openwrt/staging/blogic.git
12 years agomemcg: replace mem and mem_cont stragglers
Hugh Dickins [Wed, 21 Mar 2012 23:34:18 +0000 (16:34 -0700)]
memcg: replace mem and mem_cont stragglers

Replace mem and mem_cont stragglers in memcontrol.c by memcg.

Signed-off-by: Hugh Dickins <hughd@google.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomemcg: replace MEM_CONT by MEM_RES_CTLR
Hugh Dickins [Wed, 21 Mar 2012 23:34:18 +0000 (16:34 -0700)]
memcg: replace MEM_CONT by MEM_RES_CTLR

Correct an #endif comment in memcontrol.h from MEM_CONT to MEM_RES_CTLR.

Signed-off-by: Hugh Dickins <hughd@google.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoswap: don't do discard if no discard option added
Shaohua Li [Wed, 21 Mar 2012 23:34:17 +0000 (16:34 -0700)]
swap: don't do discard if no discard option added

When swapon() was not passed the SWAP_FLAG_DISCARD option, sys_swapon()
will still perform a discard operation.  This can cause problems if
discard is slow or buggy.

Reverse the order of the check so that a discard operation is performed
only if the sys_swapon() caller is attempting to enable discard.

Signed-off-by: Shaohua Li <shli@fusionio.com>
Reported-by: Holger Kiehl <Holger.Kiehl@dwd.de>
Tested-by: Holger Kiehl <Holger.Kiehl@dwd.de>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: forbid lumpy-reclaim in shrink_active_list()
Konstantin Khlebnikov [Wed, 21 Mar 2012 23:34:17 +0000 (16:34 -0700)]
mm: forbid lumpy-reclaim in shrink_active_list()

Reset the reclaim mode in shrink_active_list() to RECLAIM_MODE_SINGLE |
RECLAIM_MODE_ASYNC.  (sync/async sign is used only in shrink_page_list
and does not affect shrink_active_list)

Currenly shrink_active_list() sometimes works in lumpy-reclaim mode, if
RECLAIM_MODE_LUMPYRECLAIM is left over from an earlier
shrink_inactive_list().  Meanwhile, in age_active_anon()
sc->reclaim_mode is totally zero.  So the current behavior is too
complex and confusing, and this looks like bug.

In general, shrink_active_list() populates the inactive list for the
next shrink_inactive_list().  Lumpy shring_inactive_list() isolates
pages around the chosen one from both the active and inactive lists.
So, there is no reason for lumpy isolation in shrink_active_list().

See also: https://lkml.org/lkml/2012/3/15/583

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Proposed-by: Hugh Dickins <hughd@google.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agommap.c: fix comment for __insert_vm_struct()
Kautuk Consul [Wed, 21 Mar 2012 23:34:16 +0000 (16:34 -0700)]
mmap.c: fix comment for __insert_vm_struct()

The comment above __insert_vm_struct seems to suggest that this function
is also going to link the VMA with the anon_vma, but this is not true.
This function only links the VMA to the mm->mm_rb tree and the mm->mmap
linked list.

[akpm@linux-foundation.org: improve comment layout and text]
Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agonuma_emulation: fix cpumask_of_node()
Andrea Arcangeli [Wed, 21 Mar 2012 23:34:16 +0000 (16:34 -0700)]
numa_emulation: fix cpumask_of_node()

Without this fix the cpumask_of_node() for a fake=numa=2 is:

    cpumask 0 ff
    cpumask 1 ff

with the fix it's correct and it's set to:

    cpumask 0 55
    cpumask 1 aa

Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Johannes Weiner <jweiner@redhat.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agopage_alloc: remove unused find_zone_movable_pfns_for_nodes() argument
Kautuk Consul [Wed, 21 Mar 2012 23:34:15 +0000 (16:34 -0700)]
page_alloc: remove unused find_zone_movable_pfns_for_nodes() argument

find_zone_movable_pfns_for_nodes() does not use its argument.

Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com>
Cc: David Rientjes <rientjes@google.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>
12 years agopage_alloc.c: remove add_from_early_node_map()
Kautuk Consul [Wed, 21 Mar 2012 23:34:15 +0000 (16:34 -0700)]
page_alloc.c: remove add_from_early_node_map()

add_from_early_node_map() is unused.

Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohugetlbfs: return error code when initializing module
Hillf Danton [Wed, 21 Mar 2012 23:34:15 +0000 (16:34 -0700)]
hugetlbfs: return error code when initializing module

Return an errno upon failure to create inode kmem cache, and unregister
the FS upon failure to mount.

[akpm@linux-foundation.org: remove unneeded test of `error']
Signed-off-by: Hillf Danton <dhillf@gmail.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohugetlb: remove prev_vma from hugetlb_get_unmapped_area_topdown()
Xiao Guangrong [Wed, 21 Mar 2012 23:34:14 +0000 (16:34 -0700)]
hugetlb: remove prev_vma from hugetlb_get_unmapped_area_topdown()

After looking up the vma which covers or follows the cached search
address, the following condition is always true:

!prev_vma || (addr >= prev_vma->vm_end)

so we can stop checking the previous VMA altogether.

Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohugetlbfs: fix alignment of huge page requests
Steven Truelove [Wed, 21 Mar 2012 23:34:14 +0000 (16:34 -0700)]
hugetlbfs: fix alignment of huge page requests

When calling shmget() with SHM_HUGETLB, shmget aligns the request size to
PAGE_SIZE, but this is not sufficient.

Modify hugetlb_file_setup() to align requests to the huge page size, and
to accept an address argument so that all alignment checks can be
performed in hugetlb_file_setup(), rather than in its callers.  Change
newseg() and mmap_pgoff() to match the new prototype and eliminate a now
redundant alignment check.

[akpm@linux-foundation.org: fix build]
Signed-off-by: Steven Truelove <steven.truelove@utoronto.ca>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, hugetlb: add thread name and pid to SHM_HUGETLB mlock rlimit warning
David Rientjes [Wed, 21 Mar 2012 23:34:13 +0000 (16:34 -0700)]
mm, hugetlb: add thread name and pid to SHM_HUGETLB mlock rlimit warning

Add the thread name and pid of the application that is allocating shm
segments with MAP_HUGETLB without being a part of
/proc/sys/vm/hugetlb_shm_group or having CAP_IPC_LOCK.

This identifies the application so it may be fixed by avoiding using the
deprecated exception (see Documentation/feature-removal-schedule.txt).

Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Dave Jones <davej@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, counters: fold __sync_task_rss_stat() into sync_mm_rss()
David Rientjes [Wed, 21 Mar 2012 23:34:13 +0000 (16:34 -0700)]
mm, counters: fold __sync_task_rss_stat() into sync_mm_rss()

There's no difference between sync_mm_rss() and __sync_task_rss_stat(),
so fold the latter into the former.

Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, counters: remove task argument to sync_mm_rss() and __sync_task_rss_stat()
David Rientjes [Wed, 21 Mar 2012 23:34:13 +0000 (16:34 -0700)]
mm, counters: remove task argument to sync_mm_rss() and __sync_task_rss_stat()

sync_mm_rss() can only be used for current to avoid race conditions in
iterating and clearing its per-task counters.  Remove the task argument
for it and its helper function, __sync_task_rss_stat(), to avoid thinking
it can be used safely for anything other than current.

Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohugepages: fix use after free bug in "quota" handling
David Gibson [Wed, 21 Mar 2012 23:34:12 +0000 (16:34 -0700)]
hugepages: fix use after free bug in "quota" handling

hugetlbfs_{get,put}_quota() are badly named.  They don't interact with the
general quota handling code, and they don't much resemble its behaviour.
Rather than being about maintaining limits on on-disk block usage by
particular users, they are instead about maintaining limits on in-memory
page usage (including anonymous MAP_PRIVATE copied-on-write pages)
associated with a particular hugetlbfs filesystem instance.

Worse, they work by having callbacks to the hugetlbfs filesystem code from
the low-level page handling code, in particular from free_huge_page().
This is a layering violation of itself, but more importantly, if the
kernel does a get_user_pages() on hugepages (which can happen from KVM
amongst others), then the free_huge_page() can be delayed until after the
associated inode has already been freed.  If an unmount occurs at the
wrong time, even the hugetlbfs superblock where the "quota" limits are
stored may have been freed.

Andrew Barry proposed a patch to fix this by having hugepages, instead of
storing a pointer to their address_space and reaching the superblock from
there, had the hugepages store pointers directly to the superblock,
bumping the reference count as appropriate to avoid it being freed.
Andrew Morton rejected that version, however, on the grounds that it made
the existing layering violation worse.

This is a reworked version of Andrew's patch, which removes the extra, and
some of the existing, layering violation.  It works by introducing the
concept of a hugepage "subpool" at the lower hugepage mm layer - that is a
finite logical pool of hugepages to allocate from.  hugetlbfs now creates
a subpool for each filesystem instance with a page limit set, and a
pointer to the subpool gets added to each allocated hugepage, instead of
the address_space pointer used now.  The subpool has its own lifetime and
is only freed once all pages in it _and_ all other references to it (i.e.
superblocks) are gone.

subpools are optional - a NULL subpool pointer is taken by the code to
mean that no subpool limits are in effect.

Previous discussion of this bug found in:  "Fix refcounting in hugetlbfs
quota handling.". See:  https://lkml.org/lkml/2011/8/11/28 or
http://marc.info/?l=linux-mm&m=126928970510627&w=1

v2: Fixed a bug spotted by Hillf Danton, and removed the extra parameter to
alloc_huge_page() - since it already takes the vma, it is not necessary.

Signed-off-by: Andrew Barry <abarry@cray.com>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Paul Mackerras <paulus@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohugetlb: cleanup hugetlb.h
David Gibson [Wed, 21 Mar 2012 23:34:12 +0000 (16:34 -0700)]
hugetlb: cleanup hugetlb.h

Make a couple of small cleanups to linux/include/hugetlb.h.  The
set_file_hugepages() function, which was not used anywhere is removed,
and the hugetlbfs_config and hugetlbfs_inode_info structures with its
HUGETLBFS_I helper function are moved into inode.c, the only place they
were used.

These structures are really linked to the hugetlbfs filesystem
specifically not to hugepage mm handling in general, so they belong in
the filesystem code not in a generally available header.

It would be nice to move the hugetlbfs_sb_info (superblock) structure in
there as well, but it's currently needed in a number of places via the
hstate_vma() and hstate_inode().

Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Cc: Hugh Dickins <hughd@google.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Andrew Barry <abarry@cray.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: Hillf Danton <dhillf@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoksm: cleanup: introduce find_mergeable_vma()
Bob Liu [Wed, 21 Mar 2012 23:34:11 +0000 (16:34 -0700)]
ksm: cleanup: introduce find_mergeable_vma()

There are multiple places which perform the same check.  Add a new
find_mergeable_vma() to handle this.

Signed-off-by: Bob Liu <lliubbo@gmail.com>
Acked-by: Hugh Dickins <hughd@google.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocpuset: mm: reduce large amounts of memory barrier related damage v3
Mel Gorman [Wed, 21 Mar 2012 23:34:11 +0000 (16:34 -0700)]
cpuset: mm: reduce large amounts of memory barrier related damage v3

Commit c0ff7453bb5c ("cpuset,mm: fix no node to alloc memory when
changing cpuset's mems") wins a super prize for the largest number of
memory barriers entered into fast paths for one commit.

[get|put]_mems_allowed is incredibly heavy with pairs of full memory
barriers inserted into a number of hot paths.  This was detected while
investigating at large page allocator slowdown introduced some time
after 2.6.32.  The largest portion of this overhead was shown by
oprofile to be at an mfence introduced by this commit into the page
allocator hot path.

For extra style points, the commit introduced the use of yield() in an
implementation of what looks like a spinning mutex.

This patch replaces the full memory barriers on both read and write
sides with a sequence counter with just read barriers on the fast path
side.  This is much cheaper on some architectures, including x86.  The
main bulk of the patch is the retry logic if the nodemask changes in a
manner that can cause a false failure.

While updating the nodemask, a check is made to see if a false failure
is a risk.  If it is, the sequence number gets bumped and parallel
allocators will briefly stall while the nodemask update takes place.

In a page fault test microbenchmark, oprofile samples from
__alloc_pages_nodemask went from 4.53% of all samples to 1.15%.  The
actual results were

                             3.3.0-rc3          3.3.0-rc3
                             rc3-vanilla        nobarrier-v2r1
    Clients   1 UserTime       0.07 (  0.00%)   0.08 (-14.19%)
    Clients   2 UserTime       0.07 (  0.00%)   0.07 (  2.72%)
    Clients   4 UserTime       0.08 (  0.00%)   0.07 (  3.29%)
    Clients   1 SysTime        0.70 (  0.00%)   0.65 (  6.65%)
    Clients   2 SysTime        0.85 (  0.00%)   0.82 (  3.65%)
    Clients   4 SysTime        1.41 (  0.00%)   1.41 (  0.32%)
    Clients   1 WallTime       0.77 (  0.00%)   0.74 (  4.19%)
    Clients   2 WallTime       0.47 (  0.00%)   0.45 (  3.73%)
    Clients   4 WallTime       0.38 (  0.00%)   0.37 (  1.58%)
    Clients   1 Flt/sec/cpu  497620.28 (  0.00%) 520294.53 (  4.56%)
    Clients   2 Flt/sec/cpu  414639.05 (  0.00%) 429882.01 (  3.68%)
    Clients   4 Flt/sec/cpu  257959.16 (  0.00%) 258761.48 (  0.31%)
    Clients   1 Flt/sec      495161.39 (  0.00%) 517292.87 (  4.47%)
    Clients   2 Flt/sec      820325.95 (  0.00%) 850289.77 (  3.65%)
    Clients   4 Flt/sec      1020068.93 (  0.00%) 1022674.06 (  0.26%)
    MMTests Statistics: duration
    Sys Time Running Test (seconds)             135.68    132.17
    User+Sys Time Running Test (seconds)         164.2    160.13
    Total Elapsed Time (seconds)                123.46    120.87

The overall improvement is small but the System CPU time is much
improved and roughly in correlation to what oprofile reported (these
performance figures are without profiling so skew is expected).  The
actual number of page faults is noticeably improved.

For benchmarks like kernel builds, the overall benefit is marginal but
the system CPU time is slightly reduced.

To test the actual bug the commit fixed I opened two terminals.  The
first ran within a cpuset and continually ran a small program that
faulted 100M of anonymous data.  In a second window, the nodemask of the
cpuset was continually randomised in a loop.

