platform/kernel/linux-starfive.git
11 years agoncpfs: fix error return code in ncp_parse_options()
Wei Yongjun [Mon, 8 Jul 2013 23:00:44 +0000 (16:00 -0700)]
ncpfs: fix error return code in ncp_parse_options()

Fix to return -EINVAL from the option parse error handling case instead
of 0, as done elsewhere in this function.

Signed-off-by: Wei Yongjun <yongjun_wei@trendmicro.com.cn>
Cc: Petr Vandrovec <petr@vandrovec.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agocheckpatch: make the CamelCase cache work for non-git trees too
Joe Perches [Mon, 8 Jul 2013 23:00:43 +0000 (16:00 -0700)]
checkpatch: make the CamelCase cache work for non-git trees too

Might as well check include timestamps and cache the include file
CamelCase uses for the non-git case too.

The camelcase cache file is now named:

  for git:      .checkpatch-camelcase.git.<commit_id>
  for non-git:  .checkpatch-camelcase.date.<YYYYMMDDhhmm>

All .checkpatch-camelcase* files are deleted if not current.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agopanic: add cpu/pid to warn_slowpath_common in WARNING printk()s
Alex Thorlton [Mon, 8 Jul 2013 23:00:42 +0000 (16:00 -0700)]
panic: add cpu/pid to warn_slowpath_common in WARNING printk()s

Add the cpu/pid that called WARN() so that the stack traces can be
matched up with the WARNING messages.

[akpm@linux-foundation.org: remove stray quote]
Signed-off-by: Alex Thorlton <athorlton@sgi.com>
Reviewed-by: Robin Holt <holt@sgi.com>
Cc: Stephen Boyd <sboyd@codeaurora.org>
Cc: Vikram Mulukutla <markivx@codeaurora.org>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/memory_hotplug.c: fix return value of online_pages()
Toshi Kani [Mon, 8 Jul 2013 23:00:41 +0000 (16:00 -0700)]
mm/memory_hotplug.c: fix return value of online_pages()

online_pages() is called from memory_block_action() when a user requests
to online a memory block via sysfs.  This function needs to return a
proper error value in case of error.

Signed-off-by: Toshi Kani <toshi.kani@hp.com>
Cc: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Tang Chen <tangchen@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm: honor min_free_kbytes set by user
Michal Hocko [Mon, 8 Jul 2013 23:00:40 +0000 (16:00 -0700)]
mm: honor min_free_kbytes set by user

min_free_kbytes is updated during memory hotplug (by
init_per_zone_wmark_min) currently which is right thing to do in most
cases but this could be unexpected if admin increased the value to
prevent from allocation failures and the new min_free_kbytes would be
decreased as a result of memory hotadd.

This patch saves the user defined value and allows updating
min_free_kbytes only if it is higher than the saved one.

A warning is printed when the new value is ignored.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Cc: Mel Gorman <mgorman@suse.de>
Acked-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: don't need to free memcg via RCU or workqueue
Li Zefan [Mon, 8 Jul 2013 23:00:38 +0000 (16:00 -0700)]
memcg: don't need to free memcg via RCU or workqueue

Now memcg has the same life cycle with its corresponding cgroup, and a
cgroup is freed via RCU and then mem_cgroup_css_free() will be called in
a work function, so we can simply call __mem_cgroup_free() in
mem_cgroup_css_free().

This actually reverts commit 59927fb984d ("memcg: free mem_cgroup by RCU
to fix oops").

Signed-off-by: Li Zefan <lizefan@huawei.com>
Cc: Hugh Dickins <hughd@google.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: kill memcg refcnt
Li Zefan [Mon, 8 Jul 2013 23:00:37 +0000 (16:00 -0700)]
memcg: kill memcg refcnt

Now memcg has the same life cycle as its corresponding cgroup.  Kill the
useless refcnt.

Signed-off-by: Li Zefan <lizefan@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: don't need to get a reference to the parent
Li Zefan [Mon, 8 Jul 2013 23:00:36 +0000 (16:00 -0700)]
memcg: don't need to get a reference to the parent

The cgroup core guarantees it's always safe to access the parent.

Signed-off-by: Li Zefan <lizefan@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: use css_get/put for swap memcg
Li Zefan [Mon, 8 Jul 2013 23:00:34 +0000 (16:00 -0700)]
memcg: use css_get/put for swap memcg

Use css_get/put instead of mem_cgroup_get/put.  A simple replacement
will do.

The historical reason that memcg has its own refcnt instead of always
using css_get/put, is that cgroup couldn't be removed if there're still
css refs, so css refs can't be used as long-lived reference.  The
situation has changed so that rmdir a cgroup will succeed regardless css
refs, but won't be freed until css refs goes down to 0.

Signed-off-by: Li Zefan <lizefan@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: use css_get/put when charging/uncharging kmem
Li Zefan [Mon, 8 Jul 2013 23:00:33 +0000 (16:00 -0700)]
memcg: use css_get/put when charging/uncharging kmem

Use css_get/put instead of mem_cgroup_get/put.

We can't do a simple replacement, because here mem_cgroup_put() is
called during mem_cgroup_css_free(), while mem_cgroup_css_free() won't
be called until css refcnt goes down to 0.

Instead we increment css refcnt in mem_cgroup_css_offline(), and then
check if there's still kmem charges.  If not, css refcnt will be
decremented immediately, otherwise the refcnt will be released after the
last kmem allocation is uncahred.

[akpm@linux-foundation.org: tweak comment]
Signed-off-by: Li Zefan <lizefan@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: don't use mem_cgroup_get() when creating a kmemcg cache
Li Zefan [Mon, 8 Jul 2013 23:00:31 +0000 (16:00 -0700)]
memcg: don't use mem_cgroup_get() when creating a kmemcg cache

Use css_get()/css_put() instead of mem_cgroup_get()/mem_cgroup_put().

There are two things being done in the current code:

First, we acquired a css_ref to make sure that the underlying cgroup
would not go away.  That is a short lived reference, and it is put as
soon as the cache is created.

At this point, we acquire a long-lived per-cache memcg reference count
to guarantee that the memcg will still be alive.

so it is:

  enqueue: css_get
  create : memcg_get, css_put
  destroy: memcg_put

So we only need to get rid of the memcg_get, change the memcg_put to
css_put, and get rid of the now extra css_put.

(This changelog is mostly written by Glauber)

Signed-off-by: Li Zefan <lizefan@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: use css_get() in sock_update_memcg()
Li Zefan [Mon, 8 Jul 2013 23:00:30 +0000 (16:00 -0700)]
memcg: use css_get() in sock_update_memcg()

Use css_get/css_put instead of mem_cgroup_get/put.

Note, if at the same time someone is moving @current to a different
cgroup and removing the old cgroup, css_tryget() may return false, and
sock->sk_cgrp won't be initialized, which is fine.

Signed-off-by: Li Zefan <lizefan@huawei.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Hugh Dickins <hughd@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg, kmem: fix reference count handling on the error path
Michal Hocko [Mon, 8 Jul 2013 23:00:29 +0000 (16:00 -0700)]
memcg, kmem: fix reference count handling on the error path

mem_cgroup_css_online calls mem_cgroup_put if memcg_init_kmem fails.
This is not correct because only memcg_propagate_kmem takes an
additional reference while mem_cgroup_sockets_init is allowed to fail as
well (although no current implementation fails) but it doesn't take any
reference.  This all suggests that it should be memcg_propagate_kmem
that should clean up after itself so this patch moves mem_cgroup_put
over there.

Unfortunately this is not that easy (as pointed out by Li Zefan) because
memcg_kmem_mark_dead marks the group dead (KMEM_ACCOUNTED_DEAD) if it is
marked active (KMEM_ACCOUNTED_ACTIVE) which is the case even if
memcg_propagate_kmem fails so the additional reference is dropped in
that case in kmem_cgroup_destroy which means that the reference would be
dropped two times.

The easiest way then would be to simply remove mem_cgrroup_put from
mem_cgroup_css_online and rely on kmem_cgroup_destroy doing the right
thing.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Li Zefan <lizefan@huawei.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: <stable@vger.kernel.org> [3.8]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoRevert "memcg: avoid dangling reference count in creation failure"
Michal Hocko [Mon, 8 Jul 2013 23:00:27 +0000 (16:00 -0700)]
Revert "memcg: avoid dangling reference count in creation failure"

This reverts commit e4715f01be697a.

mem_cgroup_put is hierarchy aware so mem_cgroup_put(memcg) already drops
an additional reference from all parents so the additional
mem_cgrroup_put(parent) potentially causes use-after-free.

