memory cgroup enhancements: force_empty interface for dropping all account in empty cgroup

This patch adds an interface "memory.force_empty".  Any write to this file
will drop all charges in this cgroup if there is no task under.

%echo 1 > /....../memory.force_empty

will drop all charges of memory cgroup if cgroup's tasks is empty.

This is useful to invoke rmdir() against memory cgroup successfully.

Tested and worked well on x86_64/fake-NUMA system.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memory cgroup enhancements: fix zone handling in try_to_free_mem_cgroup_page

Because NODE_DATA(node)->node_zonelists[] is guaranteed to contain all
necessary zones, it is not necessary to use for_each_online_node.

And this for_each_online_node() makes reclaim routine start always
from node 0. This is not good. This patch makes reclaim start from
caller's node and just use usual (default) zonelist order.

[akpm@linux-foundation.org: fix warning]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcgroup: revert swap_state mods

If we're charging rss and we're charging cache, it seems obvious that we
should be charging swapcache - as has been done.  But in practice that
doesn't work out so well: both swapin readahead and swapoff leave the
majority of pages charged to the wrong cgroup (the cgroup that happened to
read them in, rather than the cgroup to which they belong).

(Which is why unuse_pte's GFP_KERNEL while holding pte lock never showed up
as a problem: no allocation was ever done there, every page read being
already charged to the cgroup which initiated the swapoff.)

It all works rather better if we leave the charging to do_swap_page and
unuse_pte, and do nothing for swapcache itself: revert mm/swap_state.c to
what it was before the memory-controller patches.  This also speeds up
significantly a contained process working at its limit: because it no
longer needs to keep waiting for swap writeback to complete.

Is it unfair that swap pages become uncharged once they're unmapped, even
though they're still clearly private to particular cgroups?  For a short
while, yes; but PageReclaim arranges for those pages to go to the end of
the inactive list and be reclaimed soon if necessary.

shmem/tmpfs pages are a distinct case: their charging also benefits from
this change, but their second life on the lists as swapcache pages may
prove more unfair - that I need to check next.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcgroup: fix zone isolation OOM

mem_cgroup_charge_common shows a tendency to OOM without good reason, when
a memhog goes well beyond its rss limit but with plenty of swap available.
Seen on x86 but not on PowerPC; seen when the next patch omits swapcache
from memcgroup, but we presume it can happen without.

mem_cgroup_isolate_pages is not quite satisfying reclaim's criteria for OOM
avoidance.  Already it has to scan beyond the nr_to_scan limit when it
finds a !LRU page or an active page when handling inactive or an inactive
page when handling active.  It needs to do exactly the same when it finds a
page from the wrong zone (the x86 tests had two zones, the PowerPC tests
had only one).

Don't increment scan and then decrement it in these cases, just move the
incrementation down.  Fix recent off-by-one when checking against
nr_to_scan.  Cut out "Check if the meta page went away from under us",
presumably left over from early debugging: no amount of such checks could
save us if this list really were being updated without locking.

This change does make the unlimited scan while holding two spinlocks
even worse - bad for latency and bad for containment; but that's a
separate issue which is better left to be fixed a little later.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

bugfix for memory cgroup controller: avoid !PageLRU page in mem_cgroup_isolate_pages

This patch makes mem_cgroup_isolate_pages() to be

  - ignore !PageLRU pages.
  - fixes the bug that isolation makes no progress if page_zone(page) != zone
    page once find. (just increment scan in this case.)

kswapd and memory migration removes a page from list when it handles
a page for reclaiming/migration.

Because __isolate_lru_page() doesn't moves page !PageLRU pages, it will
be safe to avoid touching !PageLRU() page and its page_cgroup.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

bugfix for memory cgroup controller: migration under memory controller fix

While using memory control cgroup, page-migration under it works as following.
==
 1. uncharge all refs at try to unmap.
 2. charge regs again remove_migration_ptes()
==
This is simple but has following problems.
==
 The page is uncharged and charged back again if *mapped*.
    - This means that cgroup before migration can be different from one after
      migration
    - If page is not mapped but charged as page cache, charge is just ignored
      (because not mapped, it will not be uncharged before migration)
      This is memory leak.
==
This patch tries to keep memory cgroup at page migration by increasing
one refcnt during it. 3 functions are added.

 mem_cgroup_prepare_migration() --- increase refcnt of page->page_cgroup
 mem_cgroup_end_migration()     --- decrease refcnt of page->page_cgroup
 mem_cgroup_page_migration() --- copy page->page_cgroup from old page to
                                 new page.

During migration
  - old page is under PG_locked.
  - new page is under PG_locked, too.
  - both old page and new page is not on LRU.

These 3 facts guarantee that page_cgroup() migration has no race.

Tested and worked well in x86_64/fake-NUMA box.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

bugfix for memory controller: add helper function for assigning cgroup to page

This patch adds following functions.
   - clear_page_cgroup(page, pc)
   - page_cgroup_assign_new_page_group(page, pc)

Mainly for cleanup.