Without the commit, the program would fail every so often (usually
within 10 seconds) and obviously with the commit everything worked fine.
With this patch applied, it also worked fine so the fix should be
functionally equivalent.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Miao Xie <miaox@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Christoph Lameter <cl@linux.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, memcg: pass charge order to oom killer
David Rientjes [Wed, 21 Mar 2012 23:34:10 +0000 (16:34 -0700)]
mm, memcg: pass charge order to oom killer

The oom killer typically displays the allocation order at the time of oom
as a part of its diangostic messages (for global, cpuset, and mempolicy
ooms).

The memory controller may also pass the charge order to the oom killer so
it can emit the same information.  This is useful in determining how large
the memory allocation is that triggered the oom killer.

Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Balbir Singh <bsingharora@gmail.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm/vmscan.c: fix spelling error
Copot Alexandru [Wed, 21 Mar 2012 23:34:10 +0000 (16:34 -0700)]
mm/vmscan.c: fix spelling error

s/noticable/noticeable/

Signed-off-by: Copot Alexandru <alex.mihai.c@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: update stale lock ordering comment for memory-failure.c
Andi Kleen [Wed, 21 Mar 2012 23:34:09 +0000 (16:34 -0700)]
mm: update stale lock ordering comment for memory-failure.c

When i_mmap_lock changed to a mutex the locking order in memory failure
was changed to take the sleeping lock first.  But the big fat mm lock
ordering comment (BFMLO) wasn't updated.  Do this here.

Pointed out by Andrew.

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Cc: 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>
12 years agomm: use global_dirty_limit in throttle_vm_writeout()
Fengguang Wu [Wed, 21 Mar 2012 23:34:09 +0000 (16:34 -0700)]
mm: use global_dirty_limit in throttle_vm_writeout()

When starting a memory hog task, a desktop box w/o swap is found to go
unresponsive for a long time.  It's solely caused by lots of congestion
waits in throttle_vm_writeout():

 gnome-system-mo-4201 553.073384: congestion_wait: throttle_vm_writeout+0x70/0x7f shrink_mem_cgroup_zone+0x48f/0x4a1
 gnome-system-mo-4201 553.073386: writeback_congestion_wait: usec_timeout=100000 usec_delayed=100000
           gtali-4237 553.080377: congestion_wait: throttle_vm_writeout+0x70/0x7f shrink_mem_cgroup_zone+0x48f/0x4a1
           gtali-4237 553.080378: writeback_congestion_wait: usec_timeout=100000 usec_delayed=100000
            Xorg-3483 553.103375: congestion_wait: throttle_vm_writeout+0x70/0x7f shrink_mem_cgroup_zone+0x48f/0x4a1
            Xorg-3483 553.103377: writeback_congestion_wait: usec_timeout=100000 usec_delayed=100000

The root cause is, the dirty threshold is knocked down a lot by the memory
hog task.  Fixed by using global_dirty_limit which decreases gradually on
such events and can guarantee we stay above (the also decreasing) nr_dirty
in the progress of following down to the new dirty threshold.

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Greg Thelen <gthelen@google.com>
Cc: Ying Han <yinghan@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
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>
12 years agomm: don't set __GFP_WRITE on ramfs/sysfs writes
Fengguang Wu [Wed, 21 Mar 2012 23:34:08 +0000 (16:34 -0700)]
mm: don't set __GFP_WRITE on ramfs/sysfs writes

There is not much point in skipping zones during allocation based on the
dirty usage which they'll never contribute to.  And we'd like to avoid
page reclaim waits when writing to ramfs/sysfs etc.

Signed-off-by: Fengguang Wu <fengguang.wu@intel.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Jan Kara <jack@suse.cz>
Cc: Greg Thelen <gthelen@google.com>
Cc: Ying Han <yinghan@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Cc: Mel Gorman <mgorman@suse.de>
Acked-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>
12 years agohugetlbfs: avoid taking i_mutex from hugetlbfs_read()
Aneesh Kumar K.V [Wed, 21 Mar 2012 23:34:08 +0000 (16:34 -0700)]
hugetlbfs: avoid taking i_mutex from hugetlbfs_read()

Taking i_mutex in hugetlbfs_read() can result in deadlock with mmap as
explained below

 Thread A:
  read() on hugetlbfs
   hugetlbfs_read() called
    i_mutex grabbed
     hugetlbfs_read_actor() called
      __copy_to_user() called
       page fault is triggered
 Thread B, sharing address space with A:
  mmap() the same file
   ->mmap_sem is grabbed on task_B->mm->mmap_sem
    hugetlbfs_file_mmap() is called
     attempt to grab ->i_mutex and block waiting for A to give it up
 Thread A:
  pagefault handled blocked on attempt to grab task_A->mm->mmap_sem,
 which happens to be the same thing as task_B->mm->mmap_sem.  Block waiting
 for B to give it up.

AFAIU the i_mutex locking was added to hugetlbfs_read() as per
http://lkml.indiana.edu/hypermail/linux/kernel/0707.2/3066.html to take
care of the race between truncate and read.  This patch fixes this by
looking at page->mapping under lock_page() (find_lock_page()) to ensure
that the inode didn't get truncated in the range during a parallel read.

Ideally we can extend the patch to make sure we don't increase i_size in
mmap.  But that will break userspace, because applications will now have
to use truncate(2) to increase i_size in hugetlbfs.

Based on the original patch from Hillf Danton.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Hugh Dickins <hughd@google.com>
Cc: <stable@kernel.org> [everything after 2007 :)]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agobootmem/sparsemem: remove limit constraint in alloc_bootmem_section
Nishanth Aravamudan [Wed, 21 Mar 2012 23:34:07 +0000 (16:34 -0700)]
bootmem/sparsemem: remove limit constraint in alloc_bootmem_section

While testing AMS (Active Memory Sharing) / CMO (Cooperative Memory
Overcommit) on powerpc, we tripped the following:

  kernel BUG at mm/bootmem.c:483!
  cpu 0x0: Vector: 700 (Program Check) at [c000000000c03940]
      pc: c000000000a62bd8: .alloc_bootmem_core+0x90/0x39c
      lr: c000000000a64bcc: .sparse_early_usemaps_alloc_node+0x84/0x29c
      sp: c000000000c03bc0
     msr: 8000000000021032
    current = 0xc000000000b0cce0
    paca    = 0xc000000001d80000
      pid   = 0, comm = swapper
  kernel BUG at mm/bootmem.c:483!
  enter ? for help
  [c000000000c03c80c000000000a64bcc
  .sparse_early_usemaps_alloc_node+0x84/0x29c
  [c000000000c03d50c000000000a64f10 .sparse_init+0x12c/0x28c
  [c000000000c03e20c000000000a474f4 .setup_arch+0x20c/0x294
  [c000000000c03ee0c000000000a4079c .start_kernel+0xb4/0x460
  [c000000000c03f90c000000000009670 .start_here_common+0x1c/0x2c

This is

        BUG_ON(limit && goal + size > limit);

and after some debugging, it seems that

goal = 0x7ffff000000
limit = 0x80000000000

and sparse_early_usemaps_alloc_node ->
sparse_early_usemaps_alloc_pgdat_section calls

return alloc_bootmem_section(usemap_size() * count, section_nr);

This is on a system with 8TB available via the AMS pool, and as a quirk
of AMS in firmware, all of that memory shows up in node 0.  So, we end
up with an allocation that will fail the goal/limit constraints.

In theory, we could "fall-back" to alloc_bootmem_node() in
sparse_early_usemaps_alloc_node(), but since we actually have HOTREMOVE
defined, we'll BUG_ON() instead.  A simple solution appears to be to
unconditionally remove the limit condition in alloc_bootmem_section,
meaning allocations are allowed to cross section boundaries (necessary
for systems of this size).

Johannes Weiner pointed out that if alloc_bootmem_section() no longer
guarantees section-locality, we need check_usemap_section_nr() to print
possible cross-dependencies between node descriptors and the usemaps
allocated through it.  That makes the two loops in
sparse_early_usemaps_alloc_node() identical, so re-factor the code a
bit.

[akpm@linux-foundation.org: code simplification]
Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com>
Cc: Dave Hansen <haveblue@us.ibm.com>
Cc: Anton Blanchard <anton@au1.ibm.com>
Cc: Paul Mackerras <paulus@samba.org>
Cc: Ben Herrenschmidt <benh@kernel.crashing.org>
Cc: Robert Jennings <rcj@linux.vnet.ibm.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: <stable@vger.kernel.org> [3.3.1]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: drain percpu lru add/rotate page-vectors on cpu hot-unplug
Konstantin Khlebnikov [Wed, 21 Mar 2012 23:34:06 +0000 (16:34 -0700)]
mm: drain percpu lru add/rotate page-vectors on cpu hot-unplug

This cpu hotplug hook was accidentally removed in commit 00a62ce91e55
("mm: fix Committed_AS underflow on large NR_CPUS environment")

The visible effect of this accident: some pages are borrowed in per-cpu
page-vectors.  Truncate can deal with it, but these pages cannot be
reused while this cpu is offline.  So this is like a temporary memory
leak.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Eric B Munson <ebmunson@us.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Christoph Lameter <cl@linux-foundation.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: fix move/migrate_pages() race on task struct
Christoph Lameter [Wed, 21 Mar 2012 23:34:06 +0000 (16:34 -0700)]
mm: fix move/migrate_pages() race on task struct

Migration functions perform the rcu_read_unlock too early.  As a result
the task pointed to may change from under us.  This can result in an oops,
as reported by Dave Hansen in https://lkml.org/lkml/2012/2/23/302.

The following patch extend the period of the rcu_read_lock until after the
permissions checks are done.  We also take a refcount so that the task
reference is stable when calling security check functions and performing
cpuset node validation (which takes a mutex).

The refcount is dropped before actual page migration occurs so there is no
change to the refcounts held during page migration.

Also move the determination of the mm of the task struct to immediately
before the do_migrate*() calls so that it is clear that we switch from
handling the task during permission checks to the mm for the actual
migration.  Since the determination is only done once and we then no
longer use the task_struct we can be sure that we operate on a specific
address space that will not change from under us.

[akpm@linux-foundation.org: checkpatch fixes]
Signed-off-by: Christoph Lameter <cl@linux.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Reported-by: Dave Hansen <dave@linux.vnet.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agothp: allow a hwpoisoned head page to be put back to LRU
Dean Nelson [Wed, 21 Mar 2012 23:34:05 +0000 (16:34 -0700)]
thp: allow a hwpoisoned head page to be put back to LRU

Andrea Arcangeli pointed out to me that a check in __memory_failure()
which was intended to prevent THP tail pages from being checked for the
absence of the PG_lru flag (something that is always the case), was also
preventing THP head pages from being checked.

A THP head page could actually benefit from the call to shake_page() by
ending up being put back to a LRU, provided it had been waiting in a
pagevec array.

Andrea suggested that the "!PageTransCompound(p)" in the if-statement
should be replaced by a "!PageTransTail(p)", thus allowing THP head pages
to be checked and possibly shaken.

Signed-off-by: Dean Nelson <dnelson@redhat.com>
Cc: Jin Dongming <jin.dongming@np.css.fujitsu.com>
Reviewed-by: Andrea Arcangeli <aarcange@redhat.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agotmpfs: security xattr setting on inode creation
Jarkko Sakkinen [Wed, 21 Mar 2012 23:34:05 +0000 (16:34 -0700)]
tmpfs: security xattr setting on inode creation

Adds to generic xattr support introduced in Linux 3.0 by implementing
initxattrs callback.  This enables consulting of security attributes from
LSM and EVM when inode is created.

[hughd@google.com: moved under CONFIG_TMPFS_XATTR, with memcpy in shmem_xattr_alloc]
Signed-off-by: Jarkko Sakkinen <jarkko.sakkinen@intel.com>
Reviewed-by: James Morris <james.l.morris@oracle.com>
Signed-off-by: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, oom: force oom kill on sysrq+f
David Rientjes [Wed, 21 Mar 2012 23:34:04 +0000 (16:34 -0700)]
mm, oom: force oom kill on sysrq+f

The oom killer chooses not to kill a thread if:

 - an eligible thread has already been oom killed and has yet to exit,
   and

 - an eligible thread is exiting but has yet to free all its memory and
   is not the thread attempting to currently allocate memory.

SysRq+F manually invokes the global oom killer to kill a memory-hogging
task.  This is normally done as a last resort to free memory when no
progress is being made or to test the oom killer itself.

For both uses, we always want to kill a thread and never defer.  This
patch causes SysRq+F to always kill an eligible thread and can be used to
force a kill even if another oom killed thread has failed to exit.

Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Pekka Enberg <penberg@kernel.org>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoprocfs: mark thread stack correctly in proc/<pid>/maps
Siddhesh Poyarekar [Wed, 21 Mar 2012 23:34:04 +0000 (16:34 -0700)]
procfs: mark thread stack correctly in proc/<pid>/maps

Stack for a new thread is mapped by userspace code and passed via
sys_clone.  This memory is currently seen as anonymous in
/proc/<pid>/maps, which makes it difficult to ascertain which mappings
are being used for thread stacks.  This patch uses the individual task
stack pointers to determine which vmas are actually thread stacks.

For a multithreaded program like the following:

#include <pthread.h>

void *thread_main(void *foo)
{
while(1);
}

int main()
{
pthread_t t;
pthread_create(&t, NULL, thread_main, NULL);
pthread_join(t, NULL);
}

proc/PID/maps looks like the following:

    00400000-00401000 r-xp 00000000 fd:0a 3671804                            /home/siddhesh/a.out
    00600000-00601000 rw-p 00000000 fd:0a 3671804                            /home/siddhesh/a.out
    019ef000-01a10000 rw-p 00000000 00:00 0                                  [heap]
    7f8a44491000-7f8a44492000 ---p 00000000 00:00 0
    7f8a44492000-7f8a44c92000 rw-p 00000000 00:00 0
    7f8a44c92000-7f8a44e3d000 r-xp 00000000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a44e3d000-7f8a4503d000 ---p 001ab000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a4503d000-7f8a45041000 r--p 001ab000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a45041000-7f8a45043000 rw-p 001af000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a45043000-7f8a45048000 rw-p 00000000 00:00 0
    7f8a45048000-7f8a4505f000 r-xp 00000000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4505f000-7f8a4525e000 ---p 00017000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4525e000-7f8a4525f000 r--p 00016000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4525f000-7f8a45260000 rw-p 00017000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a45260000-7f8a45264000 rw-p 00000000 00:00 0
    7f8a45264000-7f8a45286000 r-xp 00000000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45457000-7f8a4545a000 rw-p 00000000 00:00 0
    7f8a45484000-7f8a45485000 rw-p 00000000 00:00 0
    7f8a45485000-7f8a45486000 r--p 00021000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45486000-7f8a45487000 rw-p 00022000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45487000-7f8a45488000 rw-p 00000000 00:00 0
    7fff6273b000-7fff6275c000 rw-p 00000000 00:00 0                          [stack]
    7fff627ff000-7fff62800000 r-xp 00000000 00:00 0                          [vdso]
    ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0                  [vsyscall]

Here, one could guess that 7f8a44492000-7f8a44c92000 is a stack since
the earlier vma that has no permissions (7f8a44e3d000-7f8a4503d000) but
that is not always a reliable way to find out which vma is a thread
stack.  Also, /proc/PID/maps and /proc/PID/task/TID/maps has the same
content.