Signed-off-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Li Zefan <lizefan@huawei.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: Tejun Heo <tj@kernel.org>
Cc: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: <stable@vger.kernel.org> [3.9+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agommap: allow MAP_HUGETLB for hugetlbfs files v2
Jörn Engel [Mon, 8 Jul 2013 23:00:26 +0000 (16:00 -0700)]
mmap: allow MAP_HUGETLB for hugetlbfs files v2

It is counterintuitive at best that mmap'ing a hugetlbfs file with
MAP_HUGETLB fails, while mmap'ing it without will a) succeed and b)
return huge pages.

v2: use is_file_hugepages(), as suggested by Jianguo

Signed-off-by: Joern Engel <joern@logfs.org>
Cc: Jianguo Wu <wujianguo@huawei.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm: vmscan: do not scale writeback pages when deciding whether to set ZONE_WRITEBACK
Mel Gorman [Mon, 8 Jul 2013 23:00:25 +0000 (16:00 -0700)]
mm: vmscan: do not scale writeback pages when deciding whether to set ZONE_WRITEBACK

After the patch "mm: vmscan: Flatten kswapd priority loop" was merged
the scanning priority of kswapd changed.

The priority now rises until it is scanning enough pages to meet the
high watermark.  shrink_inactive_list sets ZONE_WRITEBACK if a number of
pages were encountered under writeback but this value is scaled based on
the priority.  As kswapd frequently scans with a higher priority now it
is relatively easy to set ZONE_WRITEBACK.  This patch removes the
scaling and treates writeback pages similar to how it treats unqueued
dirty pages and congested pages.  The user-visible effect should be that
kswapd will writeback fewer pages from reclaim context.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Dave Chinner <david@fromorbit.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>
11 years agomm: vmscan: do not continue scanning if reclaim was aborted for compaction
Mel Gorman [Mon, 8 Jul 2013 23:00:24 +0000 (16:00 -0700)]
mm: vmscan: do not continue scanning if reclaim was aborted for compaction

Direct reclaim is not aborting to allow compaction to go ahead properly.
do_try_to_free_pages is told to abort reclaim which is happily ignores
and instead increases priority instead until it reaches 0 and starts
shrinking file/anon equally.  This patch corrects the situation by
aborting reclaim when requested instead of raising priority.

Signed-off-by: Mel Gorman <mgorman@suse.de>
Cc: Rik van Riel <riel@redhat.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Dave Chinner <david@fromorbit.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>
11 years agomm/memory_hotplug.c: fix a comment typo in register_page_bootmem_info_node()
Tang Chen [Mon, 8 Jul 2013 23:00:23 +0000 (16:00 -0700)]
mm/memory_hotplug.c: fix a comment typo in register_page_bootmem_info_node()

Signed-off-by: Tang Chen <tangchen@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/memblock.c: fix wrong comment in __next_free_mem_range()
Tang Chen [Mon, 8 Jul 2013 23:00:22 +0000 (16:00 -0700)]
mm/memblock.c: fix wrong comment in __next_free_mem_range()

Remove one redundant "nid" in the comment.

Signed-off-by: Tang Chen <tangchen@cn.fujitsu.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agopage migration: fix wrong comment in address_space_operations.migratepage()
Tang Chen [Mon, 8 Jul 2013 23:00:21 +0000 (16:00 -0700)]
page migration: fix wrong comment in address_space_operations.migratepage()

There is no parameter "sync" in address_space_operations->migratepage().
It should be migrate_mode.  And the comment is for MIGRATE_ASYNC.

Signed-off-by: Tang Chen <tangchen@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/vmalloc.c: fix an overflow bug in alloc_vmap_area()
Zhang Yanfei [Mon, 8 Jul 2013 23:00:19 +0000 (16:00 -0700)]
mm/vmalloc.c: fix an overflow bug in alloc_vmap_area()

When searching a vmap area in the vmalloc space, we use (addr + size -
1) to check if the value is less than addr, which is an overflow.  But
we assign (addr + size) to vmap_area->va_end.

So if we come across the below case:

  (addr + size - 1) : not overflow
  (addr + size)     : overflow

we will assign an overflow value (e.g 0) to vmap_area->va_end, And this
will trigger BUG in __insert_vmap_area, causing system panic.

So using (addr + size) to check the overflow should be the correct
behaviour, not (addr + size - 1).

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Reported-by: Ghennadi Procopciuc <unix140@gmail.com>
Tested-by: Daniel Baluta <dbaluta@ixiacom.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm: remove unused VM_<READfoo> macros and expand other in-place
Joe Perches [Mon, 8 Jul 2013 23:00:18 +0000 (16:00 -0700)]
mm: remove unused VM_<READfoo> macros and expand other in-place

These VM_<READfoo> macros aren't used very often and three of them
aren't used at all.

Expand the ones that are used in-place, and remove all the now unused
#define VM_<foo> macros.

VM_READHINTMASK, VM_NormalReadHint and VM_ClearReadHint were added just
before 2.4 and appears have never been used.

Signed-off-by: Joe Perches <joe@perches.com>
Acked-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/pgtable: don't accumulate addr during pgd prepopulate pmd
Wanpeng Li [Mon, 8 Jul 2013 23:00:17 +0000 (16:00 -0700)]
mm/pgtable: don't accumulate addr during pgd prepopulate pmd

The old codes accumulate addr to get right pmd, however, currently pmds
are preallocated and transfered as a parameter, there is unnecessary to
accumulate addr variable any more, this patch remove it.

Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Reviewed-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/thp: fix doc for transparent huge zero page
Wanpeng Li [Mon, 8 Jul 2013 23:00:16 +0000 (16:00 -0700)]
mm/thp: fix doc for transparent huge zero page

Transparent huge zero page is used during the page fault instead of in
khugepaged.

  # ls /sys/kernel/mm/transparent_hugepage/
  defrag  enabled  khugepaged  use_zero_page
  # ls /sys/kernel/mm/transparent_hugepage/khugepaged/
  alloc_sleep_millisecs  defrag  full_scans  max_ptes_none  pages_collapsed  pages_to_scan  scan_sleep_millisecs

This patch corrects the documentation just like the codes done.

Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Acked-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/page_alloc: fix doc for numa_zonelist_order
Wanpeng Li [Mon, 8 Jul 2013 23:00:16 +0000 (16:00 -0700)]
mm/page_alloc: fix doc for numa_zonelist_order

The default zonelist order selecter will select "node" order if any nodes
DMA zone comprises greater than 70% of its local memory instead of 60%,
according to default_zonelist_order::low_kmem_size > total * 70/100.

Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/writeback: commit reason of WB_REASON_FORKER_THREAD mismatch name
Wanpeng Li [Mon, 8 Jul 2013 23:00:15 +0000 (16:00 -0700)]
mm/writeback: commit reason of WB_REASON_FORKER_THREAD mismatch name

After commit 839a8e8660b6 ("writeback: replace custom worker pool
implementation with unbound workqueue"), there is no bdi forker thread
any more.  However, WB_REASON_FORKER_THREAD is still used due to it is
TPs userland visible and we won't be exposing exactly the same
information with just a different name.

Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/writeback: don't check force_wait to handle bdi->work_list
Wanpeng Li [Mon, 8 Jul 2013 23:00:14 +0000 (16:00 -0700)]
mm/writeback: don't check force_wait to handle bdi->work_list

After commit 839a8e8660b6 ("writeback: replace custom worker pool
implementation with unbound workqueue"), bdi_writeback_workfn runs off
bdi_writeback->dwork, on each execution, it processes bdi->work_list and
reschedules if there are more things to do instead of flush any work
that race with us existing.  It is unecessary to check force_wait in
wb_do_writeback since it is always 0 after the mentioned commit.  This
patch remove the force_wait in wb_do_writeback.

Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/writeback: remove wb_reason_name
Wanpeng Li [Mon, 8 Jul 2013 23:00:12 +0000 (16:00 -0700)]
mm/writeback: remove wb_reason_name

wb_reason_name is not used any more - remove it.

Signed-off-by: Wanpeng Li <liwanp@linux.vnet.ibm.com>
Reviewed-by: Tejun Heo <tj@kernel.org>
Reviewed-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agofs/fs-writeback.c: : make wb_do_writeback() as static
Haicheng Li [Mon, 8 Jul 2013 23:00:11 +0000 (16:00 -0700)]
fs/fs-writeback.c: : make wb_do_writeback() as static

It's not used globally and could be static.