A manner "check page->cgroup again after lock_page_cgroup()" is
implemented in straight way.

A comment in mem_cgroup_uncharge() will be removed by force-empty patch

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kswapd should only wait on IO if there is IO

The current kswapd (and try_to_free_pages) code has an oddity where the
code will wait on IO, even if there is no IO in flight.  This problem is
notable especially when the system scans through many unfreeable pages,
causing unnecessary stalls in the VM.

Additionally, tasks without __GFP_FS or __GFP_IO in the direct reclaim path
will sleep if a significant number of pages are encountered that should be
written out.  This gives kswapd a chance to write out those pages, while
the direct reclaim task sleeps.

Signed-off-by: Rik van Riel <riel@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

oom: add sysctl to enable task memory dump

Adds a new sysctl, 'oom_dump_tasks', that enables the kernel to produce a
dump of all system tasks (excluding kernel threads) when performing an
OOM-killing.  Information includes pid, uid, tgid, vm size, rss, cpu,
oom_adj score, and name.

This is helpful for determining why there was an OOM condition and which
rogue task caused it.

It is configurable so that large systems, such as those with several
thousand tasks, do not incur a performance penalty associated with dumping
data they may not desire.

If an OOM was triggered as a result of a memory controller, the tasklist
shall be filtered to exclude tasks that are not a member of the same
cgroup.

Cc: Andrea Arcangeli <andrea@suse.de>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcontrol: move oom task exclusion to tasklist scan

Creates a helper function to return non-zero if a task is a member of a
memory controller:

int task_in_mem_cgroup(const struct task_struct *task,
       const struct mem_cgroup *mem);

When the OOM killer is constrained by the memory controller, the exclusion
of tasks that are not a member of that controller was previously misplaced
and appeared in the badness scoring function.  It should be excluded
during the tasklist scan in select_bad_process() instead.

[akpm@linux-foundation.org: build fix]
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mem-controller gfp-mask fix

Need to strip __GFP_HIGHMEM flag while passing to mem_container_cache_charge().

Signed-off-by: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memory controller BUG_ON()

Move mem_controller_cache_charge() above radix_tree_preload().
radix_tree_preload() disables preemption, even though the gfp_mask passed
contains __GFP_WAIT, we cannot really do __GFP_WAIT allocations, thus we
hit a BUG_ON() in kmem_cache_alloc().

This patch moves mem_controller_cache_charge() to above radix_tree_preload()
for cache charging.

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

memcgroup: reinstate swapoff mod

This patch reinstates the "swapoff: scan ptes preemptibly" mod we started
with: in due course it should be rendered down into the earlier patches,
leaving us with a more straightforward mem_cgroup_charge mod to unuse_pte,
allocating with GFP_KERNEL while holding no spinlock and no atomic kmap.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcontrol: move mm_cgroup to header file

Inline functions must preceed their use, so mm_cgroup() should be defined
in linux/memcontrol.h.

include/linux/memcontrol.h:48: warning: 'mm_cgroup' declared inline after
being called
include/linux/memcontrol.h:48: warning: previous declaration of
'mm_cgroup' was here

[akpm@linux-foundation.org: build fix]
[akpm@linux-foundation.org: nuther build fix]
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: make charging gfp mask aware

Nick Piggin pointed out that swap cache and page cache addition routines
could be called from non GFP_KERNEL contexts.  This patch makes the
charging routine aware of the gfp context.  Charging might fail if the
cgroup is over it's limit, in which case a suitable error is returned.

This patch was tested on a Powerpc box.  I am still looking at being able
to test the path, through which allocations happen in non GFP_KERNEL
contexts.

[kamezawa.hiroyu@jp.fujitsu.com: problem with ZONE_MOVABLE]
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-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>

Memory controller: make page_referenced() cgroup aware

Make page_referenced() cgroup aware.  Without this patch, page_referenced()
can cause a page to be skipped while reclaiming pages.  This patch ensures
that other cgroups do not hold pages in a particular cgroup hostage.  It
is required to ensure that shared pages are freed from a cgroup when they
are not actively referenced from the cgroup that brought them in

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: add switch to control what type of pages to limit

Choose if we want cached pages to be accounted or not.  By default both are
accounted for.  A new set of tunables are added.

echo -n 1 > mem_control_type

switches the accounting to account for only mapped pages

echo -n 3 > mem_control_type

switches the behaviour back

[bunk@kernel.org: mm/memcontrol.c: clenups]
[akpm@linux-foundation.org: fix sparc32 build]
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Adrian Bunk <bunk@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: OOM handling

Out of memory handling for cgroups over their limit. A task from the
cgroup over limit is chosen using the existing OOM logic and killed.

TODO:
1. As discussed in the OLS BOF session, consider implementing a user
space policy for OOM handling.

[akpm@linux-foundation.org: fix build due to oom-killer changes]
Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller improve user interface

Change the interface to use bytes instead of pages.  Page sizes can vary
across platforms and configurations.  A new strategy routine has been added
to the resource counters infrastructure to format the data as desired.

