mm: compaction: minimise the time IRQs are disabled while isolating pages for migration

compaction_alloc() isolates pages for migration in isolate_migratepages.
While it's scanning, IRQs are disabled on the mistaken assumption the
scanning should be short.  Tests show this to be true for the most part
but contention times on the LRU lock can be increased.  Before this patch,
the IRQ disabled times for a simple test looked like

  Total sampled time IRQs off (not real total time): 5493
  Event shrink_inactive_list..shrink_zone                  1596 us count 1
  Event shrink_inactive_list..shrink_zone                  1530 us count 1
  Event shrink_inactive_list..shrink_zone                   956 us count 1
  Event shrink_inactive_list..shrink_zone                   541 us count 1
  Event shrink_inactive_list..shrink_zone                   531 us count 1
  Event split_huge_page..add_to_swap                        232 us count 1
  Event save_args..call_softirq                              36 us count 1
  Event save_args..call_softirq                              35 us count 2
  Event __wake_up..__wake_up                                  1 us count 1

This patch reduces the worst-case IRQs-disabled latencies by releasing the
lock every SWAP_CLUSTER_MAX pages that are scanned and releasing the CPU if
necessary. The cost of this is that the processing performing compaction will
be slower but IRQs being disabled for too long a time has worse consequences
as the following report shows;

  Total sampled time IRQs off (not real total time): 4367
  Event shrink_inactive_list..shrink_zone                   881 us count 1
  Event shrink_inactive_list..shrink_zone                   875 us count 1
  Event shrink_inactive_list..shrink_zone                   868 us count 1
  Event shrink_inactive_list..shrink_zone                   555 us count 1
  Event split_huge_page..add_to_swap                        495 us count 1
  Event compact_zone..compact_zone_order                    269 us count 1
  Event split_huge_page..add_to_swap                        266 us count 1
  Event shrink_inactive_list..shrink_zone                    85 us count 1
  Event save_args..call_softirq                              36 us count 2
  Event __wake_up..__wake_up                                  1 us count 1

[akpm@linux-foundation.org: simplify with s/unlocked/locked/]
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Arthur Marsh <arthur.marsh@internode.on.net>
Cc: Clemens Ladisch <cladisch@googlemail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: compaction: minimise the time IRQs are disabled while isolating free pages

compaction_alloc() isolates free pages to be used as migration targets.
While its scanning, IRQs are disabled on the mistaken assumption the
scanning should be short.  Analysis showed that IRQs were in fact being
disabled for substantial time.  A simple test was run using large
anonymous mappings with transparent hugepage support enabled to trigger
frequent compactions.  A monitor sampled what the worst IRQ-off latencies
were and a post-processing tool found the following;

  Total sampled time IRQs off (not real total time): 22355
  Event compaction_alloc..compaction_alloc                 8409 us count 1
  Event compaction_alloc..compaction_alloc                 7341 us count 1
  Event compaction_alloc..compaction_alloc                 2463 us count 1
  Event compaction_alloc..compaction_alloc                 2054 us count 1
  Event shrink_inactive_list..shrink_zone                  1864 us count 1
  Event shrink_inactive_list..shrink_zone                    88 us count 1
  Event save_args..call_softirq                              36 us count 1
  Event save_args..call_softirq                              35 us count 2
  Event __make_request..__blk_run_queue                      24 us count 1
  Event __alloc_pages_nodemask..__alloc_pages_nodemask        6 us count 1

i.e.  compaction is disabled IRQs for a prolonged period of time - 8ms in
one instance.  The full report generated by the tool can be found at

 http://www.csn.ul.ie/~mel/postings/minfree-20110225/irqsoff-vanilla-micro.report

This patch reduces the time IRQs are disabled by simply disabling IRQs at
the last possible minute.  An updated IRQs-off summary report then looks
like;

  Total sampled time IRQs off (not real total time): 5493
  Event shrink_inactive_list..shrink_zone                  1596 us count 1
  Event shrink_inactive_list..shrink_zone                  1530 us count 1
  Event shrink_inactive_list..shrink_zone                   956 us count 1
  Event shrink_inactive_list..shrink_zone                   541 us count 1
  Event shrink_inactive_list..shrink_zone                   531 us count 1
  Event split_huge_page..add_to_swap                        232 us count 1
  Event save_args..call_softirq                              36 us count 1
  Event save_args..call_softirq                              35 us count 2
  Event __wake_up..__wake_up                                  1 us count 1

A full report is again available at

  http://www.csn.ul.ie/~mel/postings/minfree-20110225/irqsoff-minimiseirq-free-v1r4-micro.report

As should be obvious, IRQ disabled latencies due to compaction are
almost elimimnated for this particular test.

[aarcange@redhat.com: Fix initialisation of isolated]
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujisu.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Arthur Marsh <arthur.marsh@internode.on.net>
Cc: Clemens Ladisch <cladisch@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: don't return 0 too early from find_get_pages()

Callers of find_get_pages(), or its wrapper pagevec_lookup() - notably
truncate_inode_pages_range() - stop looking further when it returns 0.

But if an interrupt comes just after its radix_tree_gang_lookup_slot(),
especially if we have preemptible RCU enabled, isn't it conceivable that
all 14 pages returned could be removed from the page cache by
shrink_page_list(), before find_get_pages() gets to process them?  So
causing it to return 0 although there may be plenty more pages beyond.

Make find_get_pages() and find_get_pages_tag() check for this unlikely
case, and restart should it occur; but callers of find_get_pages_contig()
have no such expectation, it's okay for that to return 0 early.

I have not seen this in practice, just worried by the possibility.

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Nick Piggin <npiggin@kernel.dk>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Salman Qazi <sqazi@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove worrying dead code from find_get_pages()

The radix_tree_deref_retry() case in find_get_pages() has a strange little
excrescence, not seen in the other gang lookups: it looks like the start
of an abandoned attempt to guarantee forward progress in a case that
cannot arise.

ret should always be 0 here: if it isn't, then going back to restart will
leak references to pages already gotten.  There used to be a comment
saying nr_found is necessarily 1 here: that's not quite true, but the
radix_tree_deref_retry() case is peculiar to the entry at index 0, when we
race with it being moved out of the radix_tree root or back.

Remove the worrisome two lines, add a brief comment here and in
find_get_pages_contig() and find_get_pages_tag(), and a WARN_ON in
find_get_pages() should it ever be seen elsewhere than at 0.

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Nick Piggin <npiggin@kernel.dk>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Salman Qazi <sqazi@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hugetlbfs: correct handling of negative input to /proc/sys/vm/nr_hugepages

When the user inserts a negative value into /proc/sys/vm/nr_hugepages it
will cause the kernel to allocate as many hugepages as possible and to
then update /proc/meminfo to reflect this.

This changes the behavior so that the negative input will result in
nr_hugepages value being unchanged.

Signed-off-by: Petr Holasek <pholasek@redhat.com>
Signed-off-by: Anton Arapov <anton@redhat.com>
Reviewed-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Eric B Munson <emunson@mgebm.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: vmscan: kswapd should not free an excessive number of pages when balancing small zones

When reclaiming for order-0 pages, kswapd requires that all zones be
balanced.  Each cycle through balance_pgdat() does background ageing on
all zones if necessary and applies equal pressure on the inactive zone
unless a lot of pages are free already.

A "lot of free pages" is defined as a "balance gap" above the high
watermark which is currently 7*high_watermark.  Historically this was
reasonable as min_free_kbytes was small.  However, on systems using huge
pages, it is recommended that min_free_kbytes is higher and it is tuned
with hugeadm --set-recommended-min_free_kbytes.  With the introduction of
transparent huge page support, this recommended value is also applied.  On
X86-64 with 4G of memory, min_free_kbytes becomes 67584 so one would
expect around 68M of memory to be free.  The Normal zone is approximately
35000 pages so under even normal memory pressure such as copying a large
file, it gets exhausted quickly.  As it is getting exhausted, kswapd
applies pressure equally to all zones, including the DMA32 zone.  DMA32 is
approximately 700,000 pages with a high watermark of around 23,000 pages.
In this situation, kswapd will reclaim around (23000*8 where 8 is the high
watermark + balance gap of 7 * high watermark) pages or 718M of pages
before the zone is ignored.  What the user sees is that free memory far
higher than it should be.

To avoid an excessive number of pages being reclaimed from the larger
zones, explicitely defines the "balance gap" to be either 1% of the zone
or the low watermark for the zone, whichever is smaller.  While kswapd
will check all zones to apply pressure, it'll ignore zones that meets the
(high_wmark + balance_gap) watermark.