Signed-off-by: Haicheng Li <haicheng.li@linux.intel.com>
Cc: Jan Kara <jack@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/sparse.c: put clear_hwpoisoned_pages within CONFIG_MEMORY_HOTREMOVE
Zhang Yanfei [Mon, 8 Jul 2013 23:00:10 +0000 (16:00 -0700)]
mm/sparse.c: put clear_hwpoisoned_pages within CONFIG_MEMORY_HOTREMOVE

With CONFIG_MEMORY_HOTREMOVE unset, there is a compile warning:

  mm/sparse.c:755: warning: `clear_hwpoisoned_pages' defined but not used

And Bisecting it ended up pointing to 4edd7ceff ("mm, hotplug: avoid
compiling memory hotremove functions when disabled").

This is because the commit above put sparse_remove_one_section() within
the protection of CONFIG_MEMORY_HOTREMOVE but the only user of
clear_hwpoisoned_pages() is sparse_remove_one_section(), and it is not
within the protection of CONFIG_MEMORY_HOTREMOVE.

So put clear_hwpoisoned_pages within CONFIG_MEMORY_HOTREMOVE should fix
the warning.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Toshi Kani <toshi.kani@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm: remove unused __put_page()
Zhang Yanfei [Mon, 8 Jul 2013 23:00:09 +0000 (16:00 -0700)]
mm: remove unused __put_page()

This function is nowhere used, and it has a confusing name with put_page
in mm/swap.c.  So better to remove it.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agovfree: don't schedule free_work() if llist_add() returns false
Oleg Nesterov [Mon, 8 Jul 2013 23:00:08 +0000 (16:00 -0700)]
vfree: don't schedule free_work() if llist_add() returns false

vfree() only needs schedule_work(&p->wq) if p->list was empty, otherwise
vfree_deferred->wq is already pending or it is running and didn't do
llist_del_all() yet.

Signed-off-by: Oleg Nesterov <oleg@redhat.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>
11 years agomm/page_alloc.c: remove unlikely() from the current_order test
Zhang Yanfei [Mon, 8 Jul 2013 23:00:08 +0000 (16:00 -0700)]
mm/page_alloc.c: remove unlikely() from the current_order test

In __rmqueue_fallback(), current_order loops down from MAX_ORDER - 1 to
the order passed.  MAX_ORDER is typically 11 and pageblock_order is
typically 9 on x86.  Integer division truncates, so pageblock_order / 2
is 4.  For the first eight iterations, it's guaranteed that
current_order >= pageblock_order / 2 if it even gets that far!

So just remove the unlikely(), it's completely bogus.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Suggested-by: David Rientjes <rientjes@google.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>
11 years agomm: remove unused functions is_{normal_idx, normal, dma32, dma}
Zhang Yanfei [Mon, 8 Jul 2013 23:00:07 +0000 (16:00 -0700)]
mm: remove unused functions is_{normal_idx, normal, dma32, dma}

These functions are nowhere used, so remove them.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/page_alloc.c: remove zone_type argument of build_zonelists_node
Zhang Yanfei [Mon, 8 Jul 2013 23:00:06 +0000 (16:00 -0700)]
mm/page_alloc.c: remove zone_type argument of build_zonelists_node

The callers of build_zonelists_node always pass MAX_NR_ZONES -1 as the
zone_type argument, so we can directly use the value in
build_zonelists_node and remove zone_type argument.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.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>
11 years agoMAINTAINERS: add zswap and zbud maintainer
Seth Jennings [Mon, 8 Jul 2013 23:00:05 +0000 (16:00 -0700)]
MAINTAINERS: add zswap and zbud maintainer

Add maintainer information for zswap and zbud into the MAINTAINERS file.

Signed-off-by: Seth Jennings <sjenning@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoinclude/linux/gfp.h: fix the comment for GFP_ZONE_TABLE
Zhang Yanfei [Mon, 8 Jul 2013 23:00:02 +0000 (16:00 -0700)]
include/linux/gfp.h: fix the comment for GFP_ZONE_TABLE

0xc just means MOVABLE + DMA32, which results in zone DMA32.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: do not account memory used for cache creation
Glauber Costa [Mon, 8 Jul 2013 23:00:01 +0000 (16:00 -0700)]
memcg: do not account memory used for cache creation

The memory we used to hold the memcg arrays is currently accounted to
the current memcg.  But that creates a problem, because that memory can
only be freed after the last user is gone.  Our only way to know which
is the last user, is to hook up to freeing time, but the fact that we
still have some in flight kmallocs will prevent freeing to happen.  I
believe therefore to be just easier to account this memory as global
overhead.

Signed-off-by: Glauber Costa <glommer@openvz.org>
Cc: Johannes Weiner <hannes@cmpxchg.org>
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>
11 years agomemcg: also test for skip accounting at the page allocation level
Glauber Costa [Mon, 8 Jul 2013 23:00:00 +0000 (16:00 -0700)]
memcg: also test for skip accounting at the page allocation level

The memory we used to hold the memcg arrays is currently accounted to
the current memcg.  But that creates a problem, because that memory can
only be freed after the last user is gone.  Our only way to know which
is the last user, is to hook up to freeing time, but the fact that we
still have some in flight kmallocs will prevent freeing to happen.  I
believe therefore to be just easier to account this memory as global
overhead.

This patch (of 2):

Disabling accounting is only relevant for some specific memcg internal
allocations.  Therefore we would initially not have such check at
memcg_kmem_newpage_charge, since direct calls to the page allocator that
are marked with GFP_KMEMCG only happen outside memcg core.  We are
mostly concerned with cache allocations and by having this test at
memcg_kmem_get_cache we are already able to relay the allocation to the
root cache and bypass the memcg caches altogether.

There is one exception, though: the SLUB allocator does not create large
order caches, but rather service large kmallocs directly from the page
allocator.  Therefore, the following sequence, when backed by the SLUB
allocator:

memcg_stop_kmem_account();
kmalloc(<large_number>)
memcg_resume_kmem_account();

would effectively ignore the fact that we should skip accounting, since
it will drive us directly to this function without passing through the
cache selector memcg_kmem_get_cache.  Such large allocations are
extremely rare but can happen, for instance, for the cache arrays.

This was never a problem in practice, because we weren't skipping
accounting for the cache arrays.  All the allocations we were skipping
were fairly small.  However, the fact that we were not skipping those
allocations are a problem and can prevent the memcgs from going away.
As we fix that, we need to make sure that the fix will also work with
the SLUB allocator.

Signed-off-by: Glauber Costa <glommer@openvz.org>
Reported-by: Michal Hocko <mhocko@suze.cz>
Cc: Johannes Weiner <hannes@cmpxchg.org>
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>
11 years agomm/vmalloc.c: check VM_UNINITIALIZED flag in s_show instead of show_numa_info
Zhang Yanfei [Mon, 8 Jul 2013 22:59:59 +0000 (15:59 -0700)]
mm/vmalloc.c: check VM_UNINITIALIZED flag in s_show instead of show_numa_info

We should check the VM_UNITIALIZED flag in s_show().  If this flag is
set, that said, the vm_struct is not fully initialized.  So it is
unnecessary to try to show the information contained in vm_struct.

We checked this flag in show_numa_info(), but I think it's better to
check it earlier.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/vmalloc.c: rename VM_UNLIST to VM_UNINITIALIZED
Zhang Yanfei [Mon, 8 Jul 2013 22:59:58 +0000 (15:59 -0700)]
mm/vmalloc.c: rename VM_UNLIST to VM_UNINITIALIZED

VM_UNLIST was used to indicate that the vm_struct is not listed in
vmlist.

But after commit 4341fa454796 ("mm, vmalloc: remove list management of
vmlist after initializing vmalloc"), the meaning of this flag changed.
It now means the vm_struct is not fully initialized.  So renaming it to
VM_UNINITIALIZED seems more reasonable.

Also change clear_vm_unlist to clear_vm_uninitialized_flag.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/vmalloc.c: emit the failure message before return
Zhang Yanfei [Mon, 8 Jul 2013 22:59:57 +0000 (15:59 -0700)]
mm/vmalloc.c: emit the failure message before return

Use goto to jump to the fail label to give a failure message before
returning NULL.  This makes the failure handling in this function
consistent.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/vmalloc.c: remove alloc_map from vmap_block
Zhang Yanfei [Mon, 8 Jul 2013 22:59:56 +0000 (15:59 -0700)]
mm/vmalloc.c: remove alloc_map from vmap_block

As we have removed the dead code in the vb_alloc, it seems there is no
place to use the alloc_map.  So there is no reason to maintain the
alloc_map in vmap_block.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/vmalloc.c: remove unused purge_fragmented_blocks_thiscpu
Zhang Yanfei [Mon, 8 Jul 2013 22:59:55 +0000 (15:59 -0700)]
mm/vmalloc.c: remove unused purge_fragmented_blocks_thiscpu

This function is nowhere used now, so remove it.