With this patch in place, /proc/PID/task/TID/maps are treated as 'maps
as the task would see it' and hence, only the vma that that task uses as
stack is marked as [stack].  All other 'stack' vmas are marked as
anonymous memory.  /proc/PID/maps acts as a thread group level view,
where all thread stack vmas are marked as [stack:TID] where TID is the
process ID of the task that uses that vma as stack, while the process
stack is marked as [stack].

So /proc/PID/maps will look like this:

    00400000-00401000 r-xp 00000000 fd:0a 3671804                            /home/siddhesh/a.out
    00600000-00601000 rw-p 00000000 fd:0a 3671804                            /home/siddhesh/a.out
    019ef000-01a10000 rw-p 00000000 00:00 0                                  [heap]
    7f8a44491000-7f8a44492000 ---p 00000000 00:00 0
    7f8a44492000-7f8a44c92000 rw-p 00000000 00:00 0                          [stack:1442]
    7f8a44c92000-7f8a44e3d000 r-xp 00000000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a44e3d000-7f8a4503d000 ---p 001ab000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a4503d000-7f8a45041000 r--p 001ab000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a45041000-7f8a45043000 rw-p 001af000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a45043000-7f8a45048000 rw-p 00000000 00:00 0
    7f8a45048000-7f8a4505f000 r-xp 00000000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4505f000-7f8a4525e000 ---p 00017000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4525e000-7f8a4525f000 r--p 00016000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4525f000-7f8a45260000 rw-p 00017000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a45260000-7f8a45264000 rw-p 00000000 00:00 0
    7f8a45264000-7f8a45286000 r-xp 00000000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45457000-7f8a4545a000 rw-p 00000000 00:00 0
    7f8a45484000-7f8a45485000 rw-p 00000000 00:00 0
    7f8a45485000-7f8a45486000 r--p 00021000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45486000-7f8a45487000 rw-p 00022000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45487000-7f8a45488000 rw-p 00000000 00:00 0
    7fff6273b000-7fff6275c000 rw-p 00000000 00:00 0                          [stack]
    7fff627ff000-7fff62800000 r-xp 00000000 00:00 0                          [vdso]
    ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0                  [vsyscall]

Thus marking all vmas that are used as stacks by the threads in the
thread group along with the process stack.  The task level maps will
however like this:

    00400000-00401000 r-xp 00000000 fd:0a 3671804                            /home/siddhesh/a.out
    00600000-00601000 rw-p 00000000 fd:0a 3671804                            /home/siddhesh/a.out
    019ef000-01a10000 rw-p 00000000 00:00 0                                  [heap]
    7f8a44491000-7f8a44492000 ---p 00000000 00:00 0
    7f8a44492000-7f8a44c92000 rw-p 00000000 00:00 0                          [stack]
    7f8a44c92000-7f8a44e3d000 r-xp 00000000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a44e3d000-7f8a4503d000 ---p 001ab000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a4503d000-7f8a45041000 r--p 001ab000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a45041000-7f8a45043000 rw-p 001af000 fd:00 2097482                    /lib64/libc-2.14.90.so
    7f8a45043000-7f8a45048000 rw-p 00000000 00:00 0
    7f8a45048000-7f8a4505f000 r-xp 00000000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4505f000-7f8a4525e000 ---p 00017000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4525e000-7f8a4525f000 r--p 00016000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a4525f000-7f8a45260000 rw-p 00017000 fd:00 2099938                    /lib64/libpthread-2.14.90.so
    7f8a45260000-7f8a45264000 rw-p 00000000 00:00 0
    7f8a45264000-7f8a45286000 r-xp 00000000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45457000-7f8a4545a000 rw-p 00000000 00:00 0
    7f8a45484000-7f8a45485000 rw-p 00000000 00:00 0
    7f8a45485000-7f8a45486000 r--p 00021000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45486000-7f8a45487000 rw-p 00022000 fd:00 2097348                    /lib64/ld-2.14.90.so
    7f8a45487000-7f8a45488000 rw-p 00000000 00:00 0
    7fff6273b000-7fff6275c000 rw-p 00000000 00:00 0
    7fff627ff000-7fff62800000 r-xp 00000000 00:00 0                          [vdso]
    ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0                  [vsyscall]

where only the vma that is being used as a stack by *that* task is
marked as [stack].

Analogous changes have been made to /proc/PID/smaps,
/proc/PID/numa_maps, /proc/PID/task/TID/smaps and
/proc/PID/task/TID/numa_maps. Relevant snippets from smaps and
numa_maps:

    [siddhesh@localhost ~ ]$ pgrep a.out
    1441
    [siddhesh@localhost ~ ]$ cat /proc/1441/smaps | grep "\[stack"
    7f8a44492000-7f8a44c92000 rw-p 00000000 00:00 0                          [stack:1442]
    7fff6273b000-7fff6275c000 rw-p 00000000 00:00 0                          [stack]
    [siddhesh@localhost ~ ]$ cat /proc/1441/task/1442/smaps | grep "\[stack"
    7f8a44492000-7f8a44c92000 rw-p 00000000 00:00 0                          [stack]
    [siddhesh@localhost ~ ]$ cat /proc/1441/task/1441/smaps | grep "\[stack"
    7fff6273b000-7fff6275c000 rw-p 00000000 00:00 0                          [stack]
    [siddhesh@localhost ~ ]$ cat /proc/1441/numa_maps | grep "stack"
    7f8a44492000 default stack:1442 anon=2 dirty=2 N0=2
    7fff6273a000 default stack anon=3 dirty=3 N0=3
    [siddhesh@localhost ~ ]$ cat /proc/1441/task/1442/numa_maps | grep "stack"
    7f8a44492000 default stack anon=2 dirty=2 N0=2
    [siddhesh@localhost ~ ]$ cat /proc/1441/task/1441/numa_maps | grep "stack"
    7fff6273a000 default stack anon=3 dirty=3 N0=3

[akpm@linux-foundation.org: checkpatch fixes]
[akpm@linux-foundation.org: fix build]
Signed-off-by: Siddhesh Poyarekar <siddhesh.poyarekar@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Jamie Lokier <jamie@shareable.org>
Cc: Mike Frysinger <vapier@gentoo.org>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Matt Mackall <mpm@selenic.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: hugetlb: bail out unmapping after serving reference page
Hillf Danton [Wed, 21 Mar 2012 23:34:03 +0000 (16:34 -0700)]
mm: hugetlb: bail out unmapping after serving reference page

When unmapping a given VM range, we could bail out if a reference page is
supplied and is unmapped, which is a minor optimization.

Signed-off-by: Hillf Danton <dhillf@gmail.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agothp: documentation: 'transparent_hugepage=' can also be specified on cmdline
Jiri Kosina [Wed, 21 Mar 2012 23:34:02 +0000 (16:34 -0700)]
thp: documentation: 'transparent_hugepage=' can also be specified on cmdline

The behavior of THP can either be toggled through sysfs in runtime or
using a kernel cmdline parameter 'transparent_hugepage='.  Document the
latter in kernel-parameters.txt

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.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>
12 years agovmscan: handle isolated pages with lru lock released
Hillf Danton [Wed, 21 Mar 2012 23:34:02 +0000 (16:34 -0700)]
vmscan: handle isolated pages with lru lock released

When shrinking inactive lru list, isolated pages are queued on locally
private list, so the lock-hold time could be reduced if pages are counted
without lock protection.

To achieve that, firstly updating reclaim stat is delayed until the
putback stage, after reacquiring the lru lock.

Secondly, operations related to vm and zone stats are now proteced with
preemption disabled as they are per-cpu operations.

Signed-off-by: Hillf Danton <dhillf@gmail.com>
Acked-by: Hugh Dickins <hughd@google.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agormap: remove __anon_vma_link() declaration
Xiao Guangrong [Wed, 21 Mar 2012 23:34:01 +0000 (16:34 -0700)]
rmap: remove __anon_vma_link() declaration

This declaration is not used anymore, remove it.

Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agormap: anon_vma_prepare: Reduce code duplication by calling anon_vma_chain_link
Kautuk Consul [Wed, 21 Mar 2012 23:34:01 +0000 (16:34 -0700)]
rmap: anon_vma_prepare: Reduce code duplication by calling anon_vma_chain_link

Reduce code duplication by calling anon_vma_chain_link() from
anon_vma_prepare().

Also move anon_vmal_chain_link() to a more suitable location in the file.

Signed-off-by: Kautuk Consul <consul.kautuk@gmail.com>
Acked-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Hugh Dickins <hughd@google.com>
Reviewed-by: KAMEZWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mgorman@suse.de>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: hugetlb: defer freeing pages when gathering surplus pages
Hillf Danton [Wed, 21 Mar 2012 23:34:00 +0000 (16:34 -0700)]
mm: hugetlb: defer freeing pages when gathering surplus pages

When gathering surplus pages, the number of needed pages is recomputed
after reacquiring hugetlb lock to catch changes in resv_huge_pages and
free_huge_pages.  Plus it is recomputed with the number of newly allocated
pages involved.

Thus freeing pages can be deferred a bit to see if the final page request
is satisfied, though pages could be allocated less than needed.

Signed-off-by: Hillf Danton <dhillf@gmail.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: vmscan: forcibly scan highmem if there are too many buffer_heads pinning highmem
Mel Gorman [Wed, 21 Mar 2012 23:34:00 +0000 (16:34 -0700)]
mm: vmscan: forcibly scan highmem if there are too many buffer_heads pinning highmem

Stuart Foster reported on bugzilla that copying large amounts of data
from NTFS caused an OOM kill on 32-bit X86 with 16G of memory.  Andrew
Morton correctly identified that the problem was NTFS was using 512
blocks meaning each page had 8 buffer_heads in low memory pinning it.

In the past, direct reclaim used to scan highmem even if the allocating
process did not specify __GFP_HIGHMEM but not any more.  kswapd no longer
will reclaim from zones that are above the high watermark.  The intention
in both cases was to minimise unnecessary reclaim.  The downside is on
machines with large amounts of highmem that lowmem can be fully consumed
by buffer_heads with nothing trying to free them.

The following patch is based on a suggestion by Andrew Morton to extend
the buffer_heads_over_limit case to force kswapd and direct reclaim to
scan the highmem zone regardless of the allocation request or watermarks.

Addresses https://bugzilla.kernel.org/show_bug.cgi?id=42578

[hughd@google.com: move buffer_heads_over_limit check up]
[akpm@linux-foundation.org: buffer_heads_over_limit is unlikely]
Reported-by: Stuart Foster <smf.linux@ntlworld.com>
Tested-by: Stuart Foster <smf.linux@ntlworld.com>
Signed-off-by: Mel Gorman <mgorman@suse.de>
Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Rik van Riel <riel@redhat.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: replace PAGE_MIGRATION with IS_ENABLED(CONFIG_MIGRATION)
Konstantin Khlebnikov [Wed, 21 Mar 2012 23:33:59 +0000 (16:33 -0700)]
mm: replace PAGE_MIGRATION with IS_ENABLED(CONFIG_MIGRATION)

Since commit 2a11c8ea20bf ("kconfig: Introduce IS_ENABLED(),
IS_BUILTIN() and IS_MODULE()") there is a generic grep-friendly method
for checking config options in C expressions.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agopagemap: introduce data structure for pagemap entry
Naoya Horiguchi [Wed, 21 Mar 2012 23:33:59 +0000 (16:33 -0700)]
pagemap: introduce data structure for pagemap entry

Currently a local variable of pagemap entry in pagemap_pte_range() is
named pfn and typed with u64, but it's not correct (pfn should be unsigned
long.)

This patch introduces special type for pagemap entries and replaces code
with it.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agopagemap: document KPF_THP and make page-types aware of it
Naoya Horiguchi [Wed, 21 Mar 2012 23:33:58 +0000 (16:33 -0700)]
pagemap: document KPF_THP and make page-types aware of it

page-types, which is a common user of pagemap, gets aware of thp with this
patch.  This helps system admins and kernel hackers know about how thp
works.  Here is a sample output of page-types over a thp:

  $ page-types -p <pid> --raw --list

  voffset offset  len     flags
  ...
  7f9d40200       3f8400  1       ___U_lA____Ma_bH______t____________
  7f9d40201       3f8401  1ff     ________________T_____t____________

               flags      page-count       MB  symbolic-flags                     long-symbolic-flags
  0x0000000000410000             511        1  ________________T_____t____________        compound_tail,thp
  0x000000000040d868               1        0  ___U_lA____Ma_bH______t____________        uptodate,lru,active,mmap,anonymous,swapbacked,compound_head,thp

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Acked-by: Wu Fengguang <fengguang.wu@intel.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Andi Kleen <andi@firstfloor.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>
12 years agopagemap: export KPF_THP
Naoya Horiguchi [Wed, 21 Mar 2012 23:33:58 +0000 (16:33 -0700)]
pagemap: export KPF_THP

This flag shows that a given page is a subpage of a transparent hugepage.
It helps us debug and test the kernel by showing physical address of thp.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Reviewed-by: Wu Fengguang <fengguang.wu@intel.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Andi Kleen <andi@firstfloor.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>
12 years agothp: optimize away unnecessary page table locking
Naoya Horiguchi [Wed, 21 Mar 2012 23:33:57 +0000 (16:33 -0700)]
thp: optimize away unnecessary page table locking

Currently when we check if we can handle thp as it is or we need to split
it into regular sized pages, we hold page table lock prior to check
whether a given pmd is mapping thp or not.  Because of this, when it's not
"huge pmd" we suffer from unnecessary lock/unlock overhead.  To remove it,
this patch introduces a optimized check function and replace several
similar logics with it.

[akpm@linux-foundation.org: checkpatch fixes]
Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agopagemap: avoid splitting thp when reading /proc/pid/pagemap
Naoya Horiguchi [Wed, 21 Mar 2012 23:33:57 +0000 (16:33 -0700)]
pagemap: avoid splitting thp when reading /proc/pid/pagemap

Thp split is not necessary if we explicitly check whether pmds are mapping
thps or not.  This patch introduces this check and adds code to generate
pagemap entries for pmds mapping thps, which results in less performance
impact of pagemap on thp.