To test this, 80G were copied from a partition and the amount of memory
being used was recorded.  A comparison of a patch and unpatched kernel can
be seen at
http://www.csn.ul.ie/~mel/postings/minfree-20110222/memory-usage-hydra.ps
and shows that kswapd is not reclaiming as much memory with the patch
applied.

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Shaohua Li <shaohua.li@intel.com>
Cc: "Chen, Tim C" <tim.c.chen@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mempolicy: remove redundant check in __mpol_equal()

The 'flags' field is already checked, no need to do it again.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Cc: Bob Liu <lliubbo@gmail.com>
Cc: Lee Schermerhorn <lee.schermerhorn@hp.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Andi Kleen <andi@firstfloor.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

smaps: have smaps show transparent huge pages

Now that the mere act of _looking_ at /proc/$pid/smaps will not destroy
transparent huge pages, tell how much of the VMA is actually mapped with
them.

This way, we can make sure that we're getting THPs where we
expect to see them.

Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: David Rientjes <rientjes@google.com>
Reviewed-by: Eric B Munson <emunson@mgebm.net>
Tested-by: Eric B Munson <emunson@mgebm.net>
Cc: Michael J Wolf <mjwolf@us.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

smaps: teach smaps_pte_range() about THP pmds

This adds code to explicitly detect and handle pmd_trans_huge() pmds.  It
then passes HPAGE_SIZE units in to the smap_pte_entry() function instead
of PAGE_SIZE.

This means that using /proc/$pid/smaps now will no longer cause THPs to be
broken down in to small pages.

Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Reviewed-by: Eric B Munson <emunson@mgebm.net>
Tested-by: Eric B Munson <emunson@mgebm.net>
Acked-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Michael J Wolf <mjwolf@us.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

smaps: pass pte size argument in to smaps_pte_entry()

Add an argument to the new smaps_pte_entry() function to let it account in
things other than PAGE_SIZE units.  I changed all of the PAGE_SIZE sites,
even though not all of them can be reached for transparent huge pages,
just so this will continue to work without changes as THPs are improved.

Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Reviewed-by: Eric B Munson <emunson@mgebm.net>
Tested-by: Eric B Munson <emunson@mgebm.net>
Cc: Michael J Wolf <mjwolf@us.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

smaps: break out smaps_pte_entry() from smaps_pte_range()

We will use smaps_pte_entry() in a moment to handle both small and
transparent large pages.  But, we must break it out of smaps_pte_range()
first.

Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: David Rientjes <rientjes@google.com>
Reviewed-by: Eric B Munson <emunson@mgebm.net>
Tested-by: Eric B Munson <emunson@mgebm.net>
Cc: Michael J Wolf <mjwolf@us.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pagewalk: only split huge pages when necessary

Right now, if a mm_walk has either ->pte_entry or ->pmd_entry set, it will
unconditionally split any transparent huge pages it runs in to.  In
practice, that means that anyone doing a

cat /proc/$pid/smaps

will unconditionally break down every huge page in the process and depend
on khugepaged to re-collapse it later.  This is fairly suboptimal.

This patch changes that behavior.  It teaches each ->pmd_entry handler
(there are five) that they must break down the THPs themselves.  Also, the
_generic_ code will never break down a THP unless a ->pte_entry handler is
actually set.

This means that the ->pmd_entry handlers can now choose to deal with THPs
without breaking them down.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: David Rientjes <rientjes@google.com>
Reviewed-by: Eric B Munson <emunson@mgebm.net>
Tested-by: Eric B Munson <emunson@mgebm.net>
Cc: Michael J Wolf <mjwolf@us.ibm.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: reclaim invalidated page ASAP

invalidate_mapping_pages is very big hint to reclaimer.  It means user
doesn't want to use the page any more.  So in order to prevent working set
page eviction, this patch move the page into tail of inactive list by
PG_reclaim.

Please, remember that pages in inactive list are working set as well as
active list.  If we don't move pages into inactive list's tail, pages near
by tail of inactive list can be evicted although we have a big clue about
useless pages.  It's totally bad.

Now PG_readahead/PG_reclaim is shared.  fe3cba17 added ClearPageReclaim
into clear_page_dirty_for_io for preventing fast reclaiming readahead
marker page.

In this series, PG_reclaim is used by invalidated page, too.  If VM find
the page is invalidated and it's dirty, it sets PG_reclaim to reclaim
asap.  Then, when the dirty page will be writeback,
clear_page_dirty_for_io will clear PG_reclaim unconditionally.  It
disturbs this serie's goal.

I think it's okay to clear PG_readahead when the page is dirty, not
writeback time.  So this patch moves ClearPageReadahead.  In v4,
ClearPageReadahead in set_page_dirty has a problem which is reported by
Steven Barrett.  It's due to compound page.  Some driver(ex, audio) calls
set_page_dirty with compound page which isn't on LRU.  but my patch does
ClearPageRelcaim on compound page.  In non-CONFIG_PAGEFLAGS_EXTENDED, it
breaks PageTail flag.

I think it doesn't affect THP and pass my test with THP enabling but Cced
Andrea for double check.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Reported-by: Steven Barrett <damentz@liquorix.net>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: move memcg reclaimable page into tail of inactive list

The rotate_reclaimable_page function moves just written out pages, which
the VM wanted to reclaim, to the end of the inactive list.  That way the
VM will find those pages first next time it needs to free memory.

This patch applies the rule in memcg.  It can help to prevent unnecessary
working page eviction of memcg.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: deactivate invalidated pages

Recently, there are reported problem about thrashing.
(http://marc.info/?l=rsync&m=128885034930933&w=2) It happens by backup
workloads(ex, nightly rsync).  That's because the workload makes just
use-once pages and touches pages twice.  It promotes the page into active
list so that it results in working set page eviction.

Some app developer want to support POSIX_FADV_NOREUSE.  But other OSes
don't support it, either.
(http://marc.info/?l=linux-mm&m=128928979512086&w=2)

By other approach, app developers use POSIX_FADV_DONTNEED.  But it has a
problem.  If kernel meets page is writing during invalidate_mapping_pages,
it can't work.  It makes for application programmer to use it since they
always have to sync data before calling fadivse(..POSIX_FADV_DONTNEED) to
make sure the pages could be discardable.  At last, they can't use
deferred write of kernel so that they could see performance loss.
(http://insights.oetiker.ch/linux/fadvise.html)

In fact, invalidation is very big hint to reclaimer.  It means we don't
use the page any more.  So let's move the writing page into inactive
list's head if we can't truncate it right now.

Why I move page to head of lru on this patch, Dirty/Writeback page would
be flushed sooner or later.  It can prevent writeout of pageout which is
less effective than flusher's writeout.

Originally, I reused lru_demote of Peter with some change so added his
Signed-off-by.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Reported-by: Ben Gamari <bgamari.foss@gmail.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: mm_struct: remove 16 bytes of alignment padding on 64 bit builds

Reorder mm_struct to remove 16 bytes of alignment padding on 64 bit
builds.  On my config this shrinks mm_struct by enough to fit in one
fewer cache lines and allows more objects per slab in mm_struct
kmem_cache under SLUB.

slabinfo before patch :-
    Sizes (bytes)     Slabs
    --------------------------------
    Object :     848  Total  :       9
    SlabObj:     896  Full   :       2
    SlabSiz:   16384  Partial:       5
    Loss   :      48  CpuSlab:       2
    Align  :      64  Objects:      18

 slabinfo after :-
    Sizes (bytes)     Slabs
    --------------------------------
    Object :     832  Total  :       7
    SlabObj:     832  Full   :       2
    SlabSiz:   16384  Partial:       3
    Loss   :       0  CpuSlab:       2
    Align  :      64  Objects:      19

Signed-off-by: Richard Kennedy <richard@rsk.demon.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: remove unused TestSetPageLocked() interface

TestSetPageLocked() isn't being used anywhere.  Also, using it would
likely be an error, since the proper interface trylock_page() provides
stronger ordering guarantees.

Signed-off-by: Michel Lespinasse <walken@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: simplify anon_vma refcounts

This patch changes the anon_vma refcount to be 0 when the object is free.
It does this by adding 1 ref to being in use in the anon_vma structure
(iow.  the anon_vma->head list is not empty).