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/vmalloc.c: remove dead code in vb_alloc
Zhang Yanfei [Mon, 8 Jul 2013 22:59:54 +0000 (15:59 -0700)]
mm/vmalloc.c: remove dead code in vb_alloc

Space in a vmap block that was once allocated is considered dirty and
not made available for allocation again before the whole block is
recycled.  The result is that free space within a vmap block is always
contiguous.

So if a vmap block has enough free space for allocation, the allocation
is impossible to fail.  Thus, the fragmented block purging was never
invoked from vb_alloc().  So remove this dead code.

[ Same patches also sent by:

    Chanho Min <chanho.min@lge.com>
    Johannes Weiner <hannes@cmpxchg.org>

  but git doesn't do "multiple authors" ]

Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Cc: Mel Gorman <mel@csn.ul.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm/vmalloc.c: unbreak __vunmap()
Dan Carpenter [Mon, 8 Jul 2013 22:59:53 +0000 (15:59 -0700)]
mm/vmalloc.c: unbreak __vunmap()

There is an extra semi-colon so the function always returns.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm: remove duplicated call of get_pfn_range_for_nid
Zhang Yanfei [Mon, 8 Jul 2013 22:59:52 +0000 (15:59 -0700)]
mm: remove duplicated call of get_pfn_range_for_nid

When calculating pages in a node, for each zone in that node, we will
have

  zone_spanned_pages_in_node
    --> get_pfn_range_for_nid
  zone_absent_pages_in_node
    --> get_pfn_range_for_nid

That is to say, we call the get_pfn_range_for_nid to get start_pfn and
end_pfn of the node for MAX_NR_ZONES * 2 times.  And this is totally
unnecessary if we call the get_pfn_range_for_nid before
zone_*_pages_in_node add two extra arguments node_start_pfn and
node_end_pfn for zone_*_pages_in_node, then we can remove the
get_pfn_range_in_node in zone_*_pages_in_node.

[akpm@linux-foundation.org: make definitions more readable]
Signed-off-by: Zhang Yanfei <zhangyanfei@cn.fujitsu.com>
Cc: Michal Hocko <mhocko@suse.cz>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomm: invoke oom-killer from remaining unconverted page fault handlers
Johannes Weiner [Mon, 8 Jul 2013 22:59:50 +0000 (15:59 -0700)]
mm: invoke oom-killer from remaining unconverted page fault handlers

A few remaining architectures directly kill the page faulting task in an
out of memory situation.  This is usually not a good idea since that
task might not even use a significant amount of memory and so may not be
the optimal victim to resolve the situation.

Since 2.6.29's 1c0fe6e ("mm: invoke oom-killer from page fault") there
is a hook that architecture page fault handlers are supposed to call to
invoke the OOM killer and let it pick the right task to kill.  Convert
the remaining architectures over to this hook.

To have the previous behavior of simply taking out the faulting task the
vm.oom_kill_allocating_task sysctl can be set to 1.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Reviewed-by: Michal Hocko <mhocko@suse.cz>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: David Rientjes <rientjes@google.com>
Acked-by: Vineet Gupta <vgupta@synopsys.com> [arch/arc bits]
Cc: James Hogan <james.hogan@imgtec.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Jonas Bonn <jonas@southpole.se>
Cc: Chen Liqin <liqin.chen@sunplusct.com>
Cc: Lennox Wu <lennox.wu@gmail.com>
Cc: Chris Metcalf <cmetcalf@tilera.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agomemcg: clean up memcg->nodeinfo
Johannes Weiner [Mon, 8 Jul 2013 22:59:49 +0000 (15:59 -0700)]
memcg: clean up memcg->nodeinfo

Remove struct mem_cgroup_lru_info and fold its single member, the
variably sized nodeinfo[0], directly into struct mem_cgroup.  This
should make it more obvious why it has to be the last member there.

Also move the comment that's above that special last member below it, so
it is more visible to somebody that considers appending to the struct
mem_cgroup.

Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
Cc: David Rientjes <rientjes@google.com>
Acked-by: Michal Hocko <mhocko@suse.cz>
Cc: Glauber Costa <glommer@openvz.org>
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>
11 years agomm: mremap: validate input before taking lock
Rasmus Villemoes [Mon, 8 Jul 2013 22:59:48 +0000 (15:59 -0700)]
mm: mremap: validate input before taking lock

This patch is very similar to commit 84d96d897671 ("mm: madvise:
complete input validation before taking lock"): perform some basic
validation of the input to mremap() before taking the
&current->mm->mmap_sem lock.

This also makes the MREMAP_FIXED => MREMAP_MAYMOVE dependency slightly
more explicit.

Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.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>
11 years agodrivers/iommu/msm_iommu_dev.c: fix leak and clean up error paths
Libo Chen [Mon, 8 Jul 2013 22:59:47 +0000 (15:59 -0700)]
drivers/iommu/msm_iommu_dev.c: fix leak and clean up error paths

Fix two obvious problems:

1. We have registered msm_iommu_driver first, and need unregister it
   when registered msm_iommu_ctx_driver fail

2. We don't need to kfree drvdata before kzalloc was successful.

[akpm@linux-foundation.org: remove now-unneeded initialization of ctx_drvdata, remove unneeded braces]
Signed-off-by: Libo Chen <libo.chen@huawei.com>
Acked-by: David Brown <davidb@codeaurora.org>
Cc: David Woodhouse <dwmw2@infradead.org>
Cc: James Hogan <james.hogan@imgtec.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agofsnotify: update comments concerning locking scheme
Lino Sanfilippo [Mon, 8 Jul 2013 22:59:46 +0000 (15:59 -0700)]
fsnotify: update comments concerning locking scheme

There have been changes in the locking scheme of fsnotify but the
comments in the source code have not been updated yet.  This patch
corrects this.

Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Cc: Eric Paris <eparis@redhat.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>
11 years agoinotify: fix race when adding a new watch
Lino Sanfilippo [Mon, 8 Jul 2013 22:59:45 +0000 (15:59 -0700)]
inotify: fix race when adding a new watch

In inotify_new_watch() the number of watches for a group is compared
against the max number of allowed watches and increased afterwards.  The
check and incrementation is not done atomically, so it is possible for
multiple concurrent threads to pass the check and increment the number
of marks above the allowed max.

This patch uses an inotify groups mark_lock to ensure that both check
and incrementation are done atomic.  Furthermore we dont have to worry
about the race that allows a concurrent thread to add a watch just after
inotify_update_existing_watch() returned with -ENOENT anymore, since
this is also synchronized by the groups mark mutex now.

Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Cc: Eric Paris <eparis@redhat.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>
11 years agodnotify: replace dnotify_mark_mutex with mark mutex of dnotify_group
Lino Sanfilippo [Mon, 8 Jul 2013 22:59:44 +0000 (15:59 -0700)]
dnotify: replace dnotify_mark_mutex with mark mutex of dnotify_group

There is no need to use a special mutex to protect against the
fcntl/close race (see dnotify.c for a description of this race).
Instead the dnotify_groups mark mutex can be used.

Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Cc: Eric Paris <eparis@redhat.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>
11 years agofanotify: put duplicate code for adding vfsmount/inode marks into an own function
Lino Sanfilippo [Mon, 8 Jul 2013 22:59:43 +0000 (15:59 -0700)]
fanotify: put duplicate code for adding vfsmount/inode marks into an own function

The code under the groups mark_mutex in fanotify_add_inode_mark() and
fanotify_add_vfsmount_mark() is almost identical.  So put it into a
seperate function.

Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Cc: Eric Paris <eparis@redhat.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>
11 years agofanotify: fix races when adding/removing marks
Lino Sanfilippo [Mon, 8 Jul 2013 22:59:42 +0000 (15:59 -0700)]
fanotify: fix races when adding/removing marks

For both adding an event to an existing mark and destroying a mark we
first have to find it via fsnotify_find_[inode|vfsmount]_mark().  But
getting the mark and adding an event (or destroying it) is not done
atomically.  This opens a race where a thread is about to destroy a mark
while another thread still finds the same mark and adds an event to its
mask although it will be destroyed.

Another race exists concerning the excess of a groups number of marks
limit: When a mark is added the number of group marks is checked against
the max number of marks per group and increased afterwards.  Since check
and increment is also not done atomically, this may result in 2 or more
processes passing the check at the same time and increasing the number
of group marks above the allowed limit.

With this patch both races are avoided by doing the concerning
operations with the groups mark mutex locked.