Suggested by David Rientjes, Andrew Morton and Herbert Poetzl

Tested on a UML setup with the config for memory control enabled.

[kamezawa.hiroyu@jp.fujitsu.com: possible race fix in res_counter]
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-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>

Memory controller: add per cgroup LRU and reclaim

Add the page_cgroup to the per cgroup LRU.  The reclaim algorithm has
been modified to make the isolate_lru_pages() as a pluggable component.  The
scan_control data structure now accepts the cgroup on behalf of which
reclaims are carried out.  try_to_free_pages() has been extended to become
cgroup aware.

[akpm@linux-foundation.org: fix warning]
[Lee.Schermerhorn@hp.com: initialize all scan_control's isolate_pages member]
[bunk@kernel.org: make do_try_to_free_pages() static]
[hugh@veritas.com: memcgroup: fix try_to_free order]
[kamezawa.hiroyu@jp.fujitsu.com: this unlock_page_cgroup() is unnecessary]
Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-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>

Memory controller: task migration

Allow tasks to migrate from one cgroup to the other.  We migrate
mm_struct's mem_cgroup only when the thread group id migrates.

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: memory accounting

Add the accounting hooks.  The accounting is carried out for RSS and Page
Cache (unmapped) pages.  There is now a common limit and accounting for both.
The RSS accounting is accounted at page_add_*_rmap() and page_remove_rmap()
time.  Page cache is accounted at add_to_page_cache(),
__delete_from_page_cache().  Swap cache is also accounted for.

Each page's page_cgroup is protected with the last bit of the
page_cgroup pointer, this makes handling of race conditions involving
simultaneous mappings of a page easier.  A reference count is kept in the
page_cgroup to deal with cases where a page might be unmapped from the RSS
of all tasks, but still lives in the page cache.

Credits go to Vaidyanathan Srinivasan for helping with reference counting work
of the page cgroup.  Almost all of the page cache accounting code has help
from Vaidyanathan Srinivasan.

[hugh@veritas.com: fix swapoff breakage]
[akpm@linux-foundation.org: fix locking]
Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: <Valdis.Kletnieks@vt.edu>
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: accounting setup

Basic setup routines, the mm_struct has a pointer to the cgroup that
it belongs to and the the page has a page_cgroup associated with it.

Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: cgroups setup

Setup the memory cgroup and add basic hooks and controls to integrate
and work with the cgroup.

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Memory controller: resource counters

With fixes from David Rientjes <rientjes@google.com>

Introduce generic structures and routines for resource accounting.

Each resource accounting cgroup is supposed to aggregate it,
cgroup_subsystem_state and its resource-specific members within.

Signed-off-by: Pavel Emelianov <xemul@openvz.org>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: David Rientjes <rientjes@google.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcgroup: temporarily revert swapoff mod

This patch precisely reverts the "swapoff: scan ptes preemptibly" patch
just presented.  It's a temporary measure to allow existing memory
controller patches to apply without rejects: in due course they should be
rendered down into one sensible patch, and this reversion disappear.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: Balbir Singh <balbir@linux.vnet.ibm.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>

Memory controller: add document

Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Pavel Emelianov <xemul@openvz.org>
Cc: Paul Menage <menage@google.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: Kirill Korotaev <dev@sw.ru>
Cc: Herbert Poetzl <herbert@13thfloor.at>
Cc: David Rientjes <rientjes@google.com>
Cc: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/cgroup.c: make 2 functions static

cgroup_is_releasable() and notify_on_release() should be static,
not global inline.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cgroups: move cgroups destroy() callbacks to cgroup_diput()

Move the calls to the cgroup subsystem destroy() methods from
cgroup_rmdir() to cgroup_diput().  This allows control file reads and
writes to access their subsystem state without having to be concerned with
locking against cgroup destruction - the control file dentry will keep the
cgroup and its subsystem state objects alive until the file is closed.

The documentation is updated to reflect the changed semantics of destroy();
additionally the locking comments for destroy() and some other methods were
clarified and decrustified.

Signed-off-by: Paul Menage <menage@google.com>
Cc: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cgroup simplify space stripping

Simplify the space stripping code in cgroup file write.

[akpm@linux-foundation.org: s/BUG_ON/BUILD_BUG_ON/]
Signed-off-by: Paul Jackson <pj@sgi.com>
Acked-by: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

cgroup brace coding style fix

Coding style fix - one line conditionals don't get braces.

Signed-off-by: Paul Jackson <pj@sgi.com>
Acked-by: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel/cgroup.c: remove dead code

This patch removes dead code spotted by the Coverity checker
(look at the "(nbytes >= PATH_MAX)" check).

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Cc: Paul Jackson <pj@sgi.com>
Cc: Paul Menage <menage@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

email-clients.txt: sylpheed is OK at IMAP

This comment is not helpful (no reason given) and is incorrect.
Just stick to facts that are useful regarding working on Linux.