This allows a simpler release scheme without having to check both the
refcount and the list as well as avoids taking a ref for each entry on the
list.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: move anon_vma ref out from under CONFIG_foo

We need the anon_vma refcount unconditionally to simplify the anon_vma
lifetime rules.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: rename drop_anon_vma() to put_anon_vma()

The normal code pattern used in the kernel is: get/put.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Hugh Dickins <hughd@google.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: debug-pagealloc: fix kconfig dependency warning

Fix kconfig dependency warning to satisfy dependencies:

warning: (PAGE_POISONING) selects DEBUG_PAGEALLOC which has unmet
direct dependencies (DEBUG_KERNEL && ARCH_SUPPORTS_DEBUG_PAGEALLOC &&
(!HIBERNATION || !PPC && !SPARC) && !KMEMCHECK)

Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: batch-free pcp list if possible

free_pcppages_bulk() frees pages from pcp lists in a round-robin fashion
by keeping batch_free counter.  But it doesn't need to spin if there is
only one non-empty list.  This can be checked by batch_free ==
MIGRATE_PCPTYPES.

[akpm@linux-foundation.org: fix comment]
Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
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>

mm: change __remove_from_page_cache()

Now we renamed remove_from_page_cache with delete_from_page_cache.  As
consistency of __remove_from_swap_cache and remove_from_swap_cache, we
change internal page cache handling function name, too.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Christoph Hellwig <hch@infradead.org>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: goodbye remove_from_page_cache()

Now delete_from_page_cache() replaces remove_from_page_cache().  So we
remove remove_from_page_cache so fs or something out of mainline will
notice it when compile time and can fix it.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Christoph Hellwig <hch@infradead.org>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: truncate: change remove_from_page_cache

This patch series changes remove_from_page_cache()'s page ref counting
rule.  Page cache ref count is decreased in delete_from_page_cache().  So
we don't need to decrease the page reference in callers.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Dan Magenheimer <dan.magenheimer@oracle.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Nick Piggin <npiggin@kernel.dk>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: shmem: change remove_from_page_cache

This patch series changes remove_from_page_cache()'s page ref counting
rule.  Page cache ref count is decreased in delete_from_page_cache().  So
we don't need to decrease the page reference in callers.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: hugetlbfs: change remove_from_page_cache

This patch series changes remove_from_page_cache()'s page ref counting
rule.  Page cache ref count is decreased in delete_from_page_cache().  So
we don't need to decrease the page reference in callers.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Cc: William Irwin <wli@holomorphy.com>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: introduce delete_from_page_cache()

Presently we increase the page refcount in add_to_page_cache() but don't
decrease it in remove_from_page_cache().  Such asymmetry adds confusion,
requiring that callers notice it and a comment explaining why they release
a page reference.  It's not a good API.

A long time ago, Hugh tried it (http://lkml.org/lkml/2004/10/24/140) but
gave up because reiser4's drop_page() had to unlock the page between
removing it from page cache and doing the page_cache_release().  But now
the situation is changed.  I think at least things in current mainline
don't have any obstacles.  The problem is for out-of-mainline filesystems
- if they have done such things as reiser4, this patch could be a problem
but they will discover this at compile time since we remove
remove_from_page_cache().

This patch:

This function works as just wrapper remove_from_page_cache().  The
difference is that it decreases page references in itself.  So caller have
to make sure it has a page reference before calling.

This patch is ready for removing remove_from_page_cache().

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Christoph Hellwig <hch@infradead.org>
Acked-by: Hugh Dickins <hughd@google.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Edward Shishkin <edward.shishkin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: add replace_page_cache_page() function

This function basically does:

     remove_from_page_cache(old);
     page_cache_release(old);
     add_to_page_cache_locked(new);

Except it does this atomically, so there's no possibility for the "add" to
fail because of a race.

If memory cgroups are enabled, then the memory cgroup charge is also moved
from the old page to the new.

This function is currently used by fuse to move pages into the page cache
on read, instead of copying the page contents.

[minchan.kim@gmail.com: add freepage() hook to replace_page_cache_page()]
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: allow GUP to fail instead of waiting on a page

GUP user may want to try to acquire a reference to a page if it is already
in memory, but not if IO, to bring it in, is needed.  For example KVM may
tell vcpu to schedule another guest process if current one is trying to
access swapped out page.  Meanwhile, the page will be swapped in and the
guest process, that depends on it, will be able to run again.

This patch adds FAULT_FLAG_RETRY_NOWAIT (suggested by Linus) and
FOLL_NOWAIT follow_page flags.  FAULT_FLAG_RETRY_NOWAIT, when used in
conjunction with VM_FAULT_ALLOW_RETRY, indicates to handle_mm_fault that
it shouldn't drop mmap_sem and wait on a page, but return VM_FAULT_RETRY
instead.

[akpm@linux-foundation.org: improve FOLL_NOWAIT comment]
Signed-off-by: Gleb Natapov <gleb@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Hugh Dickins <hughd@google.com>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: Michel Lespinasse <walken@google.com>
Cc: Avi Kivity <avi@redhat.com>
Cc: Marcelo Tosatti <mtosatti@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: notifier_from_errno() cleanup

While looking at some other notifier callbacks I noticed this code could
use a simple cleanup.

notifier_from_errno() no longer needs the if (ret)/else conditional.  That
same conditional is now done in notifier_from_errno().

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Cc: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Pekka Enberg <penberg@cs.helsinki.fi>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oom: suppress nodes that are not allowed from meminfo on page alloc failure

Displaying extremely verbose meminfo for all nodes on the system is
overkill for page allocation failures when the context restricts that
allocation to only a subset of nodes.  We don't particularly care about
the state of all nodes when some are not allowed in the current context,
they can have an abundance of memory but we can't allocate from that part
of memory.

This patch suppresses disallowed nodes from the meminfo dump on a page
allocation failure if the context requires it.

Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
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>

oom: suppress show_mem() for many nodes in irq context on page alloc failure

When a page allocation failure occurs, show_mem() is called to dump the
state of the VM so users may understand what happened to get into that
condition.

This output, however, can be extremely verbose.  In irq context, it may
result in significant delays that incur NMI watchdog timeouts when the
machine is large (we use CONFIG_NODES_SHIFT > 8 here to define a "large"
machine since the length of the show_mem() output is proportional to the
number of possible nodes).

This patch suppresses the show_mem() call in irq context when the kernel
has CONFIG_NODES_SHIFT > 8.

Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
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>

oom: suppress nodes that are not allowed from meminfo on oom kill

The oom killer is extremely verbose for machines with a large number of
cpus and/or nodes.  This verbosity can often be harmful if it causes other
important messages to be scrolled from the kernel log and incurs a
signicant time delay, specifically for kernels with CONFIG_NODES_SHIFT >
8.

This patch causes only memory information to be displayed for nodes that
are allowed by current's cpuset when dumping the VM state.  Information
for all other nodes is irrelevant to the oom condition; we don't care if
there's an abundance of memory elsewhere if we can't access it.

This only affects the behavior of dumping memory information when an oom
is triggered.  Other dumps, such as for sysrq+m, still display the
unfiltered form when using the existing show_mem() interface.

Additionally, the per-cpu pageset statistics are extremely verbose in oom
killer output, so it is now suppressed.  This removes

nodes_weight(current->mems_allowed) * (1 + nr_cpus)

lines from the oom killer output.

Callers may use __show_mem(SHOW_MEM_FILTER_NODES) to filter disallowed
nodes.

Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
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>

kthread: use kthread_create_on_node()

ksoftirqd, kworker, migration, and pktgend kthreads can be created with
kthread_create_on_node(), to get proper NUMA affinities for their stack and
task_struct.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: David Howells <dhowells@redhat.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kthread: NUMA aware kthread_create_on_node()

All kthreads being created from a single helper task, they all use memory
from a single node for their kernel stack and task struct.

This patch suite creates kthread_create_on_node(), adding a 'cpu' parameter
to parameters already used by kthread_create().

This parameter serves in allocating memory for the new kthread on its
memory node if possible.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: David Howells <dhowells@redhat.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: NUMA aware alloc_thread_info_node()

Add a node parameter to alloc_thread_info(), and change its name to
alloc_thread_info_node()

This change is needed to allow NUMA aware kthread_create_on_cpu()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: David Howells <dhowells@redhat.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: NUMA aware alloc_task_struct_node()

All kthreads being created from a single helper task, they all use memory
from a single node for their kernel stack and task struct.

This patch suite creates kthread_create_on_cpu(), adding a 'cpu' parameter
to parameters already used by kthread_create().

This parameter serves in allocating memory for the new kthread on its
memory node if available.

Users of this new function are : ksoftirqd, kworker, migration, pktgend...