Signed-off-by: Lino Sanfilippo <LinoSanfilippo@gmx.de>
Cc: Eric Paris <eparis@redhat.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>
11 years agofanotify: info leak in copy_event_to_user()
Dan Carpenter [Mon, 8 Jul 2013 22:59:40 +0000 (15:59 -0700)]
fanotify: info leak in copy_event_to_user()

The ->reserved field isn't cleared so we leak one byte of stack
information to userspace.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Eric Paris <eparis@redhat.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>
11 years agoaudit: Fix decimal constant description
Michal Simek [Mon, 8 Jul 2013 22:59:39 +0000 (15:59 -0700)]
audit: Fix decimal constant description

Use proper decimal type for comparison with u32.

Compilation warning was introduced by 780a7654 ("audit: Make testing for
a valid loginuid explicit.")

  kernel/auditfilter.c: In function 'audit_data_to_entry':
  kernel/auditfilter.c:426:3: warning: this decimal constant is unsigned only in ISO C90 [enabled by default]
     if ((f->type == AUDIT_LOGINUID) && (f->val == 4294967295)) {

Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Eric Paris <eparis@redhat.com>
Acked-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/auditfilter.c: fix leak in audit_add_rule() error path
Chen Gang [Mon, 8 Jul 2013 22:59:38 +0000 (15:59 -0700)]
kernel/auditfilter.c: fix leak in audit_add_rule() error path

If both 'tree' and 'watch' are valid we must call audit_put_tree(), just
like the preceding code within audit_add_rule().

Signed-off-by: Chen Gang <gang.chen@asianux.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Eric Paris <eparis@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agokernel/auditfilter.c: fixing build warning
Raphael S. Carvalho [Mon, 8 Jul 2013 22:59:37 +0000 (15:59 -0700)]
kernel/auditfilter.c: fixing build warning

  kernel/auditfilter.c:426: warning: this decimal constant is unsigned only in ISO C90

Signed-off-by: Raphael S. Carvalho <raphael.scarv@gmail.com>
Cc: Eric Paris <eparis@redhat.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>
11 years agoaudit: fix mq_open and mq_unlink to add the MQ root as a hidden parent audit_names...
Jeff Layton [Mon, 8 Jul 2013 22:59:36 +0000 (15:59 -0700)]
audit: fix mq_open and mq_unlink to add the MQ root as a hidden parent audit_names record

The old audit PATH records for mq_open looked like this:

  type=PATH msg=audit(1366282323.982:869): item=1 name=(null) inode=6777
  dev=00:0c mode=041777 ouid=0 ogid=0 rdev=00:00
  obj=system_u:object_r:tmpfs_t:s15:c0.c1023
  type=PATH msg=audit(1366282323.982:869): item=0 name="test_mq" inode=26732
  dev=00:0c mode=0100700 ouid=0 ogid=0 rdev=00:00
  obj=staff_u:object_r:user_tmpfs_t:s15:c0.c1023

...with the audit related changes that went into 3.7, they now look like this:

  type=PATH msg=audit(1366282236.776:3606): item=2 name=(null) inode=66655
  dev=00:0c mode=0100700 ouid=0 ogid=0 rdev=00:00
  obj=staff_u:object_r:user_tmpfs_t:s15:c0.c1023
  type=PATH msg=audit(1366282236.776:3606): item=1 name=(null) inode=6926
  dev=00:0c mode=041777 ouid=0 ogid=0 rdev=00:00
  obj=system_u:object_r:tmpfs_t:s15:c0.c1023
  type=PATH msg=audit(1366282236.776:3606): item=0 name="test_mq"

Both of these look wrong to me.  As Steve Grubb pointed out:

 "What we need is 1 PATH record that identifies the MQ.  The other PATH
  records probably should not be there."

Fix it to record the mq root as a parent, and flag it such that it
should be hidden from view when the names are logged, since the root of
the mq filesystem isn't terribly interesting.  With this change, we get
a single PATH record that looks more like this:

  type=PATH msg=audit(1368021604.836:484): item=0 name="test_mq" inode=16914
  dev=00:0c mode=0100644 ouid=0 ogid=0 rdev=00:00
  obj=unconfined_u:object_r:user_tmpfs_t:s0

In order to do this, a new audit_inode_parent_hidden() function is
added.  If we do it this way, then we avoid having the existing callers
of audit_inode needing to do any sort of flag conversion if auditing is
inactive.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reported-by: Jiri Jaburek <jjaburek@redhat.com>
Cc: Steve Grubb <sgrubb@redhat.com>
Cc: Eric Paris <eparis@redhat.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>
11 years agodrivers/dma/iop-adma.c: fix new warnings
Olof Johansson [Mon, 8 Jul 2013 22:59:35 +0000 (15:59 -0700)]
drivers/dma/iop-adma.c: fix new warnings

The recent "drivers/dma: remove unused support for MEMSET operations"
change has fallout from lack of build testing by the author.  This
fixes:

  drivers/dma/iop-adma.c:1020:13: warning: unused variable 'dma_addr' [-Wunused-variable]
  drivers/dma/iop-adma.c:1519:2: warning: format '%s' expects a matching 'char *' argument [-Wformat=]

Signed-off-by: Olof Johansson <olof@lixom.net>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
11 years agoMerge branch 'for-linus' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Sun, 7 Jul 2013 18:11:43 +0000 (11:11 -0700)]
Merge branch 'for-linus' of git://git.infradead.org/users/vkoul/slave-dma

Pull slave-dmaengine updates from Vinod Koul:
 "Once you have some time from extended weekend celebrations please
  consider pulling the following to get:
   - Various fixes and PCI driver for dw_dmac by Andy
   - DT binding for imx-dma by Markus & imx-sdma by Shawn
   - DT fixes for dmaengine by Lars
   - jz4740 dmac driver by Lars
   - and various fixes across the drivers"

What "extended weekend celebrations"?  I'm in the merge window, who has
time for extended celebrations..

* 'for-linus' of git://git.infradead.org/users/vkoul/slave-dma: (40 commits)
  DMA: shdma: add DT support
  DMA: shdma: shdma_chan_filter() has to be in shdma-base.h
  DMA: shdma: (cosmetic) don't re-calculate a pointer
  dmaengine: at_hdmac: prepare clk before calling enable
  dmaengine/trivial: at_hdmac: add curly brackets to if/else expressions
  dmaengine: at_hdmac: remove unsuded atc_cleanup_descriptors()
  dmaengine: at_hdmac: add FIFO configuration parameter to DMA DT binding
  ARM: at91: dt: add header to define at_hdmac configuration
  MIPS: jz4740: Correct clock gate bit for DMA controller
  MIPS: jz4740: Remove custom DMA API
  MIPS: jz4740: Register jz4740 DMA device
  dma: Add a jz4740 dmaengine driver
  MIPS: jz4740: Acquire and enable DMA controller clock
  dma: mmp_tdma: disable irq when disabling dma channel
  dmaengine: PL08x: Avoid collisions with get_signal() macro
  dmaengine: dw: select DW_DMAC_BIG_ENDIAN_IO automagically
  dma: dw: add PCI part of the driver
  dma: dw: split driver to library part and platform code
  dma: move dw_dmac driver to an own directory
  dw_dmac: don't check resource with devm_ioremap_resource
  ...

11 years agoMerge branch 'cpuinit-delete' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg...
Linus Torvalds [Sun, 7 Jul 2013 18:01:19 +0000 (11:01 -0700)]
Merge branch 'cpuinit-delete' of git://git./linux/kernel/git/paulg/linux

Pull first stage of __cpuinit removal from Paul Gortmaker:
 "The two commits here 1) dummy out all the __cpuinit macros so that we
  no longer generate such sections, and then 2) remove all the section
  processing that we used to do for those sections.

  This makes all the __cpuinit and friends no-ops, so that we can remove
  the use cases of it at our leisure.  Expect stage 2, which does the
  tree wide removal sweep at the end of the merge window."

* 'cpuinit-delete' of git://git.kernel.org/pub/scm/linux/kernel/git/paulg/linux:
  modpost: remove all traces of cpuinit/cpuexit sections
  init.h: remove __cpuinit sections from the kernel

11 years agoMerge branch 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 6 Jul 2013 21:09:38 +0000 (14:09 -0700)]
Merge branch 'timers-core-for-linus' of git://git./linux/kernel/git/tip/tip

Pull timer core updates from Thomas Gleixner:
 "The timer changes contain:

   - posix timer code consolidation and fixes for odd corner cases

   - sched_clock implementation moved from ARM to core code to avoid
     duplication by other architectures

   - alarm timer updates

   - clocksource and clockevents unregistration facilities

   - clocksource/events support for new hardware

   - precise nanoseconds RTC readout (Xen feature)

   - generic support for Xen suspend/resume oddities

   - the usual lot of fixes and cleanups all over the place

  The parts which touch other areas (ARM/XEN) have been coordinated with
  the relevant maintainers.  Though this results in an handful of
  trivial to solve merge conflicts, which we preferred over nasty cross
  tree merge dependencies.