Signed-off-by: Naoya Horiguchi <n-horiguchi@ah.jp.nec.com>
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Andrea Arcangeli <aarcange@redhat.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>
12 years agomm: search from free_area_cache for the bigger size
Xiao Guangrong [Wed, 21 Mar 2012 23:33:56 +0000 (16:33 -0700)]
mm: search from free_area_cache for the bigger size

If the required size is bigger than cached_hole_size it is better to
search from free_area_cache - it is easier to get a free region,
specifically for the 64 bit process whose address space is large enough

Do it just as hugetlb_get_unmapped_area_topdown() in arch/x86/mm/hugetlbpage.c

Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: do not reset cached_hole_size when vma is unmapped
Xiao Guangrong [Wed, 21 Mar 2012 23:33:56 +0000 (16:33 -0700)]
mm: do not reset cached_hole_size when vma is unmapped

In the current code, cached_hole_size is set to the maximum value if the
unmapped vma is less that free_area_cache so the next search will search
from the base address.

Actually, we can keep cached_hole_size so that if the next required size
is more than cached_hole_size, it can search from free_area_cache.

Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohugetlb: try to search again if it is really needed
Xiao Guangrong [Wed, 21 Mar 2012 23:33:55 +0000 (16:33 -0700)]
hugetlb: try to search again if it is really needed

Search again only if some holes may be skipped in the first pass.

[akpm@linux-foundation.org: clean up crazy compound definition]
Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohugetlbfs: fix hugetlb_get_unmapped_area()
Xiao Guangrong [Wed, 21 Mar 2012 23:33:54 +0000 (16:33 -0700)]
hugetlbfs: fix hugetlb_get_unmapped_area()

Use/update cached_hole_size and free_area_cache properly to speedup
finding of a free region.

Signed-off-by: Xiao Guangrong <xiaoguangrong@linux.vnet.ibm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Hillf Danton <dhillf@gmail.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: compaction: make compact_control order signed
Dan Carpenter [Wed, 21 Mar 2012 23:33:54 +0000 (16:33 -0700)]
mm: compaction: make compact_control order signed

"order" is -1 when compacting via /proc/sys/vm/compact_memory.  Making
it unsigned causes a bug in __compact_pgdat() when we test:

if (cc->order < 0 || !compaction_deferred(zone, cc->order))
compact_zone(zone, cc);

[akpm@linux-foundation.org: make __compact_pgdat()'s comparison match other code sites]
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Minchan Kim <minchan@kernel.org>
Reviewed-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agocompact_pgdat: workaround lockdep warning in kswapd
Hugh Dickins [Wed, 21 Mar 2012 23:33:53 +0000 (16:33 -0700)]
compact_pgdat: workaround lockdep warning in kswapd

I get this lockdep warning from swapping load on linux-next, due to
"vmscan: kswapd carefully call compaction".

=================================
[ INFO: inconsistent lock state ]
3.3.0-rc2-next-20120201 #5 Not tainted
---------------------------------
inconsistent {RECLAIM_FS-ON-W} -> {IN-RECLAIM_FS-W} usage.
kswapd0/28 [HC0[0]:SC0[0]:HE1:SE1] takes:
 (pcpu_alloc_mutex){+.+.?.}, at: [<ffffffff810d6684>] pcpu_alloc+0x67/0x325
{RECLAIM_FS-ON-W} state was registered at:
  [<ffffffff81099b75>] mark_held_locks+0xd7/0x103
  [<ffffffff8109a13c>] lockdep_trace_alloc+0x85/0x9e
  [<ffffffff810f6bdc>] __kmalloc+0x6c/0x14b
  [<ffffffff810d57fd>] pcpu_mem_zalloc+0x59/0x62
  [<ffffffff810d5d16>] pcpu_extend_area_map+0x26/0xb1
  [<ffffffff810d679f>] pcpu_alloc+0x182/0x325
  [<ffffffff810d694d>] __alloc_percpu+0xb/0xd
  [<ffffffff8142ebfd>] snmp_mib_init+0x1e/0x2e
  [<ffffffff8185cd8d>] ipv4_mib_init_net+0x7a/0x184
  [<ffffffff813dc963>] ops_init.clone.0+0x6b/0x73
  [<ffffffff813dc9cc>] register_pernet_operations+0x61/0xa0
  [<ffffffff813dca8e>] register_pernet_subsys+0x29/0x42
  [<ffffffff8185d044>] inet_init+0x1ad/0x252
  [<ffffffff810002e3>] do_one_initcall+0x7a/0x12f
  [<ffffffff81832bc5>] kernel_init+0x9d/0x11e
  [<ffffffff814e51e4>] kernel_thread_helper+0x4/0x10
irq event stamp: 656613
hardirqs last  enabled at (656613): [<ffffffff814e0ddc>] __mutex_unlock_slowpath+0x104/0x128
hardirqs last disabled at (656612): [<ffffffff814e0d34>] __mutex_unlock_slowpath+0x5c/0x128
softirqs last  enabled at (655568): [<ffffffff8105b4a5>] __do_softirq+0x120/0x136
softirqs last disabled at (654757): [<ffffffff814e52dc>] call_softirq+0x1c/0x30

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(pcpu_alloc_mutex);
  <Interrupt>
    lock(pcpu_alloc_mutex);

 *** DEADLOCK ***

no locks held by kswapd0/28.

stack backtrace:
Pid: 28, comm: kswapd0 Not tainted 3.3.0-rc2-next-20120201 #5
Call Trace:
 [<ffffffff810981f4>] print_usage_bug+0x1bf/0x1d0
 [<ffffffff81096c3e>] ? print_irq_inversion_bug+0x1d9/0x1d9
 [<ffffffff810982c0>] mark_lock_irq+0xbb/0x22e
 [<ffffffff810c5399>] ? free_hot_cold_page+0x13d/0x14f
 [<ffffffff81098684>] mark_lock+0x251/0x331
 [<ffffffff81098893>] mark_irqflags+0x12f/0x141
 [<ffffffff81098e32>] __lock_acquire+0x58d/0x753
 [<ffffffff810d6684>] ? pcpu_alloc+0x67/0x325
 [<ffffffff81099433>] lock_acquire+0x54/0x6a
 [<ffffffff810d6684>] ? pcpu_alloc+0x67/0x325
 [<ffffffff8107a5b8>] ? add_preempt_count+0xa9/0xae
 [<ffffffff814e0a21>] mutex_lock_nested+0x5e/0x315
 [<ffffffff810d6684>] ? pcpu_alloc+0x67/0x325
 [<ffffffff81098f81>] ? __lock_acquire+0x6dc/0x753
 [<ffffffff810c9fb0>] ? __pagevec_release+0x2c/0x2c
 [<ffffffff810d6684>] pcpu_alloc+0x67/0x325
 [<ffffffff810c9fb0>] ? __pagevec_release+0x2c/0x2c
 [<ffffffff810d694d>] __alloc_percpu+0xb/0xd
 [<ffffffff8106c35e>] schedule_on_each_cpu+0x23/0x110
 [<ffffffff810c9fcb>] lru_add_drain_all+0x10/0x12
 [<ffffffff810f126f>] __compact_pgdat+0x20/0x182
 [<ffffffff810f15c2>] compact_pgdat+0x27/0x29
 [<ffffffff810c306b>] ? zone_watermark_ok+0x1a/0x1c
 [<ffffffff810cdf6f>] balance_pgdat+0x732/0x751
 [<ffffffff810ce0ed>] kswapd+0x15f/0x178
 [<ffffffff810cdf8e>] ? balance_pgdat+0x751/0x751
 [<ffffffff8106fd11>] kthread+0x84/0x8c
 [<ffffffff814e51e4>] kernel_thread_helper+0x4/0x10
 [<ffffffff810787ed>] ? finish_task_switch+0x85/0xea
 [<ffffffff814e3861>] ? retint_restore_args+0xe/0xe
 [<ffffffff8106fc8d>] ? __init_kthread_worker+0x56/0x56
 [<ffffffff814e51e0>] ? gs_change+0xb/0xb

The RECLAIM_FS notations indicate that it's doing the GFP_FS checking that
Nick hacked into lockdep a while back: I think we're intended to read that
"<Interrupt>" in the DEADLOCK scenario as "<Direct reclaim>".

I'm hazy, I have not reached any conclusion as to whether it's right to
complain or not; but I believe it's uneasy about kswapd now doing the
mutex_lock(&pcpu_alloc_mutex) which lru_add_drain_all() entails.  Nor have
I reached any conclusion as to whether it's important for kswapd to do
that draining or not.

But so as not to get blocked on this, with lockdep disabled from giving
further reports, here's a patch which removes the lru_add_drain_all() from
kswapd's callpath (and calls it only once from compact_nodes(), instead of
once per node).

Signed-off-by: Hugh Dickins <hughd@google.com>
Cc: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agovmscan: only defer compaction for failed order and higher
Rik van Riel [Wed, 21 Mar 2012 23:33:52 +0000 (16:33 -0700)]
vmscan: only defer compaction for failed order and higher

Currently a failed order-9 (transparent hugepage) compaction can lead to
memory compaction being temporarily disabled for a memory zone.  Even if
we only need compaction for an order 2 allocation, eg.  for jumbo frames
networking.

The fix is relatively straightforward: keep track of the highest order at
which compaction is succeeding, and only defer compaction for orders at
which compaction is failing.

Signed-off-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Hillf Danton <dhillf@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agovmscan: kswapd carefully call compaction
Rik van Riel [Wed, 21 Mar 2012 23:33:52 +0000 (16:33 -0700)]
vmscan: kswapd carefully call compaction

With CONFIG_COMPACTION enabled, kswapd does not try to free contiguous
free pages, even when it is woken for a higher order request.

This could be bad for eg.  jumbo frame network allocations, which are done
from interrupt context and cannot compact memory themselves.  Higher than
before allocation failure rates in the network receive path have been
observed in kernels with compaction enabled.

Teach kswapd to defragment the memory zones in a node, but only if
required and compaction is not deferred in a zone.

[akpm@linux-foundation.org: reduce scope of zones_need_compaction]
Signed-off-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Hillf Danton <dhillf@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agovmscan: reclaim at order 0 when compaction is enabled
Rik van Riel [Wed, 21 Mar 2012 23:33:51 +0000 (16:33 -0700)]
vmscan: reclaim at order 0 when compaction is enabled

When built with CONFIG_COMPACTION, kswapd should not try to free
contiguous pages, because it is not trying hard enough to have a real
chance at being successful, but still disrupts the LRU enough to break
other things.

Do not do higher order page isolation unless we really are in lumpy
reclaim mode.

Stop reclaiming pages once we have enough free pages that compaction can
deal with things, and we hit the normal order 0 watermarks used by kswapd.

Also remove a line of code that increments balanced right before exiting
the function.

Signed-off-by: Rik van Riel <riel@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Hillf Danton <dhillf@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: make swapin readahead skip over holes
Rik van Riel [Wed, 21 Mar 2012 23:33:50 +0000 (16:33 -0700)]
mm: make swapin readahead skip over holes

Ever since abandoning the virtual scan of processes, for scalability
reasons, swap space has been a little more fragmented than before.  This
can lead to the situation where a large memory user is killed, swap space
ends up full of "holes" and swapin readahead is totally ineffective.

On my home system, after killing a leaky firefox it took over an hour to
page just under 2GB of memory back in, slowing the virtual machines down
to a crawl.

This patch makes swapin readahead simply skip over holes, instead of
stopping at them.  This allows the system to swap things back in at rates
of several MB/second, instead of a few hundred kB/second.

The checks done in valid_swaphandles are already done in
read_swap_cache_async as well, allowing us to remove a fair amount of
code.

[akpm@linux-foundation.org: fix it for page_cluster >= 32]
Signed-off-by: Rik van Riel <riel@redhat.com>
Cc: Minchan Kim <minchan.kim@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@gmail.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Mel Gorman <mgorman@suse.de>
Cc: Adrian Drzewiecki <z@drze.net>
Cc: Hugh Dickins <hughd@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: vmscan: fix misused nr_reclaimed in shrink_mem_cgroup_zone()
Hillf Danton [Wed, 21 Mar 2012 23:33:50 +0000 (16:33 -0700)]
mm: vmscan: fix misused nr_reclaimed in shrink_mem_cgroup_zone()

The value of nr_reclaimed is the number of pages reclaimed in the current
round of the loop, whereas nr_to_reclaim should be compared with the
number of pages reclaimed in all rounds.

In each round of the loop, reclaimed pages are cut off from the reclaim
goal, and the loop stops once the goal achieved.

Signed-off-by: Hillf Danton <dhillf@gmail.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: make get_mm_counter static-inline
Konstantin Khlebnikov [Wed, 21 Mar 2012 23:33:49 +0000 (16:33 -0700)]
mm: make get_mm_counter static-inline

Make get_mm_counter() always static inline, it is simple enough for that.
And remove unused set_mm_counter()

bloat-o-meter:

add/remove: 0/1 grow/shrink: 4/12 up/down: 99/-341 (-242)
function                                     old     new   delta
try_to_unmap_one                             886     952     +66
sys_remap_file_pages                        1214    1230     +16
dup_mm                                      1684    1700     +16
do_exit                                     2277    2278      +1
zap_page_range                               208     205      -3
unmap_region                                 304     296      -8
static.oom_kill_process                      554     546      -8
try_to_unmap_file                           1716    1700     -16
getrusage                                    925     909     -16
flush_old_exec                              1704    1688     -16
static.dump_header                           416     390     -26
acct_update_integrals                        218     187     -31
do_task_stat                                2986    2954     -32
get_mm_counter                                34       -     -34
xacct_add_tsk                                371     334     -37
task_statm                                   172     118     -54
task_mem                                     383     323     -60

try_to_unmap_one() grows because update_hiwater_rss() now completely inline.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm/vmscan.c: cleanup with s/reclaim_mode/isolate_mode/
Hillf Danton [Wed, 21 Mar 2012 23:33:48 +0000 (16:33 -0700)]
mm/vmscan.c: cleanup with s/reclaim_mode/isolate_mode/

With tons of reclaim_mode (defined as one field of struct scan_control)
already in the file, it is clearer to rename the local reclaim_mode when
setting up the isolation mode.

Signed-off-by: Hillf Danton <dhillf@gmail.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: Rik van Riel <riel@redhat.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>
12 years agomm: add rss counters consistency check
Konstantin Khlebnikov [Wed, 21 Mar 2012 23:33:48 +0000 (16:33 -0700)]
mm: add rss counters consistency check

Warn about non-zero rss counters at final mmdrop.

This check will prevent reoccurences of bugs such as that fixed in "mm:
fix rss count leakage during migration".