(akpm: I've used sylpheed+imap for years)

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Paul Jackson <pj@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel-doc: warn on badly formatted short description

Make kernel-doc warn when a function/struct/union/typedef does not contain
a properly formatted short description, such as:
 * scsi_devinfo: set up the dynamic device list
or
 * scsi_devinfo -

This warning is only generated when verbose (-v) mode is used.

Also explain the -v command line option in the -h output.

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

kernel-doc: prevent duplicate description: output

Prevent duplicate output of a Description: section when there is a "blank"
("*") line between the initial function name/description line and the
"Description:" header.

Test case:  drivers/scsi/scsi_devinfo.c::scsi_init_devinfo().

Rob Landley hit this while he was producing SCSI kernel-doc.

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

REPORTING-BUGS: cc the mailing list too

People should also cc relevant mailing lists when reporting bugs.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Documentation: create new scheduler/ subdirectory

The top-level Documentation/ directory is unmanageably large, so we
should take any obvious opportunities to move stuff into subdirectories.
These sched-*.txt files seem an obvious easy case.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Cc: Ingo Molnar <mingo@elte.hu>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Documentation: move sharedsubtrees.txt to filesystems/

This documentation is also vfs-related.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Documentation: move dnotify.txt to filesystems/

I'm inclined to think dnotify belongs in filesystems/.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Acked-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

move edac.txt two levels up

There's no reason for edac.txt for being at this unusual place.

Signed-off-by: Adrian Bunk <bunk@kernel.org>
Acked-by: Alan Cox <alan@redhat.com>
Cc: Doug Thompson <norsk5@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add chapter IDs to z8530book.tmpl

Add chapter IDs to z8530book.tmpl

Signed-off-by: Rob Landley <rob@landley.net>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add table IDs to videobook.tmpl

Add table IDs to videobook.tmpl

Signed-off-by: Rob Landley <rob@landley.net>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add section IDs to rapidio.tmpl

Add section IDs to rapidio.tmpl

Signed-off-by: Rob Landley <rob@landley.net>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add missing IDs to procfs-guide.tmpl

Add missing IDs to procfs-guide.tmpl

Signed-off-by: Rob Landley <rob@landley.net>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add section IDs to mtdnand.tmpl

Add section IDs to mtdnand.tmpl

Signed-off-by: Rob Landley <rob@landley.net>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add missing section ID to lsm.tmpl

Add missing section ID to lsm.tmpl

Signed-off-by: Rob Landley <rob@landley.net>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Add missing section IDs to genericirq.tmpl

Add missing section IDs to genericirq.tmpl

Signed-off-by: Rob Landley <rob@landley.net>
Cc: "Randy.Dunlap" <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tty: Kill TTY_FLIPBUF_SIZE

This legacy define from the old buffer code is now only used in a single
power pc driver than doesn't compile anyway.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

tty: fix tty network driver interactions with TCGET/TCSET calls (x86 fix)

And to go with it Dave's type checking x86 termios headers.  I've updated
these as the original sent by Dave had some wrong types in it.

Signed-off-by: Alan Cox <alan@redhat.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

kernel-doc: fix for vunmap function prototype

Fix kernel-doc function prototype parsing which was exposed by vunmap() by
allowing more than one '*' before the function name.

Error(linux-2.6.24-mm1//mm/vmalloc.c:438): cannot understand prototype: 'struct page **vunmap(const void *addr) '

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

VFS: factor out three helpers for FIBMAP/FIONBIO/FIOASYNC file ioctls

Factor out file-specific ioctl code into smaller helper functions, away from
file_ioctl().  This helps code readability and also reduces indentation inside
case statements.

Signed-off-by: Erez Zadok <ezk@cs.sunysb.edu>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

VFS: swap do_ioctl and vfs_ioctl names

Rename old vfs_ioctl to do_ioctl, because the comment above it clearly
indicates that it is an internal function not to be exported to modules;
therefore it should have a more traditional do_XXX name.  The new do_ioctl
is exported in fs.h but not to modules.

Rename the old do_ioctl to vfs_ioctl because the names vfs_XXX should
preferably be reserved to callable VFS functions which modules may call, as
many other vfs_XXX functions already do.  Export the new vfs_ioctl to GPL
modules so others can use it (including Unionfs and eCryptfs).  Add DocBook
for new vfs_ioctl.

[akpm@linux-foundation.org: fix build]
Signed-off-by: Erez Zadok <ezk@cs.sunysb.edu>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

VFS: apply coding standards to fs/ioctl.c

Signed-off-by: Erez Zadok <ezk@cs.sunysb.edu>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

DS1WM: decouple host IRQ and INTR active state settings

The DS1WM driver incorrectly infers the IAS bit (1-wire interrupt active
high) from IRQ settings.  There are devices that have IAS=0 but still need
the IRQ to trigger on a rising edge.  With this patch, machines with DS1WM
that need IAS=1 have to set .active_high=1 in the ds1wm_platform_data.