  The patches which have been committed in the last few days are bug
  fixes plus the posix timer lot.  The latter was in akpms queue and
  next for quite some time; they just got forgotten and Frederic
  collected them last minute."

* 'timers-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (59 commits)
  hrtimer: Remove unused variable
  hrtimers: Move SMP function call to thread context
  clocksource: Reselect clocksource when watchdog validated high-res capability
  posix-cpu-timers: don't account cpu timer after stopped thread runtime accounting
  posix_timers: fix racy timer delta caching on task exit
  posix-timers: correctly get dying task time sample in posix_cpu_timer_schedule()
  selftests: add basic posix timers selftests
  posix_cpu_timers: consolidate expired timers check
  posix_cpu_timers: consolidate timer list cleanups
  posix_cpu_timer: consolidate expiry time type
  tick: Sanitize broadcast control logic
  tick: Prevent uncontrolled switch to oneshot mode
  tick: Make oneshot broadcast robust vs. CPU offlining
  x86: xen: Sync the CMOS RTC as well as the Xen wallclock
  x86: xen: Sync the wallclock when the system time is set
  timekeeping: Indicate that clock was set in the pvclock gtod notifier
  timekeeping: Pass flags instead of multiple bools to timekeeping_update()
  xen: Remove clock_was_set() call in the resume path
  hrtimers: Support resuming with two or more CPUs online (but stopped)
  timer: Fix jiffies wrap behavior of round_jiffies_common()
  ...

11 years agoMerge branch 'for-v3.11' of git://git.linaro.org/people/mszyprowski/linux-dma-mapping
Linus Torvalds [Sat, 6 Jul 2013 19:41:54 +0000 (12:41 -0700)]
Merge branch 'for-v3.11' of git://git.linaro.org/people/mszyprowski/linux-dma-mapping

Pull ARM DMA mapping updates from Marek Szyprowski:
 "This contains important bugfixes and an update for IOMMU integration
  support for ARM architecture"

* 'for-v3.11' of git://git.linaro.org/people/mszyprowski/linux-dma-mapping:
  ARM: dma: Drop __GFP_COMP for iommu dma memory allocations
  ARM: DMA-mapping: mark all !DMA_TO_DEVICE pages in unmapping as clean
  ARM: dma-mapping: NULLify dev->archdata.mapping pointer on detach
  ARM: dma-mapping: convert DMA direction into IOMMU protection attributes
  ARM: dma-mapping: Get pages if the cpu_addr is out of atomic_pool

11 years agoMerge tag 'metag-for-v3.11' of git://git.kernel.org/pub/scm/linux/kernel/git/jhogan...
Linus Torvalds [Sat, 6 Jul 2013 19:39:39 +0000 (12:39 -0700)]
Merge tag 'metag-for-v3.11' of git://git./linux/kernel/git/jhogan/metag

Pull Metag architecture changes from James Hogan:
 - Infrastructure and DT files for TZ1090 SoC (pin control drivers
   already merged via pinctrl tree).
 - Panic on boot instead of just warning if cache aliasing possible.
 - Various SMP/hotplug fixes.
 - Various other randconfig/sparse fixes.

* tag 'metag-for-v3.11' of git://git.kernel.org/pub/scm/linux/kernel/git/jhogan/metag: (24 commits)
  metag: move EXPORT_SYMBOL(csum_partial) to metag_ksyms.c
  metag: cpu hotplug: route_irq: preserve irq mask
  metag: kick: add missing irq_enter/exit to kick_handler()
  metag: smp: don't spin waiting for CPU to start
  metag: smp: enable irqs after set_cpu_online
  metag: use clear_tasks_mm_cpumask()
  metag: tz1090: select and instantiate pinctrl-tz1090-pdc
  metag: tz1090: select and instantiate pinctrl-tz1090
  metag: don't check for cache aliasing on smp cpu boot
  metag: panic if cache aliasing possible
  metag: *.dts: include using preprocessor
  metag: add <dt-bindings/> symlink
  metag/.gitignore: Extend the *.dtb pattern to match the dtb.S files
  metag/traps: include setup.h for the per_cpu_trap_init declaration
  metag/traps: Mark die() as __noreturn to match the declaration.
  metag/processor.h: Add missing cpuinfo_op declaration.
  metag/setup: Restrict scope for the capabilities variable
  metag/mm/cache: Restrict scope for metag_lnkget_probe
  metag/asm/irq.h: Declare init_IRQ
  metag/kernel/irq.c: Declare root_domain as static
  ...

11 years agoMerge tag 'xenarm-for-3.11-tag' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 6 Jul 2013 19:38:42 +0000 (12:38 -0700)]
Merge tag 'xenarm-for-3.11-tag' of git://git./linux/kernel/git/sstabellini/xen

Pull Xen ARM update rom Stefano Stabellini:
 "Just one commit this time: the implementation of the tmem hypercall
  for arm and arm64"

* tag 'xenarm-for-3.11-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/sstabellini/xen:
  xen/arm and xen/arm64: implement HYPERVISOR_tmem_op

11 years agoMerge tag 'irqdomain-for-linus' of git://git.secretlab.ca/git/linux
Linus Torvalds [Sat, 6 Jul 2013 19:37:04 +0000 (12:37 -0700)]
Merge tag 'irqdomain-for-linus' of git://git.secretlab.ca/git/linux

Pull irqdomain refactoring from Grant Likely:
 "This is the long awaited simplification of irqdomain.  It gets rid of
  the different types of irq domains and instead both linear and tree
  mappings can be supported in a single domain.  Doing this removes a
  lot of special case code and makes irq domains simpler to understand
  overall"

* tag 'irqdomain-for-linus' of git://git.secretlab.ca/git/linux:
  irq: fix checkpatch error
  irqdomain: Include hwirq number in /proc/interrupts
  irqdomain: make irq_linear_revmap() a fast path again
  irqdomain: remove irq_domain_generate_simple()
  irqdomain: Refactor irq_domain_associate_many()
  irqdomain: Beef up debugfs output
  irqdomain: Clean up aftermath of irq_domain refactoring
  irqdomain: Eliminate revmap type
  irqdomain: merge linear and tree reverse mappings.
  irqdomain: Add a name field
  irqdomain: Replace LEGACY mapping with LINEAR
  irqdomain: Relax failure path on setting up mappings

11 years agohrtimer: Remove unused variable
Thomas Gleixner [Sat, 6 Jul 2013 08:34:00 +0000 (10:34 +0200)]
hrtimer: Remove unused variable

Sigh, should have noticed myself.

Reported-by: fengguang.wu@intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
11 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Fri, 5 Jul 2013 19:12:33 +0000 (12:12 -0700)]
Merge git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto update from Herbert Xu:
 - Do not idle omap device between crypto operations in one session.
 - Added sha224/sha384 shims for SSSE3.
 - More optimisations for camellia-aesni-avx2.
 - Removed defunct blowfish/twofish AVX2 implementations.
 - Added unaligned buffer self-tests.
 - Added PCLMULQDQ optimisation for CRCT10DIF.
 - Added support for Freescale's DCP co-processor
 - Misc fixes.

* git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6: (44 commits)
  crypto: testmgr - test hash implementations with unaligned buffers
  crypto: testmgr - test AEADs with unaligned buffers
  crypto: testmgr - test skciphers with unaligned buffers
  crypto: testmgr - check that entries in alg_test_descs are in correct order
  Revert "crypto: twofish - add AVX2/x86_64 assembler implementation of twofish cipher"
  Revert "crypto: blowfish - add AVX2/x86_64 implementation of blowfish cipher"
  crypto: camellia-aesni-avx2 - tune assembly code for more performance
  hwrng: bcm2835 - fix MODULE_LICENSE tag
  hwrng: nomadik - use clk_prepare_enable()
  crypto: picoxcell - replace strict_strtoul() with kstrtoul()
  crypto: dcp - Staticize local symbols
  crypto: dcp - Use NULL instead of 0
  crypto: dcp - Use devm_* APIs
  crypto: dcp - Remove redundant platform_set_drvdata()
  hwrng: use platform_{get,set}_drvdata()
  crypto: omap-aes - Don't idle/start AES device between Encrypt operations
  crypto: crct10dif - Use PTR_RET
  crypto: ux500 - Cocci spatch "resource_size.spatch"
  crypto: sha256_ssse3 - add sha224 support
  crypto: sha512_ssse3 - add sha384 support
  ...

