GitHub/MotorolaMobilityLLC/kernel-slsi.git
14 years agodm stripe: optimize sector division
Mikulas Patocka [Thu, 12 Aug 2010 03:14:21 +0000 (04:14 +0100)]
dm stripe: optimize sector division

Optimize sector division: If the number of stripes is a power of two,
we can do shift and mask instead of division.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm stripe: move sector translation to a function
Mikulas Patocka [Thu, 12 Aug 2010 03:14:14 +0000 (04:14 +0100)]
dm stripe: move sector translation to a function

Move sector to stripe translation into a function.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: error return error for discards
Mike Snitzer [Thu, 12 Aug 2010 03:14:14 +0000 (04:14 +0100)]
dm: error return error for discards

Have the error target respond to a discard request with a hard -EIO
rather than fail the request with -EOPNOTSUPP.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm delay: support discard
Mike Snitzer [Thu, 12 Aug 2010 03:14:13 +0000 (04:14 +0100)]
dm delay: support discard

Enable discard support for the delay target.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: zero silently drop discards
Mike Snitzer [Thu, 12 Aug 2010 03:14:12 +0000 (04:14 +0100)]
dm: zero silently drop discards

Have the zero target silently drop a discard rather than fail the
request with -EOPNOTSUPP.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: use dm_target_offset macro
Alasdair G Kergon [Thu, 12 Aug 2010 03:14:11 +0000 (04:14 +0100)]
dm: use dm_target_offset macro

Use new dm_target_offset() macro to avoid most references to ti->begin
in dm targets.

Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: factor out max_io_len_target_boundary
Mike Snitzer [Thu, 12 Aug 2010 03:14:10 +0000 (04:14 +0100)]
dm: factor out max_io_len_target_boundary

Split max_io_len_target_boundary out of max_io_len so that the discard
support can make use of it without duplicating max_io_len code.

Avoiding max_io_len's split_io logic enables DM's discard support to
submit the entire discard request to a target.  But discards must still
be split on target boundaries.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: use common __issue_target_request for flush and discard support
Mike Snitzer [Thu, 12 Aug 2010 03:14:09 +0000 (04:14 +0100)]
dm: use common __issue_target_request for flush and discard support

Rename __flush_target to __issue_target_request now that it is used to
issue both flush and discard requests.

Introduce __issue_target_requests as a convenient wrapper to
__issue_target_request 'num_flush_requests' or 'num_discard_requests'
times per target.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: linear support discard
Mike Snitzer [Thu, 12 Aug 2010 03:14:08 +0000 (04:14 +0100)]
dm: linear support discard

Allow discards to be passed through to linear mappings if at least one
underlying device supports it.  Discards will be forwarded only to
devices that support them.

A target that supports discards should set num_discard_requests to
indicate how many times each discard request must be submitted to it.

Verify table's underlying devices support discards prior to setting the
associated DM device as capable of discards (via QUEUE_FLAG_DISCARD).

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Joe Thornber <thornber@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm crypt: simplify crypt_ctr
Milan Broz [Thu, 12 Aug 2010 03:14:07 +0000 (04:14 +0100)]
dm crypt: simplify crypt_ctr

Allocate cipher strings indpendently of struct crypt_config and move
cipher parsing and allocation into a separate function to prepare for
supporting the cryptoapi format e.g. "xts(aes)".

No functional change in this patch.

Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm crypt: simplify crypt_config destruction logic
Milan Broz [Thu, 12 Aug 2010 03:14:06 +0000 (04:14 +0100)]
dm crypt: simplify crypt_config destruction logic

Use just one label and reuse common destructor for crypt target.

Parse remaining argv arguments in logic order.

Also do not ignore error values from IV init and set key functions.

No functional change in this patch except changed return codes
based on above.

Signed-off-by: Milan Broz <mbroz@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: allow autoloading of dm mod
Peter Rajnoha [Thu, 12 Aug 2010 03:14:05 +0000 (04:14 +0100)]
dm: allow autoloading of dm mod

Add devname:mapper/control and MAPPER_CTRL_MINOR module alias
to support dm-mod module autoloading.

Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: Peter Rajnoha <prajnoha@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: rename map_info flush_request to target_request_nr
Mike Snitzer [Thu, 12 Aug 2010 03:14:04 +0000 (04:14 +0100)]
dm: rename map_info flush_request to target_request_nr

'target_request_nr' is a more generic name that reflects the fact that
it will be used for both flush and discard support.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm ioctl: refactor dm_table_complete
Will Drewry [Thu, 12 Aug 2010 03:14:03 +0000 (04:14 +0100)]
dm ioctl: refactor dm_table_complete

This change unifies the various checks and finalization that occurs on a
table prior to use.  By doing so, it allows table construction without
traversing the dm-ioctl interface.

Signed-off-by: Will Drewry <wad@chromium.org>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm snapshot: implement merge
Mikulas Patocka [Thu, 12 Aug 2010 03:14:02 +0000 (04:14 +0100)]
dm snapshot: implement merge

Implement merge method for the snapshot origin to improve read
performance.

Without merge method, dm asks the upper layers to submit smallest possible
bios --- one page. Submitting such small bios impacts performance negatively
when reading or writing the origin device.

Without this patch, CPU consumption when reading the origin on lvm on md-raid0
was 6 to 12%, with this patch, it drops to 1 to 4%.

Note: in my testing, it actually degraded performance in some settings, I
traced it to Maxtor disks having problems with > 512-sector requests.
Reducing the number of sectors to /sys/block/sd*/queue/max_sectors_kb to
256 fixed the read performance. I think we don't have to care about weird
disks that actually degrade performance because of large requests being
sent to them.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: do not initialise full request queue when bio based
Mike Snitzer [Thu, 12 Aug 2010 03:14:02 +0000 (04:14 +0100)]
dm: do not initialise full request queue when bio based

Change bio-based mapped devices no longer to have a fully initialized
request_queue (request_fn, elevator, etc).  This means bio-based DM
devices no longer register elevator sysfs attributes ('iosched/' tree
or 'scheduler' other than "none").

In contrast, a request-based DM device will continue to have a full
request_queue and will register elevator sysfs attributes.  Therefore
a user can determine a DM device's type by checking if elevator sysfs
attributes exist.

First allocate a minimalist request_queue structure for a DM device
(needed for both bio and request-based DM).

Initialization of a full request_queue is deferred until it is known
that the DM device is request-based, at the end of the table load
sequence.

Factor DM device's request_queue initialization:
- common to both request-based and bio-based into dm_init_md_queue().
- specific to request-based into dm_init_request_based_queue().

The md->type_lock mutex is used to protect md->queue, in addition to
md->type, during table_load().

A DM device's first table_load will establish the immutable md->type.
But md->queue initialization, based on md->type, may fail at that time
(because blk_init_allocated_queue cannot allocate memory).  Therefore
any subsequent table_load must (re)try dm_setup_md_queue independently of
establishing md->type.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm ioctl: make bio or request based device type immutable
Mike Snitzer [Thu, 12 Aug 2010 03:14:01 +0000 (04:14 +0100)]
dm ioctl: make bio or request based device type immutable

Determine whether a mapped device is bio-based or request-based when
loading its first (inactive) table and don't allow that to be changed
later.

This patch performs different device initialisation in each of the two
cases.  (We don't think it's necessary to add code to support changing
between the two types.)

Allowed md->type transitions:
  DM_TYPE_NONE to DM_TYPE_BIO_BASED
  DM_TYPE_NONE to DM_TYPE_REQUEST_BASED

We now prevent table_load from replacing the inactive table with a
conflicting type of table even after an explicit table_clear.

Introduce 'type_lock' into the struct mapped_device to protect md->type
and to prepare for the next patch that will change the queue
initialization and allocate memory while md->type_lock is held.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Acked-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
 drivers/md/dm-ioctl.c    |   15 +++++++++++++++
 drivers/md/dm.c          |   37 ++++++++++++++++++++++++++++++-------
 drivers/md/dm.h          |    5 +++++
 include/linux/dm-ioctl.h |    4 ++--
 4 files changed, 52 insertions(+), 9 deletions(-)

14 years agodm: skip second flush on bio unsupported error
Mikulas Patocka [Thu, 12 Aug 2010 03:14:00 +0000 (04:14 +0100)]
dm: skip second flush on bio unsupported error

When processing barriers, skip the second flush if processing the bio
failed with -EOPNOTSUPP.  This can happen with discard+barrier requests.
If the device doesn't support discard, there would be two useless
SYNCHRONIZE CACHE commands.  The first dm_flush cannot be so easily
optimized out, so we leave it there.