Signed-off-by: Philipp Zabel <philipp.zabel@gmail.com>
Acked-by: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Acked-by: Matt Reimer <mreimer@vpop.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gfs2: make gfs2_glock.gl_owner_pid be a struct pid *

The gl_owner_pid field is used to get the lock owning task by its pid, so make
it in a proper manner, i.e.  by using the struct pid pointer and pid_task()
function.

The pid_task() becomes exported for the gfs2 module.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

power_supply: remove capacity_level from list of sysfs attributes

This commit:

commit 8efe444038a205e79b38b7ad03878824901849a8
Author: Andres Salomon <dilinger@queued.net>
Date:   Wed Dec 12 14:12:56 2007 -0500

    power: remove POWER_SUPPLY_PROP_CAPACITY_LEVEL

Removed CAPACITY_LEVEL from every other code, leaving the array with sysfs
attributes with one more entry than the number of enums in power_supply.h.
This leads to some attributes containing the value of the attribute right
after it.  For example, temp_ambient would have the value of
time_to_empty_now.  In my case, I had time_to_full_avg have the value which
should be in model_name, when the former was usually empty.

Cc: Andres Salomon <dilinger@debian.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gfs2: make gfs2_holder.gh_owner_pid be a struct pid *

The gl_owner_pid field is used to get the holder task by its pid and check
whether the current is a holder, so make it in a proper manner, i.e.  via the
struct pid * manipulations.

Signed-off-by: Pavel Emelyanov <xemul@openvz.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Acked-by: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

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

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ericvh/v9fs:
  9p: fix p9_printfcall export
  9p: transport API reorganization
  9p: add remove function to trans_virtio
  9p: Convert semaphore to spinlock for p9_idpool
  9p: fix mmap to be read-only
  9p: add support for sticky bit
  9p: Fix soft lockup in virtio transport
  9p: fix bug in attach-per-user
  9p: block-based virtio client
  9p: create transport rpc cut-thru
  9p: fix bug in p9_clone_stat

9p: fix p9_printfcall export

ERROR: "p9_printfcall" [net/9p/9pnet_virtio.ko] undefined!

Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: transport API reorganization

This merges the mux.c (including the connection interface) with trans_fd
in preparation for transport API changes.  Ultimately, trans_fd will need
to be rewritten to clean it up and simplify the implementation, but this
reorganization is viewed as the first step.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: add remove function to trans_virtio

Request from rusty:
Just cleaning up patches for 2.6.25 merge, and noticed that
net/9p/trans_virtio.c doesn't have a remove function.  This will crash when
removing the module (console doesn't have one because it can't really be
removed).

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: Convert semaphore to spinlock for p9_idpool

When booting from v9fs, down_interruptible in p9_idpool_get() triggered a BUG
as it was being called with IRQs disabled.  A spinlock seems like the right
thing to be using since the idr functions go out of their way not to sleep.

This patch eliminates the BUG by converting the semaphore to a spinlock.

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Acked-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: fix mmap to be read-only

v9fs was allowing writable mmap which could lead to kernel BUG() cases.
This sets the mmap function to generic_file_readonly_mmap which (correctly)
returns an error to applications which open mmap for writing.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: add support for sticky bit

GDM gets unhappy if /var/gdm doesn't have the sticky bit set.  This patch adds
support for the sticky bit in much the same way setuid/setgid is supported.

With this patch, I can launch X from a v9fs rootfs (although I quickly run out
of fds in the server once gnome starts up).

Signed-off-by: Anthony Liguori <aliguori@us.ibm.com>
Acked-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: Fix soft lockup in virtio transport

This fixes a poorly placed spinlock which could result in a
soft lockup condition.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: fix bug in attach-per-user

When a new user attached at a directory other than the root, he would end
up in the parent directory of the cwd.  This was due to a logic error in
the code which attaches the user at the mount point and walks back to the
cwd.  This patch fixes that.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: block-based virtio client

This replaces the console-based virto client with a block-based
client using a single request queue.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: create transport rpc cut-thru

Add a new transport function which allows a cut-thru directly to
the transport instead of processing request through the mux if the
cut-thru exists.

Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

9p: fix bug in p9_clone_stat

This patch fixes a bug in the copying of 9P
stat information where string references
weren't being updated properly.

Signed-off-by: Martin Sava <martin.stava@gmail.com>
Signed-off-by: Eric Van Hensbergen <ericvh@gmail.com>

docbook: dmapool: fix fatal changed filename

Docbook fatal error, file was moved:
docproc: linux-2.6.24-git15/drivers/base/dmapool.c: No such file or directory

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Merge git://git./linux/kernel/git/x86/linux-2.6-x86