This patch:

Add a node parameter to alloc_task_struct(), and change its name to
alloc_task_struct_node()

This change is needed to allow NUMA aware kthread_create_on_cpu()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: David S. Miller <davem@davemloft.net>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Cc: Tony Luck <tony.luck@intel.com>
Cc: Fenghua Yu <fenghua.yu@intel.com>
Cc: David Howells <dhowells@redhat.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm/compaction: check migrate_pages's return value instead of list_empty()

Many migrate_page's caller check return value instead of list_empy by
cf608ac19c ("mm: compaction: fix COMPACTPAGEFAILED counting").  This patch
makes compaction's migrate_pages consistent with others.  This patch
should not change old behavior.

Signed-off-by: Minchan Kim <minchan.kim@gmail.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: compaction: prevent kswapd compacting memory to reduce CPU usage

This patch reverts 5a03b051 ("thp: use compaction in kswapd for GFP_ATOMIC
order > 0") due to reports stating that kswapd CPU usage was higher and
IRQs were being disabled more frequently.  This was reported at
http://www.spinics.net/linux/fedora/alsa-user/msg09885.html.

Without this patch applied, CPU usage by kswapd hovers around the 20% mark
according to the tester (Arthur Marsh:
http://www.spinics.net/linux/fedora/alsa-user/msg09899.html).  With this
patch applied, it's around 2%.

The problem is not related to THP which specifies __GFP_NO_KSWAPD but is
triggered by high-order allocations hitting the low watermark for their
order and waking kswapd on kernels with CONFIG_COMPACTION set.  The most
common trigger for this is network cards configured for jumbo frames but
it's also possible it'll be triggered by fork-heavy workloads (order-1)
and some wireless cards which depend on order-1 allocations.

The symptoms for the user will be high CPU usage by kswapd in low-memory
situations which could be confused with another writeback problem.  While
a patch like 5a03b051 may be reintroduced in the future, this patch plays
it safe for now and reverts it.

[mel@csn.ul.ie: Beefed up the changelog]
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Mel Gorman <mel@csn.ul.ie>
Reported-by: Arthur Marsh <arthur.marsh@internode.on.net>
Tested-by: Arthur Marsh <arthur.marsh@internode.on.net>
Cc: <stable@kernel.org> [2.6.38.1]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: vmap area cache

Provide a free area cache for the vmalloc virtual address allocator, based
on the algorithm used by the user virtual memory allocator.

This reduces the number of rbtree operations and linear traversals over
the vmap extents in order to find a free area, by starting off at the last
point that a free area was found.

The free area cache is reset if areas are freed behind it, or if we are
searching for a smaller area or alignment than last time.  So allocation
patterns are not changed (verified by corner-case and random test cases in
userspace testing).

This solves a regression caused by lazy vunmap TLB purging introduced in
db64fe02 (mm: rewrite vmap layer).  That patch will leave extents in the
vmap allocator after they are vunmapped, and until a significant number
accumulate that can be flushed in a single batch.  So in a workload that
vmalloc/vfree frequently, a chain of extents will build up from
VMALLOC_START address, which have to be iterated over each time (giving an
O(n) type of behaviour).

After this patch, the search will start from where it left off, giving
closer to an amortized O(1).

This is verified to solve regressions reported Steven in GFS2, and Avi in
KVM.

Hugh's update:

: I tried out the recent mmotm, and on one machine was fortunate to hit
: the BUG_ON(first->va_start < addr) which seems to have been stalling
: your vmap area cache patch ever since May.

: I can get you addresses etc, I did dump a few out; but once I stared
: at them, it was easier just to look at the code: and I cannot see how
: you would be so sure that first->va_start < addr, once you've done
: that addr = ALIGN(max(...), align) above, if align is over 0x1000
: (align was 0x8000 or 0x4000 in the cases I hit: ioremaps like Steve).

: I originally got around it by just changing the
:  if (first->va_start < addr) {
: to
:  while (first->va_start < addr) {
: without thinking about it any further; but that seemed unsatisfactory,
: why would we want to loop here when we've got another very similar
: loop just below it?

: I am never going to admit how long I've spent trying to grasp your
: "while (n)" rbtree loop just above this, the one with the peculiar
:  if (!first && tmp->va_start < addr + size)
: in.  That's unfamiliar to me, I'm guessing it's designed to save a
: subsequent rb_next() in a few circumstances (at risk of then setting
: a wrong cached_hole_size?); but they did appear few to me, and I didn't
: feel I could sign off something with that in when I don't grasp it,
: and it seems responsible for extra code and mistaken BUG_ON below it.

: I've reverted to the familiar rbtree loop that find_vma() does (but
: with va_end >= addr as you had, to respect the additional guard page):
: and then (given that cached_hole_size starts out 0) I don't see the
: need for any complications below it.  If you do want to keep that loop
: as you had it, please add a comment to explain what it's trying to do,
: and where addr is relative to first when you emerge from it.

: Aren't your tests "size <= cached_hole_size" and
: "addr + size > first->va_start" forgetting the guard page we want
: before the next area?  I've changed those.

: I have not changed your many "addr + size - 1 < addr" overflow tests,
: but have since come to wonder, shouldn't they be "addr + size < addr"
: tests - won't the vend checks go wrong if addr + size is 0?

: I have added a few comments - Wolfgang Wander's 2.6.13 description of
: 1363c3cd8603a913a27e2995dccbd70d5312d8e6 Avoiding mmap fragmentation
: helped me a lot, perhaps a pointer to that would be good too.  And I found
: it easier to understand when I renamed cached_start slightly and moved the
: overflow label down.

: This patch would go after your mm-vmap-area-cache.patch in mmotm.
: Trivially, nobody is going to get that BUG_ON with this patch, and it
: appears to work fine on my machines; but I have not given it anything like
: the testing you did on your original, and may have broken all the
: performance you were aiming for.  Please take a look and test it out
: integrate with yours if you're satisfied - thanks.

[akpm@linux-foundation.org: add locking comment]
Signed-off-by: Nick Piggin <npiggin@suse.de>
Signed-off-by: Hugh Dickins <hughd@google.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Reported-and-tested-by: Steven Whitehouse <swhiteho@redhat.com>
Reported-and-tested-by: Avi Kivity <avi@redhat.com>
Tested-by: "Barry J. Marson" <bmarson@redhat.com>
Cc: Prarit Bhargava <prarit@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pwm_backlight: add check_fb() hook

In systems with multiple framebuffer devices, one of the devices might be
blanked while another is unblanked.  In order for the backlight blanking
logic to know whether to turn off the backlight for a particular
framebuffer's blanking notification, it needs to be able to check if a
given framebuffer device corresponds to the backlight.

This plumbs the check_fb hook from core backlight through the
pwm_backlight helper to allow platform code to plug in a check_fb hook.

Signed-off-by: Robert Morell <rmorell@nvidia.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Arun Murthy <arun.murthy@stericsson.com>
Cc: Linus Walleij <linus.walleij@stericsson.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/video/backlight/jornada720_*.c: make needlessly global symbols static

The following symbols are needlessly defined global: jornada_bl_init,
jornada_bl_exit, jornada_lcd_init, jornada_lcd_exit.

Make them static.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked-by: Kristoffer Ericson <kristoffer.ericson@gmail.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

backlight: apple_bl depends on ACPI

apple_bl uses ACPI interfaces (data & code), so it should depend on ACPI.

  drivers/video/backlight/apple_bl.c:142: warning: 'struct acpi_device' declared inside parameter list
  drivers/video/backlight/apple_bl.c:142: warning: its scope is only this definition or declaration, which is probably not what you want
  drivers/video/backlight/apple_bl.c:201: warning: 'struct acpi_device' declared inside parameter list
  drivers/video/backlight/apple_bl.c:215: error: variable 'apple_bl_driver' has initializer but incomplete type
  drivers/video/backlight/apple_bl.c:216: error: unknown field 'name' specified in initializer
  ...