Previously, -EOPNOTSUPP could be received in dec_pending only with empty
barriers and we ignored that error, assuming the device not supporting
cache flushes has cache always consistent.  With the addition of discard
barriers, this -EOPNOTSUPP can also be generated by discards and we
must record it in md->barrier_error for process_barrier.

Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm snapshot: persistent use define for disk header chunk size
Tomohiro Kusumi [Thu, 12 Aug 2010 03:13:59 +0000 (04:13 +0100)]
dm snapshot: persistent use define for disk header chunk size

This patch fixes hard-coded value for the size of a chunk that includes
disk header for persistent snapshot. It should be changed to existing
macro NUM_SNAPSHOT_HDR_CHUNKS instead of using hard-coded value 1.

Signed-off-by: Tomohiro Kusumi <kusumi.tomohiro@jp.fujitsu.com>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm crypt: use kstrdup
Julia Lawall [Thu, 12 Aug 2010 03:13:58 +0000 (04:13 +0100)]
dm crypt: use kstrdup

Use kstrdup when the goal of an allocation is copy a string into the
allocated region.

The semantic patch that makes this change is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression from,to;
expression flag,E1,E2;
statement S;
@@

-  to = kmalloc(strlen(from) + 1,flag);
+  to = kstrdup(from, flag);
   ... when != \(from = E1 \| to = E1 \)
   if (to==NULL || ...) S
   ... when != \(from = E2 \| to = E2 \)
-  strcpy(to, from);
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm ioctl: use nonseekable_open
Arnd Bergmann [Thu, 12 Aug 2010 03:13:57 +0000 (04:13 +0100)]
dm ioctl: use nonseekable_open

The dm control device does not implement read/write, so it has no use for
seeking.  Using no_llseek prevents falling back to default_llseek, which
requires the BKL.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: separate device deletion from dm_put
Kiyoshi Ueda [Thu, 12 Aug 2010 03:13:56 +0000 (04:13 +0100)]
dm: separate device deletion from dm_put

This patch separates the device deletion code from dm_put()
to make sure the deletion happens in the process context.

By this patch, device deletion always occurs in an ioctl (process)
context and dm_put() can be called in interrupt context.
As a result, the request-based dm's bad dm_put() usage pointed out
by Mikulas below disappears.
    http://marc.info/?l=dm-devel&m=126699981019735&w=2

Without this patch, I confirmed there is a case to crash the system:
    dm_put() => dm_table_destroy() => vfree() => BUG_ON(in_interrupt())

Some more backgrounds and details:
In request-based dm, a device opener can remove a mapped_device
while the last request is still completing, because bios in the last
request complete first and then the device opener can close and remove
the mapped_device before the last request completes:
  CPU0                                          CPU1
  =================================================================
  <<INTERRUPT>>
  blk_end_request_all(clone_rq)
    blk_update_request(clone_rq)
      bio_endio(clone_bio) == end_clone_bio
        blk_update_request(orig_rq)
          bio_endio(orig_bio)
                                                <<I/O completed>>
                                                dm_blk_close()
                                                dev_remove()
                                                  dm_put(md)
                                                    <<Free md>>
   blk_finish_request(clone_rq)
     ....
     dm_end_request(clone_rq)
       free_rq_clone(clone_rq)
       blk_end_request_all(orig_rq)
       rq_completed(md)

So request-based dm used dm_get()/dm_put() to hold md for each I/O
until its request completion handling is fully done.
However, the final dm_put() can call the device deletion code which
must not be run in interrupt context and may cause kernel panic.

To solve the problem, this patch moves the device deletion code,
dm_destroy(), to predetermined places that is actually deleting
the mapped_device in ioctl (process) context, and changes dm_put()
just to decrement the reference count of the mapped_device.
By this change, dm_put() can be used in any context and the symmetric
model below is introduced:
    dm_create():  create a mapped_device
    dm_destroy(): destroy a mapped_device
    dm_get():     increment the reference count of a mapped_device
    dm_put():     decrement the reference count of a mapped_device

dm_destroy() waits for all references of the mapped_device to disappear,
then deletes the mapped_device.

dm_destroy() uses active waiting with msleep(1), since deleting
the mapped_device isn't performance-critical task.
And since at this point, nobody opens the mapped_device and no new
reference will be taken, the pending counts are just for racing
completing activity and will eventually decrease to zero.