I didn't hide this check under CONFIG_VM_DEBUG because it rather small and
rss counters cover whole page-table management, so this is a good
invariant.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Hugh Dickins <hughd@google.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, oom: introduce independent oom killer ratelimit state
David Rientjes [Wed, 21 Mar 2012 23:33:47 +0000 (16:33 -0700)]
mm, oom: introduce independent oom killer ratelimit state

printk_ratelimit() uses the global ratelimit state for all printks.  The
oom killer should not be subjected to this state just because another
subsystem or driver may be flooding the kernel log.

This patch introduces printk ratelimiting specifically for the oom killer.

Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, oom: do not emit oom killer warning if chosen thread is already exiting
David Rientjes [Wed, 21 Mar 2012 23:33:47 +0000 (16:33 -0700)]
mm, oom: do not emit oom killer warning if chosen thread is already exiting

If a thread is chosen for oom kill and is already PF_EXITING, then the oom
killer simply sets TIF_MEMDIE and returns.  This allows the thread to have
access to memory reserves so that it may quickly exit.  This logic is
preceeded with a comment saying there's no need to alarm the sysadmin.
This patch adds truth to that statement.

There's no need to emit any warning about the oom condition if the thread
is already exiting since it will not be killed.  In this condition, just
silently return the oom killer since its only giving access to memory
reserves and is otherwise a no-op.

Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, oom: fold oom_kill_task() into oom_kill_process()
David Rientjes [Wed, 21 Mar 2012 23:33:46 +0000 (16:33 -0700)]
mm, oom: fold oom_kill_task() into oom_kill_process()

oom_kill_task() has a single caller, so fold it into its parent function,
oom_kill_process().  Slightly reduces the number of lines in the oom
killer.

Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm, oom: avoid looping when chosen thread detaches its mm
David Rientjes [Wed, 21 Mar 2012 23:33:46 +0000 (16:33 -0700)]
mm, oom: avoid looping when chosen thread detaches its mm

oom_kill_task() returns non-zero iff the chosen process does not have any
threads with an attached ->mm.

In such a case, it's better to just return to the page allocator and retry
the allocation because memory could have been freed in the interim and the
oom condition may no longer exist.  It's unnecessary to loop in the oom
killer and find another thread to kill.

This allows both oom_kill_task() and oom_kill_process() to be converted to
void functions.  If the oom condition persists, the oom killer will be
recalled.

Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agosparc: use block_sigmask()
Matt Fleming [Wed, 21 Mar 2012 23:33:46 +0000 (16:33 -0700)]
sparc: use block_sigmask()

Use the new helper function introduced in commit 5e6292c0f28f ("signal:
add block_sigmask() for adding sigmask to current->blocked") which
centralises the code for updating current->blocked after successfully
delivering a signal and reduces the amount of duplicate code across
architectures.  In the past some architectures got this code wrong, so
using this helper function should stop that from happening again.

Acked-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: "David S. Miller" <davem@davemloft.net>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoxtensa: use set_current_blocked() and block_sigmask()
Matt Fleming [Wed, 21 Mar 2012 23:33:45 +0000 (16:33 -0700)]
xtensa: use set_current_blocked() and block_sigmask()

As described in commit e6fa16ab9c1e ("signal: sigprocmask() should do
retarget_shared_pending()") the modification of current->blocked is
incorrect as we need to check whether the signal we're about to block is
pending in the shared queue.

Also, use the new helper function introduced in commit 5e6292c0f28f
("signal: add block_sigmask() for adding sigmask to current->blocked")
which centralises the code for updating current->blocked after
successfully delivering a signal and reduces the amount of duplicate code
across architectures.  In the past some architectures got this code wrong,
so using this helper function should stop that from happening again.

Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoxtensa: don't mask signals if we fail to setup signal stack
Matt Fleming [Wed, 21 Mar 2012 23:33:45 +0000 (16:33 -0700)]
xtensa: don't mask signals if we fail to setup signal stack

setup_frame() needs to return an indication of whether it succeeded or
failed in setting up the signal stack frame.  If setup_frame() fails then
we must not modify current->blocked.

Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoxtensa: no need to reset handler if SA_ONESHOT
Matt Fleming [Wed, 21 Mar 2012 23:33:44 +0000 (16:33 -0700)]
xtensa: no need to reset handler if SA_ONESHOT

get_signal_to_deliver() already resets the signal handler if SA_ONESHOT
is set in ka->sa.sa_flags, there's no need to do it again in
handle_signal().

Furthermore, because we were modifying ka->sa.sa_handler (which is a
copy of sighand->action[]) instead of sighand->action[] the original
code actually had no effect on signal delivery.

Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoxtensa: don't reimplement force_sigsegv()
Matt Fleming [Wed, 21 Mar 2012 23:33:44 +0000 (16:33 -0700)]
xtensa: don't reimplement force_sigsegv()

Instead of open coding the sequence from force_sigsegv() just call it.
This also fixes a bug because we were modifying ka->sa.sa_handler (which
is a copy of sighand->action[]), whereas the intention of the code was to
modify sighand->action[] directly.

As the original code was working with a copy it had no effect on signal
delivery.

Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Chris Zankel <chris@zankel.net>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoseq_file: fix mishandling of consecutive pread() invocations.
Earl Chew [Wed, 21 Mar 2012 23:33:43 +0000 (16:33 -0700)]
seq_file: fix mishandling of consecutive pread() invocations.

The following program illustrates the problem:

    char buf[8192];

    int fd = open("/proc/self/maps", O_RDONLY);

    n = pread(fd, buf, sizeof(buf), 0);
    printf("%d\n", n);

    /* lseek(fd, 0, SEEK_CUR); */ /* Uncomment to work around */

    n = pread(fd, buf, sizeof(buf), 0);
    printf("%d\n", n);

The second printf() prints zero, but uncommenting the lseek() corrects its
behaviour.

To fix, make seq_read() mirror seq_lseek() when processing changes in
*ppos.  Restore m->version first, then if required traverse and update
read_pos on success.

Addresses https://bugzilla.kernel.org/show_bug.cgi?id=11856

Signed-off-by: Earl Chew <echew@ixiacom.com>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agodrivers/idle/intel_idle.c: fix confusing code identation
Marcos Paulo de Souza [Wed, 21 Mar 2012 23:33:43 +0000 (16:33 -0700)]
drivers/idle/intel_idle.c: fix confusing code identation

Fix a code indentation in the function intel_idle_cpu_init that looks
confusing.o

Suggested-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Reviewed-by: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Signed-off-by: Marcos Paulo de Souza <marcos.mage@gmail.com>
Cc: "Brown, Len" <len.brown@intel.com>
Cc: Len Brown <lenb@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agofs/namei.c: fix warnings on 32-bit
Andrew Morton [Wed, 21 Mar 2012 23:33:42 +0000 (16:33 -0700)]
fs/namei.c: fix warnings on 32-bit

i386 allnoconfig:

  fs/namei.c: In function 'has_zero':
  fs/namei.c:1617: warning: integer constant is too large for 'unsigned long' type
  fs/namei.c:1617: warning: integer constant is too large for 'unsigned long' type
  fs/namei.c: In function 'hash_name':
  fs/namei.c:1635: warning: integer constant is too large for 'unsigned long' type

There must be a tidier way of doing this.

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomm: thp: fix pmd_bad() triggering in code paths holding mmap_sem read mode
Andrea Arcangeli [Wed, 21 Mar 2012 23:33:42 +0000 (16:33 -0700)]
mm: thp: fix pmd_bad() triggering in code paths holding mmap_sem read mode

In some cases it may happen that pmd_none_or_clear_bad() is called with
the mmap_sem hold in read mode.  In those cases the huge page faults can
allocate hugepmds under pmd_none_or_clear_bad() and that can trigger a
false positive from pmd_bad() that will not like to see a pmd
materializing as trans huge.

It's not khugepaged causing the problem, khugepaged holds the mmap_sem
in write mode (and all those sites must hold the mmap_sem in read mode
to prevent pagetables to go away from under them, during code review it
seems vm86 mode on 32bit kernels requires that too unless it's
restricted to 1 thread per process or UP builds).  The race is only with
the huge pagefaults that can convert a pmd_none() into a
pmd_trans_huge().

Effectively all these pmd_none_or_clear_bad() sites running with
mmap_sem in read mode are somewhat speculative with the page faults, and
the result is always undefined when they run simultaneously.  This is
probably why it wasn't common to run into this.  For example if the
madvise(MADV_DONTNEED) runs zap_page_range() shortly before the page
fault, the hugepage will not be zapped, if the page fault runs first it
will be zapped.

Altering pmd_bad() not to error out if it finds hugepmds won't be enough
to fix this, because zap_pmd_range would then proceed to call
zap_pte_range (which would be incorrect if the pmd become a
pmd_trans_huge()).

The simplest way to fix this is to read the pmd in the local stack
(regardless of what we read, no need of actual CPU barriers, only
compiler barrier needed), and be sure it is not changing under the code
that computes its value.  Even if the real pmd is changing under the
value we hold on the stack, we don't care.  If we actually end up in
zap_pte_range it means the pmd was not none already and it was not huge,
and it can't become huge from under us (khugepaged locking explained
above).

All we need is to enforce that there is no way anymore that in a code
path like below, pmd_trans_huge can be false, but pmd_none_or_clear_bad
can run into a hugepmd.  The overhead of a barrier() is just a compiler
tweak and should not be measurable (I only added it for THP builds).  I
don't exclude different compiler versions may have prevented the race
too by caching the value of *pmd on the stack (that hasn't been
verified, but it wouldn't be impossible considering
pmd_none_or_clear_bad, pmd_bad, pmd_trans_huge, pmd_none are all inlines
and there's no external function called in between pmd_trans_huge and
pmd_none_or_clear_bad).

if (pmd_trans_huge(*pmd)) {
if (next-addr != HPAGE_PMD_SIZE) {
VM_BUG_ON(!rwsem_is_locked(&tlb->mm->mmap_sem));
split_huge_page_pmd(vma->vm_mm, pmd);
} else if (zap_huge_pmd(tlb, vma, pmd, addr))
continue;
/* fall through */
}
if (pmd_none_or_clear_bad(pmd))

Because this race condition could be exercised without special
privileges this was reported in CVE-2012-1179.

The race was identified and fully explained by Ulrich who debugged it.
I'm quoting his accurate explanation below, for reference.

====== start quote =======
      mapcount 0 page_mapcount 1
      kernel BUG at mm/huge_memory.c:1384!

    At some point prior to the panic, a "bad pmd ..." message similar to the
    following is logged on the console:

      mm/memory.c:145: bad pmd ffff8800376e1f98(80000000314000e7).

    The "bad pmd ..." message is logged by pmd_clear_bad() before it clears
    the page's PMD table entry.

        143 void pmd_clear_bad(pmd_t *pmd)
        144 {
    ->  145         pmd_ERROR(*pmd);
        146         pmd_clear(pmd);
        147 }

    After the PMD table entry has been cleared, there is an inconsistency
    between the actual number of PMD table entries that are mapping the page
    and the page's map count (_mapcount field in struct page). When the page
    is subsequently reclaimed, __split_huge_page() detects this inconsistency.

       1381         if (mapcount != page_mapcount(page))
       1382                 printk(KERN_ERR "mapcount %d page_mapcount %d\n",
       1383                        mapcount, page_mapcount(page));
    -> 1384         BUG_ON(mapcount != page_mapcount(page));

    The root cause of the problem is a race of two threads in a multithreaded
    process. Thread B incurs a page fault on a virtual address that has never
    been accessed (PMD entry is zero) while Thread A is executing an madvise()
    system call on a virtual address within the same 2 MB (huge page) range.

               virtual address space
              .---------------------.
              |                     |
              |                     |
            .-|---------------------|
            | |                     |
            | |                     |<-- B(fault)
            | |                     |
      2 MB  | |/////////////////////|-.
      huge <  |/////////////////////|  > A(range)
      page  | |/////////////////////|-'
            | |                     |
            | |                     |
            '-|---------------------|
              |                     |
              |                     |
              '---------------------'

    - Thread A is executing an madvise(..., MADV_DONTNEED) system call
      on the virtual address range "A(range)" shown in the picture.

    sys_madvise
      // Acquire the semaphore in shared mode.
      down_read(&current->mm->mmap_sem)
      ...
      madvise_vma
        switch (behavior)
        case MADV_DONTNEED:
             madvise_dontneed
               zap_page_range
                 unmap_vmas
                   unmap_page_range
                     zap_pud_range
                       zap_pmd_range
                         //
                         // Assume that this huge page has never been accessed.
                         // I.e. content of the PMD entry is zero (not mapped).
                         //
                         if (pmd_trans_huge(*pmd)) {
                             // We don't get here due to the above assumption.
                         }
                         //
                         // Assume that Thread B incurred a page fault and
             .---------> // sneaks in here as shown below.
             |           //
             |           if (pmd_none_or_clear_bad(pmd))
             |               {
             |                 if (unlikely(pmd_bad(*pmd)))
             |                     pmd_clear_bad
             |                     {
             |                       pmd_ERROR
             |                         // Log "bad pmd ..." message here.
             |                       pmd_clear
             |                         // Clear the page's PMD entry.
             |                         // Thread B incremented the map count
             |                         // in page_add_new_anon_rmap(), but
             |                         // now the page is no longer mapped
             |                         // by a PMD entry (-> inconsistency).
             |                     }
             |               }
             |
             v
    - Thread B is handling a page fault on virtual address "B(fault)" shown
      in the picture.

    ...
    do_page_fault
      __do_page_fault
        // Acquire the semaphore in shared mode.
        down_read_trylock(&mm->mmap_sem)
        ...
        handle_mm_fault
          if (pmd_none(*pmd) && transparent_hugepage_enabled(vma))
              // We get here due to the above assumption (PMD entry is zero).
              do_huge_pmd_anonymous_page
                alloc_hugepage_vma
                  // Allocate a new transparent huge page here.
                ...
                __do_huge_pmd_anonymous_page
                  ...
                  spin_lock(&mm->page_table_lock)
                  ...
                  page_add_new_anon_rmap
                    // Here we increment the page's map count (starts at -1).
                    atomic_set(&page->_mapcount, 0)
                  set_pmd_at
                    // Here we set the page's PMD entry which will be cleared
                    // when Thread A calls pmd_clear_bad().
                  ...
                  spin_unlock(&mm->page_table_lock)

    The mmap_sem does not prevent the race because both threads are acquiring
    it in shared mode (down_read).  Thread B holds the page_table_lock while
    the page's map count and PMD table entry are updated.  However, Thread A
    does not synchronize on that lock.