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Matthew Garrett <mjg@redhat.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mbp_nvidia_bl: rename to apple_bl

It works on hardware other than Macbook Pros, and it works on GPUs other
than Nvidia.  It should even work on iMacs, so change the name to match
reality more precisely and include an alias so existing users don't get
confused.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Acked-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Cc: Mourad De Clerck <mourad@aquazul.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mbp_nvidia_bl: check that the backlight control functions

The SMI-based backlight control functionality may fail to work if the
system is running under EFI rather than BIOS.  Check that the hardware
responds as expected, and exit if it doesn't.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Acked-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Cc: Mourad De Clerck <mourad@aquazul.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mbp_nvidia_bl: remove DMI dependency

This driver only has to deal with two different classes of hardware, but
right now it needs new DMI entries for every new machine. It turns out
that there's an ACPI device that uniquely identifies Apples with backlights,
so this patch reworks the driver into an ACPI one, identifies the hardware
by checking the PCI vendor of the root bridge and strips out all the DMI
code. It also changes the config text to clarify that it works on devices
other than Macbook Pros and GPUs other than nvidia.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Acked-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Cc: Mourad De Clerck <mourad@aquazul.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

acpi: tie ACPI backlight devices to PCI devices if possible

Dual-GPU machines may provide more than one ACPI backlight interface.  Tie
the backlight device to the GPU in order to allow userspace to identify
the correct interface.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Airlie <airlied@linux.ie>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Sedat Dilek <sedat.dilek@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

nouveau: change the backlight parent device to the connector, not the PCI dev

We may eventually end up with per-connector backlights, especially with
ddcci devices.  Make sure that the parent node for the backlight device is
the connector rather than the PCI device.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Airlie <airlied@linux.ie>
Cc: Alex Deucher <alexdeucher@gmail.com>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Sedat Dilek <sedat.dilek@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

radeon: expose backlight class device for legacy LVDS encoder

Allows e.g. power management daemons to control the backlight level. Inspired
by the corresponding code in radeonfb.

[mjg@redhat.com: updated to add backlight type and make the connector the parent device]
Signed-off-by: Michel Dänzer <michel@daenzer.net>
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Airlie <airlied@linux.ie>
Acked-by: Alex Deucher <alexdeucher@gmail.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Sedat Dilek <sedat.dilek@googlemail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

backlight: add backlight type

There may be multiple ways of controlling the backlight on a given
machine.  Allow drivers to expose the type of interface they are
providing, making it possible for userspace to make appropriate policy
decisions.

Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Airlie <airlied@linux.ie>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/leds/leds-lp5523.c: world-writable engine* sysfs files

Don't allow everybody to change LED settings.

Signed-off-by: Vasiliy Kulikov <segoon@openwall.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/leds/leds-lp5521.c: world-writable sysfs engine* files

Don't allow everybody to change LED settings.

Signed-off-by: Vasiliy Kulikov <segoon@openwall.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/vidfeo/backlight: ld9040 amoled driver support

Add a ld9040 amoled panel driver.

Signed-off-by: Donghwa Lee <dh09.lee@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Inki Dae <inki.dae@samsung.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

leds: make *struct gpio_led_platform_data.leds const

And fix a typo.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: Lars-Peter Clausen <lars@metafoo.de>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

leds: add driver for LM3530 ALS

Simple backlight driver for National Semiconductor LM3530.  Presently only
manual mode is supported, PWM and ALS support to be added.

Signed-off-by: Shreshtha Kumar Sahu <shreshthakumar.sahu@stericsson.com>
Cc: Linus Walleij <linus.walleij@stericsson.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

leds: convert bd2802 driver to dev_pm_ops

There is a move to deprecate bus-specific PM operations and move to using
dev_pm_ops instead in order to reduce the amount of boilerplate code in
buses and facilitiate updates to the PM core.  Do this move for the bs2802
driver.

[akpm@linux-foundation.org: fix warnings]
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Cc: Kim Kyuwon <q1.kim@samsung.com>
Cc: Kim Kyuwon <chammoru@gmail.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cgroups: if you list_empty() a head then don't list_del() it

list_del() leaves poison in the prev and next pointers.  The next
list_empty() will compare those poisons, and say the list isn't empty.
Any list operations that assume the node is on a list because of such a
check will be fooled into dereferencing poison.  One needs to INIT the
node after the del, and fortunately there's already a wrapper for that -
list_del_init().

Some of the dels are followed by deallocations, so can be ignored, and one
can be merged with an add to make a move.  Apart from that, I erred on the
side of caution in making nodes list_empty()-queriable.

Signed-off-by: Phil Carmody <ext-phil.2.carmody@nokia.com>
Reviewed-by: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oom: avoid deferring oom killer if exiting task is being traced

The oom killer naturally defers killing anything if it finds an eligible
task that is already exiting and has yet to detach its ->mm.  This avoids
unnecessarily killing tasks when one is already in the exit path and may
free enough memory that the oom killer is no longer needed.  This is
detected by PF_EXITING since threads that have already detached its ->mm
are no longer considered at all.

The problem with always deferring when a thread is PF_EXITING, however, is
that it may never actually exit when being traced, specifically if another
task is tracing it with PTRACE_O_TRACEEXIT.  The oom killer does not want
to defer in this case since there is no guarantee that thread will ever
exit without intervention.

This patch will now only defer the oom killer when a thread is PF_EXITING
and no ptracer has stopped its progress in the exit path.  It also ensures
that a child is sacrificed for the chosen parent only if it has a
different ->mm as the comment implies: this ensures that the thread group
leader is always targeted appropriately.

Signed-off-by: David Rientjes <rientjes@google.com>
Reported-by: Oleg Nesterov <oleg@redhat.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: <stable@kernel.org> [2.6.38.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oom: skip zombies when iterating tasklist

We shouldn't defer oom killing if a thread has already detached its ->mm
and still has TIF_MEMDIE set.  Memory needs to be freed, so find kill
other threads that pin the same ->mm or find another task to kill.

Signed-off-by: Andrey Vagin <avagin@openvz.org>
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: <stable@kernel.org> [2.6.38.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oom: prevent unnecessary oom kills or kernel panics

This patch prevents unnecessary oom kills or kernel panics by reverting
two commits:

495789a5 (oom: make oom_score to per-process value)
cef1d352 (oom: multi threaded process coredump don't make deadlock)

First, 495789a5 (oom: make oom_score to per-process value) ignores the
fact that all threads in a thread group do not necessarily exit at the
same time.

It is imperative that select_bad_process() detect threads that are in the
exit path, specifically those with PF_EXITING set, to prevent needlessly
killing additional tasks.  If a process is oom killed and the thread group
leader exits, select_bad_process() cannot detect the other threads that
are PF_EXITING by iterating over only processes.  Thus, it currently
chooses another task unnecessarily for oom kill or panics the machine when
nothing else is eligible.

By iterating over threads instead, it is possible to detect threads that
are exiting and nominate them for oom kill so they get access to memory
reserves.

Second, cef1d352 (oom: multi threaded process coredump don't make
deadlock) erroneously avoids making the oom killer a no-op when an
eligible thread other than current isfound to be exiting.  We want to
detect this situation so that we may allow that exiting thread time to
exit and free its memory; if it is able to exit on its own, that should
free memory so current is no loner oom.  If it is not able to exit on its
own, the oom killer will nominate it for oom kill which, in this case,
only means it will get access to memory reserves.

Without this change, it is easy for the oom killer to unnecessarily target
tasks when all threads of a victim don't exit before the thread group
leader or, in the worst case, panic the machine.

Signed-off-by: David Rientjes <rientjes@google.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Andrey Vagin <avagin@openvz.org>
Cc: <stable@kernel.org> [2.6.38.x]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mm: swap: unlock swapfile inode mutex before closing file on bad swapfiles

If an administrator tries to swapon a file backed by NFS, the inode mutex is
taken (as it is for any swapfile) but later identified to be a bad swapfile
due to the lack of bmap and tries to cleanup. During cleanup, an attempt is
made to close the file but with inode->i_mutex still held. Closing an NFS
file syncs it which tries to acquire the inode mutex leading to deadlock. If
lockdep is enabled the following appears on the console;

    =============================================
    [ INFO: possible recursive locking detected ]
    2.6.38-rc8-autobuild #1
    ---------------------------------------------
    swapon/2192 is trying to acquire lock:
     (&sb->s_type->i_mutex_key#13){+.+.+.}, at: vfs_fsync_range+0x47/0x7c

    but task is already holding lock:
     (&sb->s_type->i_mutex_key#13){+.+.+.}, at: sys_swapon+0x28d/0xae7

    other info that might help us debug this:
    1 lock held by swapon/2192:
     #0:  (&sb->s_type->i_mutex_key#13){+.+.+.}, at: sys_swapon+0x28d/0xae7

    stack backtrace:
    Pid: 2192, comm: swapon Not tainted 2.6.38-rc8-autobuild #1
    Call Trace:
        __lock_acquire+0x2eb/0x1623
        find_get_pages_tag+0x14a/0x174
        pagevec_lookup_tag+0x25/0x2e
        vfs_fsync_range+0x47/0x7c
        lock_acquire+0xd3/0x100
        vfs_fsync_range+0x47/0x7c
        nfs_flush_one+0x0/0xdf [nfs]
        mutex_lock_nested+0x40/0x2b1
        vfs_fsync_range+0x47/0x7c
        vfs_fsync_range+0x47/0x7c
        vfs_fsync+0x1c/0x1e
        nfs_file_flush+0x64/0x69 [nfs]
        filp_close+0x43/0x72
        sys_swapon+0xa39/0xae7
        sysret_check+0x2e/0x69
        system_call_fastpath+0x16/0x1b

This patch releases the mutex if its held before calling filep_close()
so swapon fails as expected without deadlock when the swapfile is backed
by NFS.  If accepted for 2.6.39, it should also be considered a -stable
candidate for 2.6.38 and 2.6.37.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: <stable@kernel.org> [2.6.37+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/asm-generic/unistd.h: fix syncfs syscall number

syncfs() is duplicating name_to_handle_at() due to a merging mistake.