For the unlikely case of the forced module unload, dm_destroy_immediate(),
which doesn't wait and forcibly deletes the mapped_device, is also
introduced and used in dm_hash_remove_all().  Otherwise, "rmmod -f"
may be stuck and never return.
And now, because the mapped_device is deleted at this point, subsequent
accesses to the mapped_device may cause NULL pointer references.

Cc: stable@kernel.org
Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm ioctl: release _hash_lock between devices in remove_all
Kiyoshi Ueda [Thu, 12 Aug 2010 03:13:55 +0000 (04:13 +0100)]
dm ioctl: release _hash_lock between devices in remove_all

This patch changes dm_hash_remove_all() to release _hash_lock when
removing a device.  After removing the device, dm_hash_remove_all()
takes _hash_lock and searches the hash from scratch again.

This patch is a preparation for the next patch, which changes device
deletion code to wait for md reference to be 0.  Without this patch,
the wait in the next patch may cause AB-BA deadlock:
  CPU0                                CPU1
  -----------------------------------------------------------------------
  dm_hash_remove_all()
    down_write(_hash_lock)
                                      table_status()
                                        md = find_device()
                                               dm_get(md)
                                                 <increment md->holders>
                                        dm_get_live_or_inactive_table()
                                          dm_get_inactive_table()
                                            down_write(_hash_lock)
    <in the md deletion code>
      <wait for md->holders to be 0>

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: stable@kernel.org
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm: prevent access to md being deleted
Kiyoshi Ueda [Thu, 12 Aug 2010 03:13:54 +0000 (04:13 +0100)]
dm: prevent access to md being deleted

This patch prevents access to mapped_device which is being deleted.

Currently, even after a mapped_device has been removed from the hash,
it could be accessed through idr_find() using minor number.
That could cause a race and NULL pointer reference below:
  CPU0                          CPU1
  ------------------------------------------------------------------
  dev_remove(param)
    down_write(_hash_lock)
    dm_lock_for_deletion(md)
      spin_lock(_minor_lock)
      set_bit(DMF_DELETING)
      spin_unlock(_minor_lock)
    __hash_remove(hc)
    up_write(_hash_lock)
                                dev_status(param)
                                  md = find_device(param)
                                         down_read(_hash_lock)
                                         __find_device_hash_cell(param)
                                           dm_get_md(param->dev)
                                             md = dm_find_md(dev)
                                                    spin_lock(_minor_lock)
                                                    md = idr_find(MINOR(dev))
                                                    spin_unlock(_minor_lock)
    dm_put(md)
      free_dev(md)
                                             dm_get(md)
                                         up_read(_hash_lock)
                                  __dev_status(md, param)
                                  dm_put(md)

This patch fixes such problems.

Signed-off-by: Kiyoshi Ueda <k-ueda@ct.jp.nec.com>
Signed-off-by: Jun'ichi Nomura <j-nomura@ce.jp.nec.com>
Cc: stable@kernel.org
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm ioctl: return uevent flag after rename
Peter Rajnoha [Thu, 12 Aug 2010 03:13:53 +0000 (04:13 +0100)]
dm ioctl: return uevent flag after rename

All the dm ioctls that generate uevents set the DM_UEVENT_GENERATED flag so
that userspace knows whether or not to wait for a uevent to be processed
before continuing,

The dm rename ioctl sets this flag but was not structured to return it
to userspace.  This patch restructures the rename ioctl processing to
behave like the other ioctls that return data and so fix this.

Signed-off-by: Peter Rajnoha <prajnoha@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm ioctl: make __dev_status void
Alasdair G Kergon [Thu, 12 Aug 2010 03:13:52 +0000 (04:13 +0100)]
dm ioctl: make __dev_status void

__dev_status() cannot fail so make it void and simplify callers.

Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm ioctl: remove __dev_status from geometry and target message
Peter Rajnoha [Thu, 12 Aug 2010 03:13:52 +0000 (04:13 +0100)]
dm ioctl: remove __dev_status from geometry and target message

Remove useless __dev_status call while processing an ioctl that sets up
device geometry and target message.  The data is not returned to
userspace so there is no point collecting it and in the case of
target_message it is collected before processing the message so if it
did return it might be stale.

Signed-off-by: Peter Rajnoha <prajnoha@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm snapshot: test chunk size against both origin and snapshot
Mikulas Patocka [Thu, 12 Aug 2010 03:13:51 +0000 (04:13 +0100)]
dm snapshot: test chunk size against both origin and snapshot

Validate chunk size against both origin and snapshot sector size

Don't allow chunk size smaller than either origin or snapshot logical
sector size. Reading or writing data not aligned to sector size is not
allowed and causes immediate errors.