====== end quote =======

[akpm@linux-foundation.org: checkpatch fixes]
Reported-by: Ulrich Obergfell <uobergfe@redhat.com>
Signed-off-by: Andrea Arcangeli <aarcange@redhat.com>
Acked-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mgorman@suse.de>
Cc: Hugh Dickins <hughd@google.com>
Cc: Dave Jones <davej@redhat.com>
Acked-by: Larry Woodman <lwoodman@redhat.com>
Acked-by: Rik van Riel <riel@redhat.com>
Cc: <stable@vger.kernel.org> [2.6.38+]
Cc: Mark Salter <msalter@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck...
Linus Torvalds [Wed, 21 Mar 2012 17:37:25 +0000 (10:37 -0700)]
Merge tag 'hwmon-for-linus' of git://git./linux/kernel/git/groeck/linux-staging

Pull hwmon changes for v3.4 from Guenter Roeck:
 "Mostly cleanup.  No new drivers this time around, but support for
  several chips added to existing drivers: TPS40400, TPS40422, MTD040,
  MAX34446, ZL9101M, ZL9117M, and LM96080.  Also, added watchdog support
  for SCH56xx, and additional attributes for a couple of drivers."

* tag 'hwmon-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/groeck/linux-staging: (137 commits)
  hwmon: (sch56xx) Add support for the integrated watchdog (v2)
  hwmon: (w83627ehf) Add support for temperature offset registers
  hwmon: (jc42) Remove unnecessary device IDs
  hwmon: (zl6100) Add support for ZL9101M and ZL9117M
  hwmon: (adm1275) Add support for ADM1075
  hwmon: (max34440) Add support for MAX34446
  hwmon: (pmbus) Add more virtual registers
  hwmon: (pmbus) Add support for Lineage Power MDT040
  hwmon: (pmbus) Add support for TI TPS40400 and TPS40422
  hwmon: (max34440) Add support for 'lowest' output voltage attribute
  hwmon: (jc42) Convert to use devm_kzalloc
  hwmon: (max16065) Convert to use devm_kzalloc
  hwmon: (smm665) Convert to use devm_kzalloc
  hwmon: (ltc4261) Convert to use devm_kzalloc
  hwmon: (pmbus) Simplify remove functions
  hwmon: (pmbus) Convert pmbus drivers to use devm_kzalloc
  hwmon: (lineage-pem) Convert to use devm_kzalloc
  hwmon: (hwmon-vid) Fix checkpatch issues
  hwmon: (hwmon-vid) Add new entries to VRM model table
  hwmon: (lm80) Add detection of NatSemi/TI LM96080
  ...

12 years agoMerge tag 'regulator-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie...
Linus Torvalds [Wed, 21 Mar 2012 17:34:56 +0000 (10:34 -0700)]
Merge tag 'regulator-3.4' of git://git./linux/kernel/git/broonie/regulator

Pull regulator updates for 3.4 from Mark Brown:
 "This has been a fairly quiet release from a regulator point of view,
  the only real framework features added were devm support and a
  convenience helper for setting up fixed voltage regulators.

  We also added a couple of drivers (but will drop the BQ240022 driver
  via the arm-soc tree as it's been replaced by the more generic
  gpio-regulator driver) and Axel Lin continued his relentless and
  generally awesome stream of fixes and cleanups."

* tag 'regulator-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator: (93 commits)
  regulator: Fix up a confusing dev_warn when DT lookup fails
  regulator: Convert tps6507x to set_voltage_sel
  regulator: Refactor tps6507x to use one tps6507x_pmic_ops for all LDOs and DCDCs
  regulator: Make s5m8767_get_voltage_register always return correct register
  regulator: s5m8767: Check pdata->buck[2|3|4]_gpiodvs earlier
  regulator: tps65910: Provide settling time for DCDC voltage change
  regulator: Add Anatop regulator driver
  regulator: Simplify implementation of tps65912_get_voltage_dcdc
  regulator: Use tps65912_set_voltage_sel for both DCDCx and LDOx
  regulator: tps65910: Provide settling time for enabling rails
  regulator: max8925: Use DIV_ROUND_UP macro
  regulator: tps65912: Use simple equations to get register address
  regulator: Fix the logic of tps65910_get_mode
  regulator: Merge tps65217_pmic_ldo234_ops and tps65217_pmic_dcdc_ops to tps65217_pmic_ops
  regulator: Use DIV_ROUND_CLOSEST in wm8350_isink_get_current
  regulator: Use array to store dcdc_range settings for tps65912
  regulator: Rename s5m8767_convert_voltage to s5m8767_convert_voltage_to_sel
  regulator: tps6524x: Remove unneeded comment for N_REGULATORS
  regulator: Rename set_voltage_sel callback function name to *_sel
  regulator: Fix s5m8767_set_voltage_time_sel calculation value
  ...

12 years agoMerge tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland...
Linus Torvalds [Wed, 21 Mar 2012 17:33:42 +0000 (10:33 -0700)]
Merge tag 'rdma-for-linus' of git://git./linux/kernel/git/roland/infiniband

Pull InfiniBand/RDMA changes for the 3.4 merge window from Roland Dreier:
 "Nothing big really stands out; by patch count lots of fixes to the
  mlx4 driver plus some cleanups and fixes to the core and other
  drivers."

* tag 'rdma-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (28 commits)
  mlx4_core: Scale size of MTT table with system RAM
  mlx4_core: Allow dynamic MTU configuration for IB ports
  IB/mlx4: Fix info returned when querying IBoE ports
  IB/mlx4: Fix possible missed completion event
  mlx4_core: Report thermal error events
  mlx4_core: Fix one more static exported function
  IB: Change CQE "csum_ok" field to a bit flag
  RDMA/iwcm: Reject connect requests if cmid is not in LISTEN state
  RDMA/cxgb3: Don't pass irq flags to flush_qp()
  mlx4_core: Get rid of redundant ext_port_cap flags
  RDMA/ucma: Fix AB-BA deadlock
  IB/ehca: Fix ilog2() compile failure
  IB: Use central enum for speed instead of hard-coded values
  IB/iser: Post initial receive buffers before sending the final login request
  IB/iser: Free IB connection resources in the proper place
  IB/srp: Consolidate repetitive sysfs code
  IB/srp: Use pr_fmt() and pr_err()/pr_warn()
  IB/core: Fix SDR rates in sysfs
  mlx4: Enforce device max FMR maps in FMR alloc
  IB/mlx4: Set bad_wr for invalid send opcode
  ...

12 years agoMerge tag 'spi-for-linus' of git://git.secretlab.ca/git/linux-2.6
Linus Torvalds [Wed, 21 Mar 2012 17:32:00 +0000 (10:32 -0700)]
Merge tag 'spi-for-linus' of git://git.secretlab.ca/git/linux-2.6

Pull SPI changes for v3.4 from Grant Likely:
 "Mostly a bunch of new drivers and driver bug fixes; but this also
  includes a few patches that create a core message queue infrastructure
  for the spi subsystem instead of making each driver open code it."

* tag 'spi-for-linus' of git://git.secretlab.ca/git/linux-2.6: (34 commits)
  spi/fsl-espi: Make sure pm is within 2..32
  spi/fsl-espi: make the clock computation easier to read
  spi: sh-hspi: modify write/read method
  spi: sh-hspi: control spi clock more correctly
  spi: sh-hspi: convert to using core message queue
  spi: s3c64xx: Fix build
  spi: s3c64xx: remove unnecessary callback msg->complete
  spi: remove redundant variable assignment
  spi: release lock on error path in spi_pump_messages()
  spi: Compatibility with direction which is used in samsung DMA operation
  spi-topcliff-pch: add recovery processing in case wait-event timeout
  spi-topcliff-pch: supports a spi mode setup and bit order setup by IO control
  spi-topcliff-pch: Fix issue for transmitting over 4KByte
  spi-topcliff-pch: Modify pci-bus number dynamically to get DMA device info
  spi/imx: simplify error handling to free gpios
  spi: Convert to DEFINE_PCI_DEVICE_TABLE
  spi: add Broadcom BCM63xx SPI controller driver
  SPI: add CSR SiRFprimaII SPI controller driver
  spi-topcliff-pch: fix -Wuninitialized warning
  spi: Mark spi_register_board_info() __devinit
  ...

12 years agoMerge tag 'dt-for-linus' of git://git.secretlab.ca/git/linux-2.6
Linus Torvalds [Wed, 21 Mar 2012 17:30:03 +0000 (10:30 -0700)]
Merge tag 'dt-for-linus' of git://git.secretlab.ca/git/linux-2.6

Pull core device tree changes for Linux v3.4 from Grant Likely:
 "This branch contains a minor documentation addition, a utility
  function for parsing string properties needed by some of the new ARM
  platforms, disables dynamic DT code that isn't used anywhere but on a
  few PPC machines, and exports DT node compatible data to userspace via
  UEVENT properties.  Nothing earth shattering here."

* tag 'dt-for-linus' of git://git.secretlab.ca/git/linux-2.6:
  of: Only compile OF_DYNAMIC on PowerPC pseries and iseries
  arm/dts: OMAP3: Add omap3evm and am335xevm support
  drivercore: Output common devicetree information in uevent
  of: Add of_property_match_string() to find index into a string list

12 years agoMerge tag 'irqdomain-for-linus' of git://git.secretlab.ca/git/linux-2.6
Linus Torvalds [Wed, 21 Mar 2012 17:27:19 +0000 (10:27 -0700)]
Merge tag 'irqdomain-for-linus' of git://git.secretlab.ca/git/linux-2.6

Pull irq_domain support for all architectures from Grant Likely:
 "Generialize powerpc's irq_host as irq_domain

  This branch takes the PowerPC irq_host infrastructure (reverse mapping
  from Linux IRQ numbers to hardware irq numbering), generalizes it,
  renames it to irq_domain, and makes it available to all architectures.

  Originally the plan has been to create an all-new irq_domain
  implementation which addresses some of the powerpc shortcomings such
  as not handling 1:1 mappings well, but doing that proved to be far
  more difficult and invasive than generalizing the working code and
  refactoring it in-place.  So, this branch rips out the 'new'
  irq_domain and replaces it with the modified powerpc version (in a
  fully bisectable way of course).  It converts all users over to the
  new API and makes irq_domain selectable on any architecture.

  No architecture is forced to enable irq_domain, but the infrastructure
  is required for doing OpenFirmware style irq translations.  It will
  even work on SPARC even though SPARC has it's own mechanism for
  translating irqs at boot time.  MIPS, microblaze, embedded x86 and c6x
  are converted too.

  The resulting irq_domain code is probably still too verbose and can be
  optimized more, but that can be done incrementally and is a task for
  follow-on patches."

* tag 'irqdomain-for-linus' of git://git.secretlab.ca/git/linux-2.6: (31 commits)
  dt: fix twl4030 for non-dt compile on x86
  mfd: twl-core: Add IRQ_DOMAIN dependency
  devicetree: Add empty of_platform_populate() for !CONFIG_OF_ADDRESS (sparc)
  irq_domain: Centralize definition of irq_dispose_mapping()
  irq_domain/mips: Allow irq_domain on MIPS
  irq_domain/x86: Convert x86 (embedded) to use common irq_domain
  ppc-6xx: fix build failure in flipper-pic.c and hlwd-pic.c
  irq_domain/microblaze: Convert microblaze to use irq_domains
  irq_domain/powerpc: Replace custom xlate functions with library functions
  irq_domain/powerpc: constify irq_domain_ops
  irq_domain/c6x: Use library of xlate functions
  irq_domain/c6x: constify irq_domain structures
  irq_domain/c6x: Convert c6x to use generic irq_domain support.
  irq_domain: constify irq_domain_ops
  irq_domain: Create common xlate functions that device drivers can use
  irq_domain: Remove irq_domain_add_simple()
  irq_domain: Remove 'new' irq_domain in favour of the ppc one
  mfd: twl-core.c: Fix the number of interrupts managed by twl4030
  of/address: add empty static inlines for !CONFIG_OF
  irq_domain: Add support for base irq and hwirq in legacy mappings
  ...

12 years agoMerge tag 'pm-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Wed, 21 Mar 2012 17:15:51 +0000 (10:15 -0700)]
Merge tag 'pm-for-3.4' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management updates for 3.4 from Rafael Wysocki:
 "Assorted extensions and fixes including:

  * Introduction of early/late suspend/hibernation device callbacks.
  * Generic PM domains extensions and fixes.
  * devfreq updates from Axel Lin and MyungJoo Ham.
  * Device PM QoS updates.
  * Fixes of concurrency problems with wakeup sources.
  * System suspend and hibernation fixes."

* tag 'pm-for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm: (43 commits)
  PM / Domains: Check domain status during hibernation restore of devices
  PM / devfreq: add relation of recommended frequency.
  PM / shmobile: Make MTU2 driver use pm_genpd_dev_always_on()
  PM / shmobile: Make CMT driver use pm_genpd_dev_always_on()
  PM / shmobile: Make TMU driver use pm_genpd_dev_always_on()
  PM / Domains: Introduce "always on" device flag
  PM / Domains: Fix hibernation restore of devices, v2
  PM / Domains: Fix handling of wakeup devices during system resume
  sh_mmcif / PM: Use PM QoS latency constraint
  tmio_mmc / PM: Use PM QoS latency constraint
  PM / QoS: Make it possible to expose PM QoS latency constraints
  PM / Sleep: JBD and JBD2 missing set_freezable()
  PM / Domains: Fix include for PM_GENERIC_DOMAINS=n case
  PM / Freezer: Remove references to TIF_FREEZE in comments
  PM / Sleep: Add more wakeup source initialization routines
  PM / Hibernate: Enable usermodehelpers in hibernate() error path
  PM / Sleep: Make __pm_stay_awake() delete wakeup source timers
  PM / Sleep: Fix race conditions related to wakeup source timer function
  PM / Sleep: Fix possible infinite loop during wakeup source destruction
  PM / Hibernate: print physical addresses consistently with other parts of kernel
  ...

12 years agoMerge branch 'kmap_atomic' of git://github.com/congwang/linux
Linus Torvalds [Wed, 21 Mar 2012 16:40:26 +0000 (09:40 -0700)]
Merge branch 'kmap_atomic' of git://github.com/congwang/linux

Pull kmap_atomic cleanup from Cong Wang.

It's been in -next for a long time, and it gets rid of the (no longer
used) second argument to k[un]map_atomic().

Fix up a few trivial conflicts in various drivers, and do an "evil
merge" to catch some new uses that have come in since Cong's tree.