Cc: Sage Weil <sage@newdream.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge branch 'slab/urgent' of git://git./linux/kernel/git/penberg/slab-2.6

* 'slab/urgent' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/slab-2.6:
  slub: Add statistics for this_cmpxchg_double failures
  slub: Add missing irq restore for the OOM path

Merge branch 'for-linus' of git://git./linux/kernel/git/ericvh/v9fs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs:
  [net/9p]: Introduce basic flow-control for VirtIO transport.
  9p: use the updated offset given by generic_write_checks
  [net/9p] Don't re-pin pages on retrying virtqueue_add_buf().
  [net/9p] Set the condition just before waking up.
  [net/9p] unconditional wake_up to proc waiting for space on VirtIO ring
  fs/9p: Add v9fs_dentry2v9ses
  fs/9p: Attach writeback_fid on first open with WR flag
  fs/9p: Open writeback fid in O_SYNC mode
  fs/9p: Use truncate_setsize instead of vmtruncate
  net/9p: Fix compile warning
  net/9p: Convert the in the 9p rpc call path to GFP_NOFS
  fs/9p: Fix race in initializing writeback fid

Merge git://git./linux/kernel/git/sage/ceph-client

* git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: use watch/notify for changes in rbd header
  libceph: add lingering request and watch/notify event framework
  rbd: update email address in Documentation
  ceph: rename dentry_release -> d_release, fix comment
  ceph: add request to the tail of unsafe write list
  ceph: remove request from unsafe list if it is canceled/timed out
  ceph: move readahead default to fs/ceph from libceph
  ceph: add ino32 mount option
  ceph: update common header files
  ceph: remove debugfs debug cruft
  libceph: fix osd request queuing on osdmap updates
  ceph: preserve I_COMPLETE across rename
  libceph: Fix base64-decoding when input ends in newline.

tty: stop using "delayed_work" in the tty layer

Using delayed-work for tty flip buffers ends up causing us to wait for
the next tick to complete some actions.  That's usually not all that
noticeable, but for certain latency-critical workloads it ends up being
totally unacceptable.

As an extreme case of this, passing a token back-and-forth over a pty
will take two ticks per iteration, so even just a thousand iterations
will take 8 seconds assuming a common 250Hz configuration.

Avoiding the whole delayed work issue brings that ping-pong test-case
down to 0.009s on my machine.

In more practical terms, this latency has been a performance problem for
things like dive computer simulators (simulating the serial interface
using the ptys) and for other environments (Alan mentions a CP/M emulator).

Reported-by: Jef Driesen <jefdriesen@telenet.be>
Acked-by: Greg KH <gregkh@suse.de>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

[net/9p]: Introduce basic flow-control for VirtIO transport.

Recent zerocopy work in the 9P VirtIO transport maps and pins
user buffers into kernel memory for the server to work on them.
Since the user process can initiate this kind of pinning with a simple
read/write call, thousands of IO threads initiated by the user process can
hog the system resources and could result into denial of service.

This patch introduces flow control to avoid that extreme scenario.

The ceiling limit to avoid denial of service attacks is set to relatively
high (nr_free_pagecache_pages()/4) so that it won't interfere with
regular usage, but can step in extreme cases to limit the total system
hang. Since we don't have a global structure to accommodate this variable,
I choose the virtio_chan as the home for this.

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Reviewed-by: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: use the updated offset given by generic_write_checks

Without this fix, even if a file is opened in O_APPEND mode, data will be
written at current file position instead of end of file.

Signed-off-by: M. Mohan Kumar <mohan@in.ibm.com>
Reviewed-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

[net/9p] Don't re-pin pages on retrying virtqueue_add_buf().

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

[net/9p] Set the condition just before waking up.

Given that the sprious wake-ups are common, we need to move the
condition setting right next to the wake_up().  After setting the condition
to req->status = REQ_STATUS_RCVD, sprious wakeups may cause the
virtqueue back on the free list for someone else to use.
This may result in kernel panic while relasing the pinned pages
in p9_release_req_pages().

Also rearranged the while loop in req_done() for better redability.

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

[net/9p] unconditional wake_up to proc waiting for space on VirtIO ring

Process may wait to get space on VirtIO ring to send a transaction to
VirtFS server. Current code just does a conditional wake_up() which
means only one process will be woken up even if multiple processes
are waiting.

This fix makes the wake_up unconditional. Hence we won't have any
processes waiting for-ever.

Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

fs/9p: Add v9fs_dentry2v9ses

Add the new static inline and use the same

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

fs/9p: Attach writeback_fid on first open with WR flag

We don't need writeback fid if we are only doing O_RDONLY open

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

fs/9p: Open writeback fid in O_SYNC mode

Older version of protocol don't support tsyncfs operation.
So for them force a O_SYNC flag on the server

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

fs/9p: Use truncate_setsize instead of vmtruncate

convert vmtruncate usage to truncate_setsize. We also writeback
all dirty pages before doing 9p operations and on success call truncate_setsize.
This ensure that we continue sanely on failed truncate on the server. The
disadvantage is that we are now going to write back the content that get
thrown away later as a part of truncate.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

net/9p: Fix compile warning

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

net/9p: Convert the in the 9p rpc call path to GFP_NOFS

Without this we can cause reclaim allocation in writepage.