11 years agoMerge tag 'upstream-3.11-rc1' of git://git.infradead.org/linux-ubi
Linus Torvalds [Fri, 5 Jul 2013 19:09:48 +0000 (12:09 -0700)]
Merge tag 'upstream-3.11-rc1' of git://git.infradead.org/linux-ubi

Pull ubi fixes from Artem Bityutskiy:
 "A couple of fixes and clean-ups, allow for assigning user-defined UBI
  device numbers when attaching MTD devices by using the "mtd=" module
  parameter"

* tag 'upstream-3.11-rc1' of git://git.infradead.org/linux-ubi:
  UBI: support ubi_num on mtd.ubi command line
  UBI: fastmap break out of used PEB search
  UBI: document UBI_IOCVOLUP better in user header
  UBI: do not abort init when ubi.mtd devices cannot be found
  UBI: drop redundant "UBI error" string

11 years agoMerge tag 'upstream-3.11-rc1' of git://git.infradead.org/linux-ubifs
Linus Torvalds [Fri, 5 Jul 2013 19:08:47 +0000 (12:08 -0700)]
Merge tag 'upstream-3.11-rc1' of git://git.infradead.org/linux-ubifs

Pull ubifs fix from Artem Bityutskiy:
 "Only a single patch which fixes a message"

* tag 'upstream-3.11-rc1' of git://git.infradead.org/linux-ubifs:
  UBIFS: correct mount message

11 years agohrtimers: Move SMP function call to thread context
Thomas Gleixner [Fri, 5 Jul 2013 10:09:18 +0000 (12:09 +0200)]
hrtimers: Move SMP function call to thread context

smp_call_function_* must not be called from softirq context.

But clock_was_set() which calls on_each_cpu() is called from softirq
context to implement a delayed clock_was_set() for the timer interrupt
handler. Though that almost never gets invoked. A recent change in the
resume code uses the softirq based delayed clock_was_set to support
Xens resume mechanism.

linux-next contains a new warning which warns if smp_call_function_*
is called from softirq context which gets triggered by that Xen
change.

Fix this by moving the delayed clock_was_set() call to a work context.

Reported-and-tested-by: Artem Savkov <artem.savkov@gmail.com>
Reported-by: Sasha Levin <sasha.levin@oracle.com>
Cc: David Vrabel <david.vrabel@citrix.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: H. Peter Anvin <hpa@zytor.com>,
Cc: Konrad Wilk <konrad.wilk@oracle.com>
Cc: John Stultz <john.stultz@linaro.org>
Cc: xen-devel@lists.xen.org
Cc: stable@vger.kernel.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
11 years agoclocksource: Reselect clocksource when watchdog validated high-res capability
Thomas Gleixner [Thu, 4 Jul 2013 20:46:45 +0000 (22:46 +0200)]
clocksource: Reselect clocksource when watchdog validated high-res capability

Up to commit 5d33b883a (clocksource: Always verify highres capability)
we had no sanity check when selecting a clocksource, which prevented
that a non highres capable clocksource is used when the system already
switched to highres/nohz mode.

The new sanity check works as Alex and Tim found out. It prevents the
TSC from being used. This happens because on x86 the boot process
looks like this:

 tsc_start_freqency_validation(TSC);
 clocksource_register(HPET);
 clocksource_done_booting();
clocksource_select()
Selects HPET which is valid for high-res

 switch_to_highres();

 clocksource_register(TSC);
  TSC is not selected, because it is not yet
flagged as VALID_HIGH_RES

 clocksource_watchdog()
Validates TSC for highres, but that does not make TSC
the current clocksource.

Before the sanity check was added, we installed TSC unvalidated which
worked most of the time. If the TSC was really detected as unstable,
then the unstable logic removed it and installed HPET again.

The sanity check is correct and needed. So the watchdog needs to kick
a reselection of the clocksource, when it qualifies TSC as a valid
high res clocksource.

To solve this, we mark the clocksource which got the flag
CLOCK_SOURCE_VALID_FOR_HRES set by the watchdog with an new flag
CLOCK_SOURCE_RESELECT and trigger the watchdog thread. The watchdog
thread evaluates the flag and invokes clocksource_select() when set.

To avoid that the clocksource_done_booting() code, which is about to
install the first real clocksource anyway, needs to go through
clocksource_select and tick_oneshot_notify() pointlessly, split out
the clocksource_watchdog_kthread() list walk code and invoke the
select/notify only when called from clocksource_watchdog_kthread().

So clocksource_done_booting() can utilize the same splitout code
without the select/notify invocation and the clocksource_mutex
unlock/relock dance.

Reported-and-tested-by: Alex Shi <alex.shi@intel.com>
Cc: Hans Peter Anvin <hpa@linux.intel.com>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: Andi Kleen <andi.kleen@intel.com>
Tested-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Davidlohr Bueso <davidlohr.bueso@hp.com>
Cc: John Stultz <john.stultz@linaro.org>
Link: http://lkml.kernel.org/r/alpine.DEB.2.02.1307042239150.11637@ionos.tec.linutronix.de
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
11 years agoDMA: shdma: add DT support
Guennadi Liakhovetski [Tue, 18 Jun 2013 16:16:57 +0000 (18:16 +0200)]
DMA: shdma: add DT support

This patch adds Device Tree support to the shdma driver. No special DT
properties are used, only standard DMA DT bindings are implemented. Since
shdma controllers reside on SoCs, their configuration is SoC-specific and
shall be passed to the driver from the SoC platform data, using the
auxdata procedure.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agoDMA: shdma: shdma_chan_filter() has to be in shdma-base.h
Guennadi Liakhovetski [Thu, 6 Jun 2013 15:37:14 +0000 (17:37 +0200)]
DMA: shdma: shdma_chan_filter() has to be in shdma-base.h

shdma_chan_filter() is a function, provided by the shdma-base.c module,
move its declaration to the appropriate header.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agoDMA: shdma: (cosmetic) don't re-calculate a pointer
Guennadi Liakhovetski [Thu, 6 Jun 2013 15:37:13 +0000 (17:37 +0200)]
DMA: shdma: (cosmetic) don't re-calculate a pointer

Use an existing pointer instead of retrieving it again.

Signed-off-by: Guennadi Liakhovetski <g.liakhovetski+renesas@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodmaengine: at_hdmac: prepare clk before calling enable
Boris BREZILLON [Wed, 19 Jun 2013 11:14:54 +0000 (13:14 +0200)]
dmaengine: at_hdmac: prepare clk before calling enable

Replace clk_enable/disable with clk_prepare_enable/disable_unprepare to
avoid common clk framework warnings.

Signed-off-by: Boris BREZILLON <b.brezillon@overkiz.com>
[nicolas.ferre@atmel.com: remove return code checking in at_dma_resume_noirq()]
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
11 years agodmaengine/trivial: at_hdmac: add curly brackets to if/else expressions
Nicolas Ferre [Fri, 7 Jun 2013 15:26:14 +0000 (17:26 +0200)]
dmaengine/trivial: at_hdmac: add curly brackets to if/else expressions

Correct coding style following the patch:
7c407d3e54dcc0c79119553c8d5ef176c1d5bc3a (DMA: AT91:
Get residual bytes in dma buffer).

Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
11 years agodmaengine: at_hdmac: remove unsuded atc_cleanup_descriptors()
Nicolas Ferre [Fri, 7 Jun 2013 15:10:53 +0000 (17:10 +0200)]
dmaengine: at_hdmac: remove unsuded atc_cleanup_descriptors()

Since patch 7c407d3e54dcc0c79119553c8d5ef176c1d5bc3a (DMA: AT91:
Get residual bytes in dma buffer), the function
atc_cleanup_descriptors() is not used anymore. We remove it to prevent
warnings.

Reported-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
11 years agodmaengine: at_hdmac: add FIFO configuration parameter to DMA DT binding
Ludovic Desroches [Thu, 13 Jun 2013 08:39:39 +0000 (10:39 +0200)]
dmaengine: at_hdmac: add FIFO configuration parameter to DMA DT binding

For most devices the FIFO configuration is the same i.e. when half FIFO size is
available/filled, a source/destination request is serviced. But USART devices
have to do it when there is enough space/data available to perform a single
AHB access so the ASAP configuration.

Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Acked-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
11 years agoARM: at91: dt: add header to define at_hdmac configuration
Ludovic Desroches [Thu, 13 Jun 2013 08:39:38 +0000 (10:39 +0200)]
ARM: at91: dt: add header to define at_hdmac configuration

DMA-cell content is a concatenation of several values. In order to keep this
stuff human readable, macros are introduced.