* git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-x86:
  x86: fix deadlock, make pgd_lock irq-safe
  virtio: fix trivial build bug
  x86: fix mttr trimming
  x86: delay CPA self-test and repeat it
  x86: fix 64-bit sections
  generic: add __FINITDATA
  x86: remove suprious ifdefs from pageattr.c
  x86: mark the .rodata section also NX
  x86: fix iret exception recovery on 64-bit
  cpuidle: dubious one-bit signed bitfield in cpuidle.h
  x86: fix sparse warnings in powernow-k8.c
  x86: fix sparse error in traps_32.c
  x86: trivial sparse/checkpatch in quirks.c
  x86 ptrace: disallow null cs/ss
  MAINTAINERS: RDC R-321x SoC maintainer
  brk randomization: introduce CONFIG_COMPAT_BRK
  brk: check the lower bound properly
  x86: remove X2 workaround
  x86: make spurious fault handler aware of large mappings
  x86: make traps on entry code be debuggable in user space, 64-bit

x86: fix deadlock, make pgd_lock irq-safe

lockdep just caught this one:

=================================
[ INFO: inconsistent lock state ]
2.6.24 #38
---------------------------------
inconsistent {in-softirq-W} -> {softirq-on-W} usage.
swapper/1 [HC0[0]:SC0[0]:HE1:SE1] takes:
 (pgd_lock){-+..}, at: [<ffffffff8022a9ea>] mm_init+0x1da/0x250
{in-softirq-W} state was registered at:
  [<ffffffffffffffff>] 0xffffffffffffffff
irq event stamp: 394559
hardirqs last  enabled at (394559): [<ffffffff80267f0a>] get_page_from_freelist+0x30a/0x4c0
hardirqs last disabled at (394558): [<ffffffff80267d25>] get_page_from_freelist+0x125/0x4c0
softirqs last  enabled at (393952): [<ffffffff80232f8e>] __do_softirq+0xce/0xe0
softirqs last disabled at (393945): [<ffffffff8020c57c>] call_softirq+0x1c/0x30

other info that might help us debug this:
no locks held by swapper/1.

stack backtrace:
Pid: 1, comm: swapper Not tainted 2.6.24 #38

Call Trace:
 [<ffffffff8024e1fb>] print_usage_bug+0x18b/0x190
 [<ffffffff8024f55d>] mark_lock+0x53d/0x560
 [<ffffffff8024fffa>] __lock_acquire+0x3ca/0xed0
 [<ffffffff80250ba8>] lock_acquire+0xa8/0xe0
 [<ffffffff8022a9ea>] ? mm_init+0x1da/0x250
 [<ffffffff809bcd10>] _spin_lock+0x30/0x70
 [<ffffffff8022a9ea>] mm_init+0x1da/0x250
 [<ffffffff8022aa99>] mm_alloc+0x39/0x50
 [<ffffffff8028b95a>] bprm_mm_init+0x2a/0x1a0
 [<ffffffff8028d12b>] do_execve+0x7b/0x220
 [<ffffffff80209776>] sys_execve+0x46/0x70
 [<ffffffff8020c214>] kernel_execve+0x64/0xd0
 [<ffffffff8020901e>] ? _stext+0x1e/0x20
 [<ffffffff802090ba>] init_post+0x9a/0xf0
 [<ffffffff809bc5f6>] ? trace_hardirqs_on_thunk+0x35/0x3a
 [<ffffffff8024f75a>] ? trace_hardirqs_on+0xba/0xd0
 [<ffffffff8020c1a8>] ? child_rip+0xa/0x12
 [<ffffffff8020bcbc>] ? restore_args+0x0/0x44
 [<ffffffff8020c19e>] ? child_rip+0x0/0x12

turns out that pgd_lock has been used on 64-bit x86 in an irq-unsafe
way for almost two years, since commit 8c914cb704a11460e.

Signed-off-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

virtio: fix trivial build bug

fix build bug:

  drivers/virtio/virtio_balloon.c: In function 'fill_balloon':
  drivers/virtio/virtio_balloon.c:98: error: implicit declaration of function 'msleep'

Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: fix mttr trimming

Pavel Emelyanov reported that his networking card did not work
and bisected it down to:

"
The commit

  093af8d7f0ba3c6be1485973508584ef081e9f93
  x86_32: trim memory by updating e820

broke my e1000 card: on loading driver says that

  e1000: probe of 0000:04:03.0 failed with error -5

and the interface doesn't appear.
"

on a 32-bit kernel, base will overflow when try to do PAGE_SHIFT,
and highest_addr will always less 4G.

So use pfn instead of address to avoid the overflow when more than
4g RAM is installed on a 32-bit kernel.

Many thanks to Pavel Emelyanov for reporting and testing it.

