped tasks unmolested.

Changing this fixes the longstanding bug of seeing after resuming from
suspend/hibernate your shell report "[1] Stopped" and the like for all
your jobs stopped by ^Z et al, as if you had freshly fg'd and ^Z'd them.
It also removes from the freezer the arcane special case treatment for
ptrace'd tasks, which relied on intimate knowledge of ptrace internals.

Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

x86: disable KVM for Voyager and friends

Most classic Pentiums don't have hardware virtualization extension,
and building kvm with Voyager, Visual Workstation, or NUMAQ
generates spurious failures.

Signed-off-by: Avi Kivity <avi@qumranet.com>
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>

KVM: VMX: Avoid rearranging switched guest msrs while they are loaded

KVM tries to run as much as possible with the guest msrs loaded instead of
host msrs, since switching msrs is very expensive.  It also tries to minimize
the number of msrs switched according to the guest mode; for example,
MSR_LSTAR is needed only by long mode guests.  This optimization is done by
setup_msrs().

However, we must not change which msrs are switched while we are running with
guest msr state:

 - switch to guest msr state
 - call setup_msrs(), removing some msrs from the list
 - switch to host msr state, leaving a few guest msrs loaded

An easy way to trigger this is to kexec an x86_64 linux guest.  Early during
setup, the guest will switch EFER to not include SCE.  KVM will stop saving
MSR_LSTAR, and on the next msr switch it will leave the guest LSTAR loaded.
The next host syscall will end up in a random location in the kernel.

Fix by reloading the host msrs before changing the msr list.

Signed-off-by: Avi Kivity <avi@qumranet.com>

KVM: MMU: Fix race when instantiating a shadow pte

For improved concurrency, the guest walk is performed concurrently with other
vcpus.  This means that we need to revalidate the guest ptes once we have
write-protected the guest page tables, at which point they can no longer be
modified.

The current code attempts to avoid this check if the shadow page table is not
new, on the assumption that if it has existed before, the guest could not have
modified the pte without the shadow lock.  However the assumption is incorrect,
as the racing vcpu could have modified the pte, then instantiated the shadow
page, before our vcpu regains control:

  vcpu0        vcpu1

  fault
  walk pte

               modify pte
               fault in same pagetable
               instantiate shadow page

  lookup shadow page
  conclude it is old
  instantiate spte based on stale guest pte

We could do something clever with generation counters, but a test run by
Marcelo suggests this is unnecessary and we can just do the revalidation
unconditionally.  The pte will be in the processor cache and the check can
be quite fast.

Signed-off-by: Avi Kivity <avi@qumranet.com>