The values for the FIFO configuration are not the same as the ones used in the
configuration register in order to keep backward compatibility. Most devices
use the half FIFO configuration but USART ones have to use the ASAP
configuration. This parameter was not initially planed to be into the at91 dma
dt binding. The third cell will be used to store this parameter, it will
become a concatenation of the FIFO configuration and of the peripheral ID. In
order to keep backward compatibility i.e. FIFO configuration is equal to 0, we
have to perform a translation since the value to put in the register to set
half FIFO is 1.

Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
Signed-off-by: Ludovic Desroches <ludovic.desroches@atmel.com>
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com>
11 years agoMIPS: jz4740: Correct clock gate bit for DMA controller
Maarten ter Huurne [Thu, 30 May 2013 16:25:00 +0000 (18:25 +0200)]
MIPS: jz4740: Correct clock gate bit for DMA controller

Signed-off-by: Maarten ter Huurne <maarten@treewalker.org>
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agoMIPS: jz4740: Remove custom DMA API
Lars-Peter Clausen [Thu, 30 May 2013 16:25:05 +0000 (18:25 +0200)]
MIPS: jz4740: Remove custom DMA API

Now that all users of the custom jz4740 DMA API have been converted to use
the dmaengine API instead we can remove the custom API and move all the code
talking to the hardware to the dmaengine driver.

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agoMIPS: jz4740: Register jz4740 DMA device
Lars-Peter Clausen [Thu, 30 May 2013 16:25:03 +0000 (18:25 +0200)]
MIPS: jz4740: Register jz4740 DMA device

Register a device for the newly added jz4740 dmaengine driver.

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
[manually edited to align struct assignment]
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: Add a jz4740 dmaengine driver
Lars-Peter Clausen [Thu, 30 May 2013 16:25:02 +0000 (18:25 +0200)]
dma: Add a jz4740 dmaengine driver

This patch adds dmaengine support for the JZ4740 DMA controller. For now the
driver will be a wrapper around the custom JZ4740 DMA API. Once all users of the
custom JZ4740 DMA API have been converted to the dmaengine API the custom API
will be removed and direct hardware access will be added to the dmaengine
driver.

Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agoMIPS: jz4740: Acquire and enable DMA controller clock
Maarten ter Huurne [Thu, 30 May 2013 16:25:01 +0000 (18:25 +0200)]
MIPS: jz4740: Acquire and enable DMA controller clock

Previously, it was assumed that the DMA controller clock is not gated
when the kernel starts running. While that is the power-on state, it is
safer to not rely on that.

Signed-off-by: Maarten ter Huurne <maarten@treewalker.org>
Signed-off-by: Lars-Peter Clausen <lars@metafoo.de>
Acked-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: mmp_tdma: disable irq when disabling dma channel
Qiao Zhou [Sat, 15 Jun 2013 04:51:48 +0000 (12:51 +0800)]
dma: mmp_tdma: disable irq when disabling dma channel

mask dma irq when disabling dma channel, so that interrupt status
will not be set and interrupt won't come again.

Signed-off-by: Qiao Zhou <zhouqiao@marvell.com>
Acked-by: Zhangfei Gao <zhangfei.gao@gmail.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodmaengine: PL08x: Avoid collisions with get_signal() macro
Mark Brown [Wed, 19 Jun 2013 19:38:28 +0000 (20:38 +0100)]
dmaengine: PL08x: Avoid collisions with get_signal() macro

As pointed out by Arnd Bergmann there is a get_signal macro definied in
linux/signal.h which can conflict with the platform data callback
function of the same name leading to confusing errors from the compiler
(especially if signal.h manages to get pulled into the driver itself due
to header dependencies).  Avoid such errors by renaming get_signal and
put_signal in the platform data to get_xfer_signal and put_xfer_signal.

Signed-off-by: Mark Brown <broonie@linaro.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodmaengine: dw: select DW_DMAC_BIG_ENDIAN_IO automagically
Vinod Koul [Wed, 12 Jun 2013 08:09:57 +0000 (13:39 +0530)]
dmaengine: dw: select DW_DMAC_BIG_ENDIAN_IO automagically

Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: dw: add PCI part of the driver
Andy Shevchenko [Wed, 5 Jun 2013 12:26:46 +0000 (15:26 +0300)]
dma: dw: add PCI part of the driver

This is the PCI part of the DesignWare DMAC driver. The controller is usually
used in the Intel hardware such as Intel Medfield.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: dw: split driver to library part and platform code
Andy Shevchenko [Wed, 5 Jun 2013 12:26:45 +0000 (15:26 +0300)]
dma: dw: split driver to library part and platform code

To simplify the driver development let's split driver to library and platform
code parts. It helps us to add PCI driver in future.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
[Fixed compile error and few checkpatch issues]
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: move dw_dmac driver to an own directory
Andy Shevchenko [Wed, 5 Jun 2013 12:26:44 +0000 (15:26 +0300)]
dma: move dw_dmac driver to an own directory

The dw_dmac driver is going to be split into multiple files. To make this more
convenient move it to an own directory.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodw_dmac: don't check resource with devm_ioremap_resource
Andy Shevchenko [Wed, 5 Jun 2013 12:26:43 +0000 (15:26 +0300)]
dw_dmac: don't check resource with devm_ioremap_resource

devm_ioremap_resource does sanity checks on the given resource. No need to
duplicate this in the driver.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: pl330: use dma_addr_t for describing bus addresses
Will Deacon [Mon, 10 Jun 2013 18:34:38 +0000 (19:34 +0100)]
dma: pl330: use dma_addr_t for describing bus addresses

The microcode bus address (pl330_dmac.mcode_bus) is currently a u32,
which fails to compile when building on a system with 64-bit bus
addresses.

This patch uses dma_addr_t to represent the address instead.

Cc: Jassi Brar <jaswinder.singh@linaro.org>
Cc: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Acked-by: Jassi Brar <jaswinder.singh@linaro.org>
Acked-by: Grant Likely <grant.likely@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: pl330: rip out broken, redundant ID probing
Will Deacon [Mon, 10 Jun 2013 18:34:37 +0000 (19:34 +0100)]
dma: pl330: rip out broken, redundant ID probing

The PL330 driver probes the peripheral and primecell IDs of the device to
make sure that it is indeed an AMBA PL330. However, it does this by
making byte accesses to a device mapping of the word-aligned ID
registers, which is either UNPREDICTABLE or generates an alignment fault
(depending on the presence of the virtualisation extensions).

Rather than fix this code, we can actually rip most of it out and let
the AMBA bus driver correctly do the probing for us.

Cc: Jassi Brar <jaswinder.singh@linaro.org>
Cc: Vinod Koul <vinod.koul@intel.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Acked-by: Jassi Brar <jaswinder.singh@linaro.org>
Acked-by: Grant Likely <grant.likely@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: imx-sdma: move to generic device tree bindings
Shawn Guo [Thu, 30 May 2013 14:23:32 +0000 (22:23 +0800)]
dma: imx-sdma: move to generic device tree bindings

Update imx-sdma driver to adopt generic DMA device tree bindings.  It
calls of_dma_controller_register() with imx-sdma specific of_dma_xlate
to get the generic DMA device tree helper support.  The #dma-cells for
imx-sdma must be 3, which includes request ID, peripheral type and
priority.

The existing way of requesting channel, clients directly call
dma_request_channel(), still work there, and will be removed after
all imx-sdma clients get converted to generic DMA device tree helper.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
Acked-by: Arnd Bergmann <arnd@arndb.de>
11 years agodmaengine: sirf: set dma residue based on the current dma transfer position
Rongjun Ying [Tue, 14 May 2013 15:03:20 +0000 (23:03 +0800)]
dmaengine: sirf: set dma residue based on the current dma transfer position

read SIRFSOC_DMA_CH_ADDR register to get current dma transfer position, then
update dma residue so that things like ALSA drivers work as ALSA drivers need
the right residue value.

Signed-off-by: Rongjun Ying <Rongjun.Ying@csr.com>
Signed-off-by: Barry Song <Baohua.Song@csr.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>
11 years agodma: tegra: avoid channel lock up after free
Dmitry Osipenko [Sat, 11 May 2013 16:30:53 +0000 (20:30 +0400)]
dma: tegra: avoid channel lock up after free

Lock scenario: Channel 1 was allocated and prepared as slave_sg, used and freed.
Now preparation of cyclic dma on channel 1 will fail with err "DMA configuration
conflict" because tdc->isr_handler still setted to handle_once_dma_done.

This happens because tegra_dma_abort_all() won't be called on channel freeing
if pending list is empty and channel not busy. We need to clear isr_handler
on channel freeing to avoid locking.

Signed-off-by: Dmitry Osipenko <digetx@gmail.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Acked-by: Laxman Dewangan <ldewangan@nvidia.com>
Signed-off-by: Vinod Koul <vinod.koul@intel.com>