[ 3433.448430] =================================
[ 3433.449117] [ INFO: inconsistent lock state ]
[ 3433.449117] 2.6.38-rc5+ #84
[ 3433.449117] ---------------------------------
[ 3433.449117] inconsistent {RECLAIM_FS-ON-W} -> {IN-RECLAIM_FS-R} usage.
[ 3433.449117] kswapd0/505 [HC0[0]:SC0[0]:HE1:SE1] takes:
[ 3433.449117]  (iprune_sem){+++++-}, at: [<ffffffff810ebbab>] shrink_icache_memory+0x45/0x2b1
[ 3433.449117] {RECLAIM_FS-ON-W} state was registered at:
[ 3433.449117]   [<ffffffff8107fe5f>] mark_held_locks+0x52/0x70
[ 3433.449117]   [<ffffffff8107ff02>] lockdep_trace_alloc+0x85/0x9f
[ 3433.449117]   [<ffffffff810d353d>] slab_pre_alloc_hook+0x18/0x3c
[ 3433.449117]   [<ffffffff810d3fd5>] kmem_cache_alloc+0x23/0xa2
[ 3433.449117]   [<ffffffff8127be77>] idr_pre_get+0x2d/0x6f
[ 3433.449117]   [<ffffffff815434eb>] p9_idpool_get+0x30/0xae
[ 3433.449117]   [<ffffffff81540123>] p9_client_rpc+0xd7/0x9b0
[ 3433.449117]   [<ffffffff815427b0>] p9_client_clunk+0x88/0xdb
[ 3433.449117]   [<ffffffff811d56e5>] v9fs_evict_inode+0x3c/0x48
[ 3433.449117]   [<ffffffff810eb511>] evict+0x1f/0x87
[ 3433.449117]   [<ffffffff810eb5c0>] dispose_list+0x47/0xe3
[ 3433.449117]   [<ffffffff810eb8da>] evict_inodes+0x138/0x14f
[ 3433.449117]   [<ffffffff810d90e2>] generic_shutdown_super+0x57/0xe8
[ 3433.449117]   [<ffffffff810d91e8>] kill_anon_super+0x11/0x50
[ 3433.449117]   [<ffffffff811d4951>] v9fs_kill_super+0x49/0xab
[ 3433.449117]   [<ffffffff810d926e>] deactivate_locked_super+0x21/0x46
[ 3433.449117]   [<ffffffff810d9e84>] deactivate_super+0x40/0x44
[ 3433.449117]   [<ffffffff810ef848>] mntput_no_expire+0x100/0x109
[ 3433.449117]   [<ffffffff810f0aeb>] sys_umount+0x2f1/0x31c
[ 3433.449117]   [<ffffffff8102c87b>] system_call_fastpath+0x16/0x1b
[ 3433.449117] irq event stamp: 192941
[ 3433.449117] hardirqs last  enabled at (192941): [<ffffffff81568dcf>] _raw_spin_unlock_irq+0x2b/0x30
[ 3433.449117] hardirqs last disabled at (192940): [<ffffffff810b5f97>] shrink_inactive_list+0x290/0x2f5
[ 3433.449117] softirqs last  enabled at (188470): [<ffffffff8105fd65>] __do_softirq+0x133/0x152
[ 3433.449117] softirqs last disabled at (188455): [<ffffffff8102d7cc>] call_softirq+0x1c/0x28
[ 3433.449117]
[ 3433.449117] other info that might help us debug this:
[ 3433.449117] 1 lock held by kswapd0/505:
[ 3433.449117]  #0:  (shrinker_rwsem){++++..}, at: [<ffffffff810b52e2>] shrink_slab+0x38/0x15f
[ 3433.449117]
[ 3433.449117] stack backtrace:
[ 3433.449117] Pid: 505, comm: kswapd0 Not tainted 2.6.38-rc5+ #84
[ 3433.449117] Call Trace:
[ 3433.449117]  [<ffffffff8107fbce>] ? valid_state+0x17e/0x191
[ 3433.449117]  [<ffffffff81036896>] ? save_stack_trace+0x28/0x45
[ 3433.449117]  [<ffffffff81080426>] ? check_usage_forwards+0x0/0x87
[ 3433.449117]  [<ffffffff8107fcf4>] ? mark_lock+0x113/0x22c
[ 3433.449117]  [<ffffffff8108105f>] ? __lock_acquire+0x37a/0xcf7
[ 3433.449117]  [<ffffffff8107fc0e>] ? mark_lock+0x2d/0x22c
[ 3433.449117]  [<ffffffff81081077>] ? __lock_acquire+0x392/0xcf7
[ 3433.449117]  [<ffffffff810b14d2>] ? determine_dirtyable_memory+0x15/0x28
[ 3433.449117]  [<ffffffff81081a33>] ? lock_acquire+0x57/0x6d
[ 3433.449117]  [<ffffffff810ebbab>] ? shrink_icache_memory+0x45/0x2b1
[ 3433.449117]  [<ffffffff81567d85>] ? down_read+0x47/0x5c
[ 3433.449117]  [<ffffffff810ebbab>] ? shrink_icache_memory+0x45/0x2b1
[ 3433.449117]  [<ffffffff810ebbab>] ? shrink_icache_memory+0x45/0x2b1
[ 3433.449117]  [<ffffffff810b5385>] ? shrink_slab+0xdb/0x15f
[ 3433.449117]  [<ffffffff810b69bc>] ? kswapd+0x574/0x96a
[ 3433.449117]  [<ffffffff810b6448>] ? kswapd+0x0/0x96a
[ 3433.449117]  [<ffffffff810714e2>] ? kthread+0x7d/0x85
[ 3433.449117]  [<ffffffff8102d6d4>] ? kernel_thread_helper+0x4/0x10
[ 3433.449117]  [<ffffffff81569200>] ? restore_args+0x0/0x30
[ 3433.449117]  [<ffffffff81071465>] ? kthread+0x0/0x85
[ 3433.449117]  [<ffffffff8102d6d0>] ? kernel_thread_helper+0x0/0x10

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

fs/9p: Fix race in initializing writeback fid

When two process open the same file we can end up with both of them
allocating the writeback_fid. Add a new mutex which can be used
for synchronizing v9fs_inode member values.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Signed-off-by: Venkateswararao Jujjuri <jvrao@linux.vnet.ibm.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

slub: Add statistics for this_cmpxchg_double failures

Add some statistics for debugging.

Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>

slub: Add missing irq restore for the OOM path

OOM path is missing the irq restore in the CONFIG_CMPXCHG_LOCAL case.

Signed-off-by: Christoph Lameter <cl@linux.com>
Signed-off-by: Pekka Enberg <penberg@kernel.org>

rbd: use watch/notify for changes in rbd header

Send notifications when we change the rbd header (e.g. create a snapshot)
and wait for such notifications.  This allows synchronizing the snapshot
creation between different rbd clients/rools.

Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>

libceph: add lingering request and watch/notify event framework

Lingering requests are requests that are sent to the OSD normally but
tracked also after we get a successful request.  This keeps the OSD
connection open and resends the original request if the object moves to
another OSD.  The OSD can then send notification messages back to us
if another client initiates a notify.

This framework will be used by RBD so that the client gets notification
when a snapshot is created by another node or tool.

Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>

Merge branch 'for-linus' of git://git./linux/kernel/git/mszeredi/fuse

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
  fuse: make fuse_dentry_revalidate() RCU aware
  fuse: make fuse_permission() RCU aware
  fuse: wakeup pollers on connection release/abort
  fuse: reduce size of struct fuse_request

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

* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  xen: update mask_rw_pte after kernel page tables init changes
  xen: set max_pfn_mapped to the last pfn mapped
  x86: Cleanup highmap after brk is concluded

Fix up trivial onflict (added header file includes) in
arch/x86/mm/init_64.c

Merge branch 'next-samsung' of git://git.fluff.org/bjdooks/linux

* 'next-samsung' of git://git.fluff.org/bjdooks/linux:
  ARM: H1940/RX1950: Change default LED triggers
  ARM: S3C2442: RX1950: Add support for LED blinking
  ARM: S3C2442: RX1950: Retain LEDs state in suspend
  ARM: S3C2410: H1940: Fix lcd_power_set function
  ARM: S3C2410: H1940: Add battery support
  ARM: S3C2410: H1940: Use leds-gpio driver for LEDs managing
  ARM: S3C2410: H1940: Make h1940-bluetooth.c compile again
  ARM: S3C2410: H1940: Add keys device

Merge branch 'for-linus/2639/i2c-2' of git://git.fluff.org/bjdooks/linux

* 'for-linus/2639/i2c-2' of git://git.fluff.org/bjdooks/linux:
  i2c-pxa2xx: Don't clear isr bits too early
  i2c-pxa2xx: Fix register offsets
  i2c-pxa2xx: pass of_node from platform driver to adapter and publish
  i2c-pxa2xx: check timeout correctly
  i2c-pxa2xx: add support for shared IRQ handler
  i2c-pxa2xx: Add PCI support for PXA I2C controller
  ARM: pxa2xx: reorganize I2C files
  i2c-pxa2xx: use dynamic register layout
  i2c-mxs: set controller to pio queue mode after reset
  i2c-eg20t: support new device OKI SEMICONDUCTOR ML7213 IOH
  i2c/busses: Add support for Diolan U2C-12 USB-I2C adapter

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

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/penberg/slab-2.6:
  slub: Dont define useless label in the !CONFIG_CMPXCHG_LOCAL case
  slab,rcu: don't assume the size of struct rcu_head
  slub,rcu: don't assume the size of struct rcu_head
  slub: automatically reserve bytes at the end of slab
  Lockless (and preemptless) fastpaths for slub
  slub: Get rid of slab_free_hook_irq()
  slub: min_partial needs to be in first cacheline
  slub: fix ksize() build error
  slub: fix kmemcheck calls to match ksize() hints
  Revert "slab: Fix missing DEBUG_SLAB last user"
  mm: Remove support for kmem_cache_name()

sd: Fail discard requests when logical block provisioning has been disabled

Ensure that we kill discard requests after logical block provisioning
has been disabled in sysfs.

Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge git://git./linux/kernel/git/davem/net-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (33 commits)
  IPVS: Use global mutex in ip_vs_app.c
  ipvs: fix a typo in __ip_vs_control_init()
  veth: Fix the byte counters
  net ipv6: Fix duplicate /proc/sys/net/ipv6/neigh directory entries.
  macvlan: Fix use after free of struct macvlan_port.
  net: fix incorrect spelling in drop monitor protocol
  can: c_can: Do basic c_can configuration _before_ enabling the interrupts
  net/appletalk: fix atalk_release use after free
  ipx: fix ipx_release()
  snmp: SNMP_UPD_PO_STATS_BH() always called from softirq
  l2tp: fix possible oops on l2tp_eth module unload
  xfrm: Fix initialize repl field of struct xfrm_state
  netfilter: ipt_CLUSTERIP: fix buffer overflow
  netfilter: xtables: fix reentrancy
  netfilter: ipset: fix checking the type revision at create command
  netfilter: ipset: fix address ranges at hash:*port* types
  niu: Rename NIU parent platform device name to fix conflict.
  r8169: fix a bug in rtl8169_init_phy()
  bonding: fix a typo in a comment
  ftmac100: use resource_size()
  ...

IPVS: Use global mutex in ip_vs_app.c

As part of the work to make IPVS network namespace aware
__ip_vs_app_mutex was replaced by a per-namespace lock,
ipvs->app_mutex. ipvs->app_key is also supplied for debugging purposes.

Unfortunately this implementation results in ipvs->app_key residing
in non-static storage which at the very least causes a lockdep warning.

This patch takes the rather heavy-handed approach of reinstating
__ip_vs_app_mutex which will cover access to the ipvs->list_head
of all network namespaces.

[   12.610000] IPVS: Creating netns size=2456 id=0
[   12.630000] IPVS: Registered protocols (TCP, UDP, SCTP, AH, ESP)
[   12.640000] BUG: key ffff880003bbf1a0 not in .data!
[   12.640000] ------------[ cut here ]------------
[   12.640000] WARNING: at kernel/lockdep.c:2701 lockdep_init_map+0x37b/0x570()
[   12.640000] Hardware name: Bochs
[   12.640000] Pid: 1, comm: swapper Tainted: G        W 2.6.38-kexec-06330-g69b7efe-dirty #122
[   12.650000] Call Trace:
[   12.650000]  [<ffffffff8102e685>] warn_slowpath_common+0x75/0xb0
[   12.650000]  [<ffffffff8102e6d5>] warn_slowpath_null+0x15/0x20
[   12.650000]  [<ffffffff8105967b>] lockdep_init_map+0x37b/0x570
[   12.650000]  [<ffffffff8105829d>] ? trace_hardirqs_on+0xd/0x10
[   12.650000]  [<ffffffff81055ad8>] debug_mutex_init+0x38/0x50
[   12.650000]  [<ffffffff8104bc4c>] __mutex_init+0x5c/0x70
[   12.650000]  [<ffffffff81685ee7>] __ip_vs_app_init+0x64/0x86
[   12.660000]  [<ffffffff81685a3b>] ? ip_vs_init+0x0/0xff
[   12.660000]  [<ffffffff811b1c33>] T.620+0x43/0x170
[   12.660000]  [<ffffffff811b1e9a>] ? register_pernet_subsys+0x1a/0x40
[   12.660000]  [<ffffffff81685a3b>] ? ip_vs_init+0x0/0xff
[   12.660000]  [<ffffffff81685a3b>] ? ip_vs_init+0x0/0xff
[   12.660000]  [<ffffffff811b1db7>] register_pernet_operations+0x57/0xb0
[   12.660000]  [<ffffffff81685a3b>] ? ip_vs_init+0x0/0xff
[   12.670000]  [<ffffffff811b1ea9>] register_pernet_subsys+0x29/0x40
[   12.670000]  [<ffffffff81685f19>] ip_vs_app_init+0x10/0x12
[   12.670000]  [<ffffffff81685a87>] ip_vs_init+0x4c/0xff
[   12.670000]  [<ffffffff8166562c>] do_one_initcall+0x7a/0x12e
[   12.670000]  [<ffffffff8166583e>] kernel_init+0x13e/0x1c2
[   12.670000]  [<ffffffff8128c134>] kernel_thread_helper+0x4/0x10
[   12.670000]  [<ffffffff8128ad40>] ? restore_args+0x0/0x30
[   12.680000]  [<ffffffff81665700>] ? kernel_init+0x0/0x1c2
[   12.680000]  [<ffffffff8128c130>] ? kernel_thread_helper+0x0/0x1global0

Signed-off-by: Simon Horman <horms@verge.net.au>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Julian Anastasov <ja@ssi.bg>
Cc: Hans Schillstrom <hans@schillstrom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipvs: fix a typo in __ip_vs_control_init()

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Cc: Simon Horman <horms@verge.net.au>
Cc: Julian Anastasov <ja@ssi.bg>
Acked-by: Simon Horman <horms@verge.net.au>
Signed-off-by: David S. Miller <davem@davemloft.net>

veth: Fix the byte counters

Commit 44540960 "veth: move loopback logic to common location" introduced
a bug in the packet counters.  I don't understand why that happened as it
is not explained in the comments and the mut check in dev_forward_skb
retains the assumption that skb->len is the total length of the packet.

I just measured this emperically by setting up a veth pair between two
noop network namespaces setting and attempting a telnet connection between
the two.  I saw three packets in each direction and the byte counters were
exactly 14*3 = 42 bytes high in each direction.  I got the actual
packet lengths with tcpdump.

So remove the extra ETH_HLEN from the veth byte count totals.

Signed-off-by: Eric W. Biederman <ebiederm@aristanetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net ipv6: Fix duplicate /proc/sys/net/ipv6/neigh directory entries.

When I was fixing issues with unregisgtering tables under /proc/sys/net/ipv6/neigh
by adding a mount point it appears I missed a critical ordering issue, in the
ipv6 initialization.  I had not realized that ipv6_sysctl_register is called
at the very end of the ipv6 initialization and in particular after we call
neigh_sysctl_register from ndisc_init.

"neigh" needs to be initialized in ipv6_static_sysctl_register which is
the first ipv6 table to initialized, and definitely before ndisc_init.
This removes the weirdness of duplicate tables while still providing a
"neigh" mount point which prevents races in sysctl unregistering.

This was initially reported at https://bugzilla.kernel.org/show_bug.cgi?id=31232
Reported-by: sunkan@zappa.cx
Signed-off-by: Eric W. Biederman <ebiederm@aristanetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

macvlan: Fix use after free of struct macvlan_port.

When the macvlan driver was extended to call unregisgter_netdevice_queue
in 23289a37e2b127dfc4de1313fba15bb4c9f0cd5b, a use after free of struct
macvlan_port was introduced.  The code in dellink relied on unregister_netdevice
actually unregistering the net device so it would be safe to free macvlan_port.

Since unregister_netdevice_queue can just queue up the unregister instead of
performing the unregiser immediately we free the macvlan_port too soon and
then the code in macvlan_stop removes the macaddress for the set of macaddress
to listen for and uses memory that has already been freed.

To fix this add a reference count to track when it is safe to free the macvlan_port
and move the call of macvlan_port_destroy into macvlan_uninit which is guaranteed
to be called after the final macvlan_port_close.

Signed-off-by: Eric W. Biederman <ebiederm@aristanetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix incorrect spelling in drop monitor protocol

It was pointed out to me recently that my spelling could be better :)

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

can: c_can: Do basic c_can configuration _before_ enabling the interrupts

I ran into some trouble while testing the SocketCAN driver for the BOSCH
C_CAN controller. The interface is not correctly initialized, if I put
some CAN traffic on the line, _while_ the interface is being started
(which means: the interface doesn't come up correcty, if there's some RX
traffic while doing 'ifconfig can0 up').

The current implementation enables the controller interrupts _before_
doing the basic c_can configuration. I think, this should be done the
other way round.

The patch below fixes things for me.

Signed-off-by: Jan Altenberg <jan@linutronix.de>
Acked-by: Kurt Van Dijck <kurt.van.dijck@eia.be>
Acked-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/appletalk: fix atalk_release use after free

The BKL removal in appletalk introduced a use-after-free problem,
where atalk_destroy_socket frees a sock, but we still release
the socket lock on it.

An easy fix is to take an extra reference on the sock and sock_put
it when returning from atalk_release.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipx: fix ipx_release()

Commit b0d0d915d1d1a0 (remove the BKL) added a regression, because
sock_put() can free memory while we are going to use it later.

Fix is to delay sock_put() _after_ release_sock().

Reported-by: Ingo Molnar <mingo@elte.hu>
Tested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: David S. Miller <davem@davemloft.net>

snmp: SNMP_UPD_PO_STATS_BH() always called from softirq

We dont need to test if we run from softirq context, we definitely are.

This saves few instructions in ip_rcv() & ip_rcv_finish()

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>