This requires us to open the origin before initialising the
exception store and to export dm_snap_origin.

Cc: stable@kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm snapshot: iterate origin and cow devices
Mikulas Patocka [Thu, 12 Aug 2010 03:13:50 +0000 (04:13 +0100)]
dm snapshot: iterate origin and cow devices

Iterate both origin and snapshot devices

iterate_devices method should call the callback for all the devices where
the bio may be remapped. Thus, snapshot_iterate_devices should call the callback
for both snapshot and origin underlying devices because it remaps some bios
to the snapshot and some to the origin.

snapshot_iterate_devices called the callback only for the origin device.
This led to badly calculated device limits if snapshot and origin were placed
on different types of disks.

Cc: stable@kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agodm mpath: fix NULL pointer dereference when path parameters missing
Alasdair G Kergon [Thu, 12 Aug 2010 03:13:49 +0000 (04:13 +0100)]
dm mpath: fix NULL pointer dereference when path parameters missing

multipath_ctr() forgets to return an error after detecting
missing path parameters.  Fix this.

Signed-off-by: Patrick LoPresti <lopresti@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Alasdair G Kergon <agk@redhat.com>
14 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
Linus Torvalds [Wed, 11 Aug 2010 16:23:32 +0000 (09:23 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  isofs: Fix lseek() to position beyond 4 GB
  vfs: remove unused MNT_STRICTATIME
  vfs: show unreachable paths in getcwd and proc
  vfs: only add " (deleted)" where necessary
  vfs: add prepend_path() helper
  vfs: __d_path: dont prepend the name of the root dentry
  ia64: perfmon: add d_dname method
  vfs: add helpers to get root and pwd
  cachefiles: use path_get instead of lone dget
  fs/sysv/super.c: add support for non-PDP11 v7 filesystems
  V7: Adjust sanity checks for some volumes
  Add v7 alias
  v9fs: fixup for inode_setattr being removed

Manual merge to take Al's version of the fs/sysv/super.c file: it merged
cleanly, but Al had removed an unnecessary header include, so his side
was better.

14 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/pkl/squashfs-linus
Linus Torvalds [Wed, 11 Aug 2010 16:20:13 +0000 (09:20 -0700)]
Merge git://git./linux/kernel/git/pkl/squashfs-linus

* git://git.kernel.org/pub/scm/linux/kernel/git/pkl/squashfs-linus:
  Squashfs: fix checkpatch.pl warnings
  Squashfs: fix filename typo
  Squashfs: update Kconfig and documentation for LZO
  Squashfs: fix block size use in LZO decompressor
  Squashfs: Add LZO compression support
  squashfs: fix filename in header comment
  Squashfs: Make XATTR config name consistent with other file systems
  squashfs: fix compiler inline warning

14 years agoMerge branch 'for-linus' of git://git.open-osd.org/linux-open-osd
Linus Torvalds [Wed, 11 Aug 2010 16:19:43 +0000 (09:19 -0700)]
Merge branch 'for-linus' of git://git.open-osd.org/linux-open-osd

* 'for-linus' of git://git.open-osd.org/linux-open-osd:
  exofs: Fix groups code when num_devices is not divisible by group_width
  exofs: Remove useless optimization
  exofs: exofs_file_fsync and exofs_file_flush correctness
  exofs: Remove superfluous dependency on buffer_head and writeback

14 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Wed, 11 Aug 2010 16:18:32 +0000 (09:18 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (39 commits)
  ceph: generalize mon requests, add pool op support
  ceph: only queue async writeback on cap revocation if there is dirty data
  ceph: do not ignore osd_idle_ttl mount option
  ceph: constify dentry_operations
  ceph: whitespace cleanup
  ceph: add flock/fcntl lock support
  ceph: define on-wire types, constants for file locking support
  ceph: add CEPH_FEATURE_FLOCK to the supported feature bits
  ceph: support v2 reconnect encoding
  ceph: support v2 client_caps encoding
  ceph: move AES iv definition to shared header
  ceph: fix decoding of pool snap info
  ceph: make ->sync_fs not wait if wait==0
  ceph: warn on missing snap realm
  ceph: print useful error message when crush rule not found
  ceph: use %pU to print uuid (fsid)
  ceph: sync header defs with server code
  ceph: clean up header guards
  ceph: strip misleading/obsolete version, feature info
  ceph: specify supported features in super.h
  ...