Bisected-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Yinghai Lu <yinghai.lu@sun.com>
Tested-by: Pavel Emelyanov <xemul@openvz.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: delay CPA self-test and repeat it

delay the CPA self-test so that any impact (corruption) of
user-space pagetables can be triggered. Repeat the test
every 30 seconds.

this would have prevented the bug fixed by 8cb2a7c1e95e472b5,
at its source.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: fix 64-bit sections

fix 64-bit section warnings.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

generic: add __FINITDATA

Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: remove suprious ifdefs from pageattr.c

The .rodata section really should just be read only; the config option
is there to make breaking up the 2Mb page an option (so people whos machines
give more performance for the 2Mb case can opt to do so).
But when the page gets split anyway, this is no longer an issue, so
clean up the code and remove the ifdefs

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: mark the .rodata section also NX

The .rodata section shouldn't just be read-only,
but also non-executable. This is free since we've broken
up the 2MB page already anyway.

also update test_nx to check for this.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: fix iret exception recovery on 64-bit

This change broke recovery of exceptions in iret:

   commit 72fe4858544292ad64600765cb78bc02298c6b1c
   Author: Glauber de Oliveira Costa <gcosta@redhat.com>

       x86: replace privileged instructions with paravirt macros

The ENTRY(native_iret) macro adds alignment padding before the iretq
instruction, so "iret_label" no longer points exactly at the instruction.
It was sloppy to leave the old "iret_label" label behind when replacing
its nearby use.  Removing it would have revealed the other use of the
label later in the file, and upon noticing that use, anyone exercising
the minimum of attention to detail expected of anyone touching this
subtle code would realize it needed to change as well.

Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

cpuidle: dubious one-bit signed bitfield in cpuidle.h

fix these sparse warnings:

  CHECK   arch/x86/kernel/acpi/cstate.c
include/linux/cpuidle.h:82:17: error: dubious one-bit signed bitfield
  CHECK   arch/x86/kernel/acpi/processor.c
include/linux/cpuidle.h:82:17: error: dubious one-bit signed bitfield
  CHECK   arch/x86/kernel/cpu/cpufreq/powernow-k7.c
include/linux/cpuidle.h:82:17: error: dubious one-bit signed bitfield
  CHECK   arch/x86/kernel/cpu/cpufreq/powernow-k8.c
include/linux/cpuidle.h:82:17: error: dubious one-bit signed bitfield
  CHECK   arch/x86/kernel/cpu/cpufreq/longhaul.c
include/linux/cpuidle.h:82:17: error: dubious one-bit signed bitfield
  CHECK   arch/x86/kernel/cpu/cpufreq/acpi-cpufreq.c
include/linux/cpuidle.h:82:17: error: dubious one-bit signed bitfield

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: fix sparse warnings in powernow-k8.c

arch/x86/kernel/cpu/cpufreq/powernow-k8.c:830:7: warning: symbol 'hi' shadows an earlier one
arch/x86/kernel/cpu/cpufreq/powernow-k8.c:824:6: originally declared here
arch/x86/kernel/cpu/cpufreq/powernow-k8.c:830:15: warning: symbol 'lo' shadows an earlier one
arch/x86/kernel/cpu/cpufreq/powernow-k8.c:824:14: originally declared here

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: fix sparse error in traps_32.c

This was being used to ensure the proper alignment of the FXSAVE/FXRSTOR data.
This would create a sparse error in the _correct_ cases, hiding further
warnings.  Use BUILD_BUG_ON instead.

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: trivial sparse/checkpatch in quirks.c

arch/x86/kernel/quirks.c:384:3: warning: returning void-valued expression
arch/x86/kernel/quirks.c:387:3: warning: returning void-valued expression
arch/x86/kernel/quirks.c:390:3: warning: returning void-valued expression
arch/x86/kernel/quirks.c:393:3: warning: returning void-valued expression

Signed-off-by: Harvey Harrison <harvey.harrison@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86 ptrace: disallow null cs/ss

In my revamp of the x86 ptrace code for setting register values,
I accidentally omitted a check that was there in the old code.
Allowing %cs to be 0 causes a bad crash in recovery from iret failure.
This patch fixes that regression against 2.6.24, and adds a comment
that should help prevent this subtlety from being overlooked again.

Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

MAINTAINERS: RDC R-321x SoC maintainer

Signed-off-by: Florian Fainelli <florian.fainelli@telecomint.eu>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

brk randomization: introduce CONFIG_COMPAT_BRK

based on similar patch from: Pavel Machek <pavel@ucw.cz>

Introduce CONFIG_COMPAT_BRK. If disabled then the kernel is free
(but not obliged to) randomize the brk area.

Heap randomization breaks ancient binaries, so we keep COMPAT_BRK
enabled by default.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

brk: check the lower bound properly

There is a check in sys_brk(), that tries to make sure that we do not
underflow the area that is dedicated to brk heap.

The check is however wrong, as it assumes that brk area starts immediately
after the end of the code (+bss), which is wrong for example in
environments with randomized brk start. The proper way is to check whether
the address is not below the start_brk address.

Signed-off-by: Jiri Kosina <jkosina@suse.cz>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: remove X2 workaround

With the spurious handler fix, the X2 does not lock up anymore.

Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: make spurious fault handler aware of large mappings

In very rare cases, on certain CPUs, we could end up in the spurious
fault handler and ignore a large pud/pmd mapping. The resulting pte
pointer points into the mapped physical space and dereferencing it
will fault recursively.

Make the code aware of large mappings and do the permission check
on the pmd/pud entry, when a large pud/pmd mapping is detected.

Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

x86: make traps on entry code be debuggable in user space, 64-bit

Unify the x86-64 behavior for 32-bit processes that set
bogus %cs/%ss values (the only ones that can fault in iret)
match what the native i386 behavior is. (do not kill the task
via do_exit but generate a SIGSEGV signal)

[ tglx@linutronix.de: build fix ]

Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

Merge branch 'async-tx-for-linus' of git://lost.foo-projects.org/~dwillia2/git/iop into fix

* 'async-tx-for-linus' of git://lost.foo-projects.org/~dwillia2/git/iop:
  async_tx: allow architecture specific async_tx_find_channel implementations
  async_tx: replace 'int_en' with operation preparation flags
  async_tx: kill tx_set_src and tx_set_dest methods
  async_tx: kill ASYNC_TX_ASSUME_COHERENT
  iop-adma: use LIST_HEAD instead of LIST_HEAD_INIT
  async_tx: use LIST_HEAD instead of LIST_HEAD_INIT
  async_tx: fix compile breakage, mark do_async_xor __always_inline

scsi: megaraid: trivial drop duplicate mutex.h include

Signed-off-by: Daniel Walker <dwalker@mvista.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

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

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/selinux-2.6:
  SELinux: Remove security_get_policycaps()
  security: allow Kconfig to set default mmap_min_addr protection

Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  ata_piix.c:piix_init_one() must be __devinit
  sata_via.c: Remove missleading comment.
  libata-core: unblacklist HITACHI drives
  sata_nv: fix ATAPI issues with memory over 4GB (v7)
  ata: drivers/ata/sata_mv.c needs dmapool.h
  libata: kill now unused n_iter and fix sata_fsl
  ahci: fix CAP.NP and PI handling
  sata_mv: Support SoC controllers
  Rename: linux/pata_platform.h to linux/ata_platform.h

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

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (35 commits)
  virtio net: fix oops on interface-up
  Fix PHY Lib support for gianfar and ucc_geth
  forcedeth: preserve registers
  forcedeth: phy status fix
  forcedeth: restart tx/rx
  ipvs: Make wrr "no available servers" error message rate-limited
  [PPPOL2TP]: Label unused warning when CONFIG_PROC_FS is not set.
  [NET_SCHED]: cls_flow: support classification based on VLAN tag
  [VLAN]: Constify skb argument to vlan_get_tag()
  [NET_SCHED]: cls_flow: fix key mask validity check
  [NET_SCHED]: em_meta: fix compile warning
  b43: Fix DMA for 30/32-bit DMA engines
  b43: fix build with CONFIG_SSB_PCIHOST=n
  mac80211: Is not EXPERIMENTAL anymore
  iwl3945-base.c: fix off-by-one errors
  b43legacy: fix DMA slot resource leakage
  b43legacy: drop packets we are not able to encrypt
  b43legacy: fix suspend/resume
  b43legacy: fix PIO crash
  Generic HDLC - use random_ether_addr()
  ...

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

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
  [SPARC64]: Temporarily remove IOMMU merging code.
  [SPARC64]: Update defconfig.
  [SPARC]: Add new timerfd syscall entries.

fb: fix warning: no return statement in function returning non-void

Warning is reproducible with selected FB_CFB_REV_PIXELS_IN_BYTE.

  CC      drivers/video/sysfillrect.o
In file included from drivers/video/sysfillrect.c:18:
drivers/video/fb_draw.h: In function `fb_rev_pixels_in_long':
drivers/video/fb_draw.h:94: warning: no return statement in function returning non-void
  CC      drivers/video/syscopyarea.o
In file included from drivers/video/syscopyarea.c:22:
drivers/video/fb_draw.h: In function `fb_rev_pixels_in_long':
drivers/video/fb_draw.h:94: warning: no return statement in function returning non-void

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Cc: "Antonino A. Daplas" <adaplas@pol.net>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

virtio: add missing #include <linux/delay.h>

Include linux/delay.h to fix compiler error:

drivers/virtio/virtio_balloon.c: In function 'fill_balloon':
drivers/virtio/virtio_balloon.c:98: error: implicit declaration of function 'msleep'

Signed-off-by: Johann Felix Soden <johfel@users.sourceforge.net>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ext3: fix lock inversion in direct IO

We cannot start transaction in ext3_direct_IO() and just let it last during
the whole write because dio_get_page() acquires mmap_sem which ranks above
transaction start (e.g.  because we have dependency chain
mmap_sem->PageLock->journal_start, or because we update atime while holding
mmap_sem) and thus deadlocks could happen.  We solve the problem by
starting a transaction separately for each ext3_get_block() call.

We *could* have a problem that we allocate a block and before its data are
written out the machine crashes and thus we expose stale data.  But that
does not happen because for hole-filling generic code falls back to
buffered writes and for file extension, we add inode to orphan list and
thus in case of crash, journal replay will truncate inode back to the
original size.

[akpm@linux-foundation.org: build fix]
Signed-off-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>