* 'kmap_atomic' of git://github.com/congwang/linux: (59 commits)
  feature-removal-schedule.txt: schedule the deprecated form of kmap_atomic() for removal
  highmem: kill all __kmap_atomic() [swarren@nvidia.com: highmem: Fix ARM build break due to __kmap_atomic rename]
  drbd: remove the second argument of k[un]map_atomic()
  zcache: remove the second argument of k[un]map_atomic()
  gma500: remove the second argument of k[un]map_atomic()
  dm: remove the second argument of k[un]map_atomic()
  tomoyo: remove the second argument of k[un]map_atomic()
  sunrpc: remove the second argument of k[un]map_atomic()
  rds: remove the second argument of k[un]map_atomic()
  net: remove the second argument of k[un]map_atomic()
  mm: remove the second argument of k[un]map_atomic()
  lib: remove the second argument of k[un]map_atomic()
  power: remove the second argument of k[un]map_atomic()
  kdb: remove the second argument of k[un]map_atomic()
  udf: remove the second argument of k[un]map_atomic()
  ubifs: remove the second argument of k[un]map_atomic()
  squashfs: remove the second argument of k[un]map_atomic()
  reiserfs: remove the second argument of k[un]map_atomic()
  ocfs2: remove the second argument of k[un]map_atomic()
  ntfs: remove the second argument of k[un]map_atomic()
  ...

12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial
Linus Torvalds [Wed, 21 Mar 2012 04:12:50 +0000 (21:12 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/trivial

Pull trivial tree from Jiri Kosina:
 "It's indeed trivial -- mostly documentation updates and a bunch of
  typo fixes from Masanari.

  There are also several linux/version.h include removals from Jesper."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial: (101 commits)
  kcore: fix spelling in read_kcore() comment
  constify struct pci_dev * in obvious cases
  Revert "char: Fix typo in viotape.c"
  init: fix wording error in mm_init comment
  usb: gadget: Kconfig: fix typo for 'different'
  Revert "power, max8998: Include linux/module.h just once in drivers/power/max8998_charger.c"
  writeback: fix fn name in writeback_inodes_sb_nr_if_idle() comment header
  writeback: fix typo in the writeback_control comment
  Documentation: Fix multiple typo in Documentation
  tpm_tis: fix tis_lock with respect to RCU
  Revert "media: Fix typo in mixer_drv.c and hdmi_drv.c"
  Doc: Update numastat.txt
  qla4xxx: Add missing spaces to error messages
  compiler.h: Fix typo
  security: struct security_operations kerneldoc fix
  Documentation: broken URL in libata.tmpl
  Documentation: broken URL in filesystems.tmpl
  mtd: simplify return logic in do_map_probe()
  mm: fix comment typo of truncate_inode_pages_range
  power: bq27x00: Fix typos in comment
  ...

12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid
Linus Torvalds [Wed, 21 Mar 2012 04:11:42 +0000 (21:11 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jikos/hid

Pull HID updates from Jiri Kosina:
 "It contains HID driver updates all over the place -- a lot of new
  hardware support especially in the multitouch area, including generic
  handling of all multitouch devices by the hid-multitiouch driver
  automatically."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid: (42 commits)
  HID: multitouch: add PID for Fructel product
  HID: wacom: Add reporting of wheel for Intuos4 WL
  HID: wacom: Replace __set_bit with input_set_capability
  HID: tivo: add support for BT-version (0x1200)
  HID: wacom: Reset stylus buttons - Intuos4 WL
  HID: multitouch: detect serial protocol
  HID: handle all multitouch devices through hid-multitouch
  HID: multitouch: fix handling of buggy reports descriptors for Dell ST2220T
  HID: make it possible to force hid-core claim the device
  HID: multitouch: add support for eGalax 0x722a
  HID: usbhid: add quirk no_get for quanta 3008 devices
  HID: multitouch: add more eGalax devices
  HID: multitouch: add new PID from Ideacom
  HID: multitouch: add support for Atmel maXTouch 03eb:2118
  HID: waltop: Add support for tablet with PID 0038
  HID: waltop: Replace original rdescs with links
  HID: uclogic: Replace original rdescs with links
  HID: wacom: Add pad buttons reporting on Intuos4 WL
  HID: wacom: report distance for Intuos4 WL
  HID: kye: Add support for 3 tablets
  ...

12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next
Linus Torvalds [Wed, 21 Mar 2012 04:04:47 +0000 (21:04 -0700)]
Merge git://git./linux/kernel/git/davem/net-next

Pull networking merge from David Miller:
 "1) Move ixgbe driver over to purely page based buffering on receive.
     From Alexander Duyck.

  2) Add receive packet steering support to e1000e, from Bruce Allan.

  3) Convert TCP MD5 support over to RCU, from Eric Dumazet.

  4) Reduce cpu usage in handling out-of-order TCP packets on modern
     systems, also from Eric Dumazet.

  5) Support the IP{,V6}_UNICAST_IF socket options, making the wine
     folks happy, from Erich Hoover.

  6) Support VLAN trunking from guests in hyperv driver, from Haiyang
     Zhang.

  7) Support byte-queue-limtis in r8169, from Igor Maravic.

  8) Outline code intended for IP_RECVTOS in IP_PKTOPTIONS existed but
     was never properly implemented, Jiri Benc fixed that.

  9) 64-bit statistics support in r8169 and 8139too, from Junchang Wang.

  10) Support kernel side dump filtering by ctmark in netfilter
      ctnetlink, from Pablo Neira Ayuso.

  11) Support byte-queue-limits in gianfar driver, from Paul Gortmaker.

  12) Add new peek socket options to assist with socket migration, from
      Pavel Emelyanov.

  13) Add sch_plug packet scheduler whose queue is controlled by
      userland daemons using explicit freeze and release commands.  From
      Shriram Rajagopalan.

  14) Fix FCOE checksum offload handling on transmit, from Yi Zou."

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-next: (1846 commits)
  Fix pppol2tp getsockname()
  Remove printk from rds_sendmsg
  ipv6: fix incorrent ipv6 ipsec packet fragment
  cpsw: Hook up default ndo_change_mtu.
  net: qmi_wwan: fix build error due to cdc-wdm dependecy
  netdev: driver: ethernet: Add TI CPSW driver
  netdev: driver: ethernet: add cpsw address lookup engine support
  phy: add am79c874 PHY support
  mlx4_core: fix race on comm channel
  bonding: send igmp report for its master
  fs_enet: Add MPC5125 FEC support and PHY interface selection
  net: bpf_jit: fix BPF_S_LDX_B_MSH compilation
  net: update the usage of CHECKSUM_UNNECESSARY
  fcoe: use CHECKSUM_UNNECESSARY instead of CHECKSUM_PARTIAL on tx
  net: do not do gso for CHECKSUM_UNNECESSARY in netif_needs_gso
  ixgbe: Fix issues with SR-IOV loopback when flow control is disabled
  net/hyperv: Fix the code handling tx busy
  ixgbe: fix namespace issues when FCoE/DCB is not enabled
  rtlwifi: Remove unused ETH_ADDR_LEN defines
  igbvf: Use ETH_ALEN
  ...

Fix up fairly trivial conflicts in drivers/isdn/gigaset/interface.c and
drivers/net/usb/{Kconfig,qmi_wwan.c} as per David.

12 years agoMerge branch 'for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
Linus Torvalds [Wed, 21 Mar 2012 01:13:22 +0000 (18:13 -0700)]
Merge branch 'for-3.4' of git://git./linux/kernel/git/tj/wq

Pull workqueue changes from Tejun Heo:
 "This contains only one commit which cleans up UP allocation path a
  bit."

* 'for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
  workqueue: use percpu allocator for cwq on UP

12 years agoMerge branch 'for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup
Linus Torvalds [Wed, 21 Mar 2012 01:11:21 +0000 (18:11 -0700)]
Merge branch 'for-3.4' of git://git./linux/kernel/git/tj/cgroup

Pull cgroup changes from Tejun Heo:
 "Out of the 8 commits, one fixes a long-standing locking issue around
  tasklist walking and others are cleanups."

* 'for-3.4' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  cgroup: Walk task list under tasklist_lock in cgroup_enable_task_cg_list
  cgroup: Remove wrong comment on cgroup_enable_task_cg_list()
  cgroup: remove cgroup_subsys argument from callbacks
  cgroup: remove extra calls to find_existing_css_set
  cgroup: replace tasklist_lock with rcu_read_lock
  cgroup: simplify double-check locking in cgroup_attach_proc
  cgroup: move struct cgroup_pidlist out from the header file
  cgroup: remove cgroup_attach_task_current_cg()

12 years agoexec: move de_thread()->setmax_mm_hiwater_rss() into exec_mmap()
Oleg Nesterov [Mon, 19 Mar 2012 16:04:01 +0000 (17:04 +0100)]
exec: move de_thread()->setmax_mm_hiwater_rss() into exec_mmap()

Minor cleanup. de_thread()->setmax_mm_hiwater_rss() looks a bit
strange, move it into exec_mmap() which plays with old_mm.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoexit_signal: fix the "parent has changed security domain" logic
Oleg Nesterov [Mon, 19 Mar 2012 16:03:41 +0000 (17:03 +0100)]
exit_signal: fix the "parent has changed security domain" logic

exit_notify() changes ->exit_signal if the parent already did exec.
This doesn't really work, we are not going to send the signal now
if there is another live thread or the exiting task is traced. The
parent can exec before the last dies or the tracer detaches.

Move this check into do_notify_parent() which actually sends the
signal.

The user-visible change is that we do not change ->exit_signal,
and thus the exiting task is still "clone children" for
do_wait()->eligible_child(__WCLONE). Hopefully this is fine, the
current logic is racy anyway.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoexit_signal: simplify the "we have changed execution domain" logic
Oleg Nesterov [Mon, 19 Mar 2012 16:03:22 +0000 (17:03 +0100)]
exit_signal: simplify the "we have changed execution domain" logic

exit_notify() checks "tsk->self_exec_id != tsk->parent_exec_id"
to handle the "we have changed execution domain" case.

We can change do_thread() to always set ->exit_signal = SIGCHLD
and remove this check to simplify the code.

We could change setup_new_exec() instead, this looks more logical
because it increments ->self_exec_id. But note that de_thread()
already resets ->exit_signal if it changes the leader, let's keep
both changes close to each other.

Note that we change ->exit_signal lockless, this changes the rules.
Thereafter ->exit_signal is not stable under tasklist but this is
fine, the only possible change is OLDSIG -> SIGCHLD. This can race
with eligible_child() but the race is harmless. We can race with
reparent_leader() which changes our ->exit_signal in parallel, but
it does the same change to SIGCHLD.

The noticeable user-visible change is that the execing task is not
"visible" to do_wait()->eligible_child(__WCLONE) right after exec.
To me this looks more logical, and this is consistent with mt case.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoCLONE_PARENT shouldn't allow to set ->exit_signal
Oleg Nesterov [Wed, 14 Mar 2012 18:55:38 +0000 (19:55 +0100)]
CLONE_PARENT shouldn't allow to set ->exit_signal

The child must not control its ->exit_signal, it is the parent who
decides which signal the child should use for notification.

This means that CLONE_PARENT should not use "clone_flags & CSIGNAL",
the forking task is the sibling of the new process and their parent
doesn't control exit_signal in this case.

This patch uses ->exit_signal of the forking process, but perhaps
we should simply use SIGCHLD.

We read group_leader->exit_signal lockless, this can race with the
ORIGINAL_SIGNAL -> SIGCHLD transition, but this is fine.

Potentially this change allows to kill self_exec_id/parent_exec_id.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoFix pppol2tp getsockname()
Benjamin LaHaise [Tue, 20 Mar 2012 03:57:54 +0000 (03:57 +0000)]
Fix pppol2tp getsockname()

While testing L2TP functionality, I came across a bug in getsockname().  The
IP address returned within the pppol2tp_addr's addr memember was not being
set to the IP  address in use.  This bug is caused by using inet_sk() on the
wrong socket (the L2TP socket rather than the underlying UDP socket), and was
likely introduced during the addition of L2TPv3 support.

Signed-off-by: Benjamin LaHaise <bcrl@kvack.org>
Signed-off-by: James Chapman <jchapman@katalix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoRemove printk from rds_sendmsg
Dave Jones [Mon, 19 Mar 2012 13:01:07 +0000 (13:01 +0000)]
Remove printk from rds_sendmsg

no socket layer outputs a message for this error and neither should rds.

Signed-off-by: Dave Jones <davej@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoMerge tag 'usb-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Tue, 20 Mar 2012 18:26:30 +0000 (11:26 -0700)]
Merge tag 'usb-3.3' of git://git./linux/kernel/git/gregkh/usb

Pull USB merge for 3.4-rc1 from Greg KH:
 "Here's the big USB merge for the 3.4-rc1 merge window.

  Lots of gadget driver reworks here, driver updates, xhci changes, some
  new drivers added, usb-serial core reworking to fix some bugs, and
  other various minor things.

  There are some patches touching arch code, but they have all been
  acked by the various arch maintainers."

* tag 'usb-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (302 commits)
  net: qmi_wwan: add support for ZTE MF820D
  USB: option: add ZTE MF820D
  usb: gadget: f_fs: Remove lock is held before freeing checks
  USB: option: make interface blacklist work again
  usb/ub: deprecate & schedule for removal the "Low Performance USB Block" driver
  USB: ohci-pxa27x: add clk_prepare/clk_unprepare calls
  USB: use generic platform driver on ath79
  USB: EHCI: Add a generic platform device driver
  USB: OHCI: Add a generic platform device driver
  USB: ftdi_sio: new PID: LUMEL PD12
  USB: ftdi_sio: add support for FT-X series devices
  USB: serial: mos7840: Fixed MCS7820 device attach problem
  usb: Don't make USB_ARCH_HAS_{XHCI,OHCI,EHCI} depend on USB_SUPPORT.
  usb gadget: fix a section mismatch when compiling g_ffs with CONFIG_USB_FUNCTIONFS_ETH
  USB: ohci-nxp: Remove i2c_write(), use smbus
  USB: ohci-nxp: Support for LPC32xx
  USB: ohci-nxp: Rename symbols from pnx4008 to nxp
  USB: OHCI-HCD: Rename ohci-pnx4008 to ohci-nxp
  usb: gadget: Kconfig: fix typo for 'different'
  usb: dwc3: pci: fix another failure path in dwc3_pci_probe()
  ...

12 years agoMerge tag 'tty-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Tue, 20 Mar 2012 18:24:39 +0000 (11:24 -0700)]
Merge tag 'tty-3.3' of git://git./linux/kernel/git/gregkh/tty

Pull TTY/serial patches from Greg KH:
 "tty and serial merge for 3.4-rc1

  Here's the big serial and tty merge for the 3.4-rc1 tree.

  There's loads of fixes and reworks in here from Jiri for the tty
  layer, and a number of patches from Alan to help try to wrestle the vt
  layer into a sane model.

  Other than that, lots of driver updates and fixes, and other minor
  stuff, all detailed in the shortlog."

* tag 'tty-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty: (132 commits)
  serial: pxa: add clk_prepare/clk_unprepare calls
  TTY: Wrong unicode value copied in con_set_unimap()
  serial: PL011: clear pending interrupts
  serial: bfin-uart: Don't access tty circular buffer in TX DMA interrupt after it is reset.
  vt: NULL dereference in vt_do_kdsk_ioctl()
  tty: serial: vt8500: fix annotations for probe/remove
  serial: remove back and forth conversions in serial_out_sync
  serial: use serial_port_in/out vs serial_in/out in 8250
  serial: introduce generic port in/out helpers
  serial: reduce number of indirections in 8250 code
  serial: delete useless void casts in 8250.c
  serial: make 8250's serial_in shareable to other drivers.
  serial: delete last unused traces of pausing I/O in 8250
  pch_uart: Add module parameter descriptions
  pch_uart: Use existing default_baud in setup_console
  pch_uart: Add user_uartclk parameter
  pch_uart: Add Fish River Island II uart clock quirks
  pch_uart: Use uartclk instead of base_baud
  mpc5200b/uart: select more tolerant uart prescaler on low baudrates
  tty: moxa: fix bit test in moxa_start()
  ...

12 years agoMerge tag 'staging-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging
Linus Torvalds [Tue, 20 Mar 2012 18:23:18 +0000 (11:23 -0700)]
Merge tag 'staging-3.3' of git://git./linux/kernel/git/gregkh/staging

Pull big staging driver updates from Greg KH:
 "Here is the big drivers/staging/ merge for 3.4-rc1

  Lots of new driver updates here, with the addition of a few new ones,
  and only one moving out of the staging tree to the "real" part of the
  kernel (the hyperv scsi driver, acked by the scsi maintainer).

  There are also loads of cleanups, fixes, and other minor things in
  here, all self-contained in the drivers/staging/ tree.

  Overall we reversed the recent trend by adding more lines than we
  removed:
   379 files changed, 37952 insertions(+), 14153 deletions(-)"

* tag 'staging-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging: (360 commits)
  staging/zmem: Use lockdep_assert_held instead of spin_is_locked
  Staging: rtl8187se: r8180_wx.c: Cleaned up comments
  Staging: rtl8187se: r8180_wx.c: Removed old comments
  Staging: rtl8187se: r8180_dm.c: Removed old comments
  Staging: android: ram_console.c:
  Staging: rtl8187se: r8180_dm.c: Fix comments
  Staging: rtl8187se: r8180_dm.c: Fix spacing issues
  Staging: rtl8187se: r8180_dm.c Fixed indentation issues
  Staging: rtl8187se: r8180_dm.c: Fix brackets
  Staging: rtl8187se: r8180_dm.c: Removed spaces before tab stop
  staging: vme: fix section mismatches in linux-next 20120314
  Staging: rtl8187se: r8180_core.c: Fix some long line issues
  Staging: rtl8187se: r8180_core.c: Fix some spacing issues
  Staging: rtl8187se: r8180_core.c: Removed trailing spaces
  staging: mei: remove driver internal versioning
  Staging: rtl8187se: r8180_core.c: Cleaned up if statement
  staging: ozwpan depends on NET
  staging: ozwpan: added maintainer for ozwpan driver
  staging/mei: propagate error codes up in the write flow
  drivers:staging:mei Fix some typos in staging/mei
  ...

12 years agoMerge tag 'driver-core-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Tue, 20 Mar 2012 18:16:20 +0000 (11:16 -0700)]
Merge tag 'driver-core-3.3' of git://git./linux/kernel/git/gregkh/driver-core

Pull driver core patches for 3.4-rc1 from Greg KH:
 "Here's the big driver core merge for 3.4-rc1.

  Lots of various things here, sysfs fixes/tweaks (with the nlink
  breakage reverted), dynamic debugging updates, w1 drivers, hyperv
  driver updates, and a variety of other bits and pieces, full
  information in the shortlog."

* tag 'driver-core-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/driver-core: (78 commits)
  Tools: hv: Support enumeration from all the pools
  Tools: hv: Fully support the new KVP verbs in the user level daemon
  Drivers: hv: Support the newly introduced KVP messages in the driver
  Drivers: hv: Add new message types to enhance KVP
  regulator: Support driver probe deferral
  Revert "sysfs: Kill nlink counting."
  uevent: send events in correct order according to seqnum (v3)
  driver core: minor comment formatting cleanups
  driver core: move the deferred probe pointer into the private area
  drivercore: Add driver probe deferral mechanism
  DS2781 Maxim Stand-Alone Fuel Gauge battery and w1 slave drivers
  w1_bq27000: Only one thread can access the bq27000 at a time.
  w1_bq27000 - remove w1_bq27000_write
  w1_bq27000: remove unnecessary NULL test.
  sysfs: Fix memory leak in sysfs_sd_setsecdata().
  intel_idle: Revert change of auto_demotion_disable_flags for Nehalem
  w1: Fix w1_bq27000
  driver-core: documentation: fix up Greg's email address
  powernow-k6: Really enable auto-loading
  powernow-k7: Fix CPU family number
  ...

12 years agoMerge tag 'char-misc-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Tue, 20 Mar 2012 18:15:18 +0000 (11:15 -0700)]
Merge tag 'char-misc-3.3' of git://git./linux/kernel/git/gregkh/char-misc

Pull char and misc patches for 3.4-rc1 from Greg KH:
 "Not much here, just a few minor fixes and some conversions to the
  module_*_driver() functions, making the codebase smaller."

* tag 'char-misc-3.3' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  misc: bmp085: Use unsigned long to store jiffies
  char/ramoops: included linux/err.h twice
  misc: bmp085: Handle jiffies overflow correctly
  misc: fsa9480: Remove obsolete cleanup for clientdata
  char: Fix typo in tlclk.c
  char: Fix typo in viotape.c
  cs5535-mfgpt: don't call __init function from __devinit
  MISC: convert drivers/misc/* to use module_spi_driver()
  MISC: convert drivers/misc/* to use module_i2c_driver()
  MISC: convert drivers/misc/* to use module_platform_driver()

12 years agoAFS: checking wrong bit in afs_readpages()
Dan Carpenter [Tue, 20 Mar 2012 16:58:06 +0000 (16:58 +0000)]
AFS: checking wrong bit in afs_readpages()

We should be testing "if (vnode->flags & (1 << 4))" instead of
"if (vnode->flags & 4) {".  The current test checks if the data was
modified instead of deleted.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agoMerge branch 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 20 Mar 2012 17:32:09 +0000 (10:32 -0700)]
Merge branch 'timers-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer changes for v3.4 from Ingo Molnar

* 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (32 commits)
  ntp: Fix integer overflow when setting time
  math: Introduce div64_long
  cs5535-clockevt: Allow the MFGPT IRQ to be shared
  cs5535-clockevt: Don't ignore MFGPT on SMP-capable kernels
  x86/time: Eliminate unused irq0_irqs counter
  clocksource: scx200_hrt: Fix the build
  x86/tsc: Reduce the TSC sync check time for core-siblings
  timer: Fix bad idle check on irq entry
  nohz: Remove ts->Einidle checks before restarting the tick
  nohz: Remove update_ts_time_stat from tick_nohz_start_idle
  clockevents: Leave the broadcast device in shutdown mode when not needed
  clocksource: Load the ACPI PM clocksource asynchronously
  clocksource: scx200_hrt: Convert scx200 to use clocksource_register_hz
  clocksource: Get rid of clocksource_calc_mult_shift()
  clocksource: dbx500: convert to clocksource_register_hz()
  clocksource: scx200_hrt:  use pr_<level> instead of printk
  time: Move common updates to a function
  time: Reorder so the hot data is together
  time: Remove most of xtime_lock usage in timekeeping.c
  ntp: Add ntp_lock to replace xtime_locking
  ...

12 years agoMerge branch 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Tue, 20 Mar 2012 17:31:44 +0000 (10:31 -0700)]
Merge branch 'sched-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler changes for v3.4 from Ingo Molnar

* 'sched-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (27 commits)
  printk: Make it compile with !CONFIG_PRINTK
  sched/x86: Fix overflow in cyc2ns_offset
  sched: Fix nohz load accounting -- again!
  sched: Update yield() docs
  printk/sched: Introduce special printk_sched() for those awkward moments
  sched/nohz: Correctly initialize 'next_balance' in 'nohz' idle balancer
  sched: Cleanup cpu_active madness
  sched: Fix load-balance wreckage
  sched: Clean up parameter passing of proc_sched_autogroup_set_nice()
  sched: Ditch per cgroup task lists for load-balancing
  sched: Rename load-balancing fields
  sched: Move load-balancing arguments into helper struct
  sched/rt: Do not submit new work when PI-blocked
  sched/rt: Prevent idle task boosting
  sched/wait: Add __wake_up_all_locked() API
  sched/rt: Document scheduler related skip-resched-check sites
  sched/rt: Use schedule_preempt_disabled()
  sched/rt: Add schedule_preempt_disabled()
  sched/rt: Do not throttle when PI boosting
  sched/rt: Keep period timer ticking when rt throttling is active
  ...

12 years agoMerge branch 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 20 Mar 2012 17:29:15 +0000 (10:29 -0700)]
Merge branch 'perf-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf events changes for v3.4 from Ingo Molnar:

 - New "hardware based branch profiling" feature both on the kernel and
   the tooling side, on CPUs that support it.  (modern x86 Intel CPUs
   with the 'LBR' hardware feature currently.)

   This new feature is basically a sophisticated 'magnifying glass' for
   branch execution - something that is pretty difficult to extract from
   regular, function histogram centric profiles.

   The simplest mode is activated via 'perf record -b', and the result
   looks like this in perf report:

$ perf record -b any_call,u -e cycles:u branchy

$ perf report -b --sort=symbol
    52.34%  [.] main                   [.] f1
    24.04%  [.] f1                     [.] f3
    23.60%  [.] f1                     [.] f2
     0.01%  [k] _IO_new_file_xsputn    [k] _IO_file_overflow
     0.01%  [k] _IO_vfprintf_internal  [k] _IO_new_file_xsputn
     0.01%  [k] _IO_vfprintf_internal  [k] strchrnul
     0.01%  [k] __printf               [k] _IO_vfprintf_internal
     0.01%  [k] main                   [k] __printf

   This output shows from/to branch columns and shows the highest
   percentage (from,to) jump combinations - i.e.  the most likely taken
   branches in the system.  "branches" can also include function calls
   and any other synchronous and asynchronous transitions of the
   instruction pointer that are not 'next instruction' - such as system
   calls, traps, interrupts, etc.

   This feature comes with (hopefully intuitive) flat ascii and TUI
   support in perf report.

 - Various 'perf annotate' visual improvements for us assembly junkies.
   It will now recognize function calls in the TUI and by hitting enter
   you can follow the call (recursively) and back, amongst other
   improvements.

 - Multiple threads/processes recording support in perf record, perf
   stat, perf top - which is activated via a comma-list of PIDs:

perf top -p 21483,21485
perf stat -p 21483,21485 -ddd
perf record -p 21483,21485

 - Support for per UID views, via the --uid paramter to perf top, perf
   report, etc.  For example 'perf top --uid mingo' will only show the
   tasks that I am running, excluding other users, root, etc.

 - Jump label restructurings and improvements - this includes the
   factoring out of the (hopefully much clearer) include/linux/static_key.h
   generic facility:

struct static_key key = STATIC_KEY_INIT_FALSE;

...

if (static_key_false(&key))
        do unlikely code
else
        do likely code

...
static_key_slow_inc();
...
static_key_slow_inc();
...

   The static_key_false() branch will be generated into the code with as
   little impact to the likely code path as possible.  the
   static_key_slow_*() APIs flip the branch via live kernel code patching.

   This facility can now be used more widely within the kernel to
   micro-optimize hot branches whose likelihood matches the static-key
   usage and fast/slow cost patterns.

 - SW function tracer improvements: perf support and filtering support.

 - Various hardenings of the perf.data ABI, to make older perf.data's
   smoother on newer tool versions, to make new features integrate more
   smoothly, to support cross-endian recording/analyzing workflows
   better, etc.

 - Restructuring of the kprobes code, the splitting out of 'optprobes',
   and a corner case bugfix.

 - Allow the tracing of kernel console output (printk).

 - Improvements/fixes to user-space RDPMC support, allowing user-space
   self-profiling code to extract PMU counts without performing any
   system calls, while playing nice with the kernel side.

 - 'perf bench' improvements

 - ... and lots of internal restructurings, cleanups and fixes that made
   these features possible.  And, as usual this list is incomplete as
   there were also lots of other improvements

* 'perf-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (120 commits)
  perf report: Fix annotate double quit issue in branch view mode
  perf report: Remove duplicate annotate choice in branch view mode
  perf/x86: Prettify pmu config literals
  perf report: Enable TUI in branch view mode
  perf report: Auto-detect branch stack sampling mode
  perf record: Add HEADER_BRANCH_STACK tag
  perf record: Provide default branch stack sampling mode option
  perf tools: Make perf able to read files from older ABIs
  perf tools: Fix ABI compatibility bug in print_event_desc()
  perf tools: Enable reading of perf.data files from different ABI rev
  perf: Add ABI reference sizes
  perf report: Add support for taken branch sampling
  perf record: Add support for sampling taken branch
  perf tools: Add code to support PERF_SAMPLE_BRANCH_STACK
  x86/kprobes: Split out optprobe related code to kprobes-opt.c
  x86/kprobes: Fix a bug which can modify kernel code permanently
  x86/kprobes: Fix instruction recovery on optimized path
  perf: Add callback to flush branch_stack on context switch
  perf: Disable PERF_SAMPLE_BRANCH_* when not supported
  perf/x86: Add LBR software filter support for Intel CPUs
  ...

12 years agoMerge branch 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 20 Mar 2012 17:28:56 +0000 (10:28 -0700)]
Merge branch 'irq-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq/core changes for v3.4 from Ingo Molnar

* 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Remove paranoid warnons and bogus fixups
  genirq: Flush the irq thread on synchronization
  genirq: Get rid of unnecessary IRQTF_DIED flag
  genirq: No need to check IRQTF_DIED before stopping a thread handler
  genirq: Get rid of unnecessary irqaction field in task_struct
  genirq: Fix incorrect check for forced IRQ thread handler
  softirq: Reduce invoke_softirq() code duplication
  genirq: Fix long-term regression in genirq irq_set_irq_type() handling
  x86-32/irq: Don't switch to irq stack for a user-mode irq