14 years agoMerge branch 'msm-video' of git://codeaurora.org/quic/kernel/dwalker/linux-msm
Linus Torvalds [Wed, 11 Aug 2010 16:18:06 +0000 (09:18 -0700)]
Merge branch 'msm-video' of git://codeaurora.org/quic/kernel/dwalker/linux-msm

* 'msm-video' of git://codeaurora.org/quic/kernel/dwalker/linux-msm:
  video: msm: Fix section mismatch in mddi.c.
  drivers: video: msm: drop some unused variables

14 years agoMerge branch 'ixp4xx' of git://git.kernel.org/pub/scm/linux/kernel/git/chris/linux-2.6
Linus Torvalds [Wed, 11 Aug 2010 16:17:27 +0000 (09:17 -0700)]
Merge branch 'ixp4xx' of git://git./linux/kernel/git/chris/linux-2.6

* 'ixp4xx' of git://git.kernel.org/pub/scm/linux/kernel/git/chris/linux-2.6:
  IXP4xx: Fix LL debugging on little-endian CPU.
  IXP4xx: Fix sparse warnings in I/O primitives.
  IXP4xx: Make mdio_bus struct static in the Ethernet driver.
  IXP4xx: Fix ixp4xx_crypto little-endian operation.
  IXP4xx: Prevent HSS transmitter lockup by disabling FRaMe signals.
  ixp4xx/vulcan: add PCI support
  ixp4xx: base support for Arcom Vulcan

14 years agoMerge branch 'for-linus' of master.kernel.org:/home/rmk/linux-2.6-arm
Linus Torvalds [Wed, 11 Aug 2010 16:13:19 +0000 (09:13 -0700)]
Merge branch 'for-linus' of /home/rmk/linux-2.6-arm

* 'for-linus' of master.kernel.org:/home/rmk/linux-2.6-arm: (226 commits)
  ARM: 6323/1: cam60: don't use __init for cam60_spi_{flash_platform_data,partitions}
  ARM: 6324/1: cam60: move cam60_spi_devices to .init.data
  ARM: 6322/1: imx/pca100: Fix name of spi platform data
  ARM: 6321/1: fix syntax error in main Kconfig file
  ARM: 6297/1: move U300 timer to dynamic clock lookup
  ARM: 6296/1: clock U300 intcon and timer properly
  ARM: 6295/1: fix U300 apb_pclk split
  ARM: 6306/1: fix inverted MMC card detect in U300
  ARM: 6299/1: errata: TLBIASIDIS and TLBIMVAIS operations can broadcast a faulty ASID
  ARM: 6294/1: etm: do a dummy read from OSSRR during initialization
  ARM: 6292/1: coresight: add ETM management registers
  ARM: 6288/1: ftrace: document mcount formats
  ARM: 6287/1: ftrace: clean up mcount assembly indentation
  ARM: 6286/1: fix Thumb-2 decompressor broken by "Auto calculate ZRELADDR"
  ARM: 6281/1: video/imxfb.c: allow usage without BACKLIGHT_CLASS_DEVICE
  ARM: 6280/1: imx: Fix build failure when including <mach/gpio.h> without <linux/spinlock.h>
  ARM: S5PV210: Fix on missing s3c-sdhci card detection method for hsmmc3
  ARM: S5P: Fix on missing S5P_DEV_FIMC in plat-s5p/Kconfig
  ARM: S5PV210: Override FIMC driver name on Aquila board
  ARM: S5PC100: enable FIMC on SMDKC100
  ...

Fix up conflicts in arch/arm/mach-{s5pc100,s5pv210}/cpu.c due to
different subsystem 'setname' calls, and trivial port types in
include/linux/serial_core.h

14 years agolib/decompress_bunzip2.c: fix checkstack warning
Prarit Bhargava [Wed, 11 Aug 2010 01:03:40 +0000 (18:03 -0700)]
lib/decompress_bunzip2.c: fix checkstack warning

Fix checkstack error:

lib/decompress_bunzip2.c: In function `get_next_block':
lib/decompress_bunzip2.c:511: warning: the frame size of 1932 bytes is larger than 1024 bytes

byteCount, symToByte, and mtfSymbol cannot be declared static or allocated
dynamically so place them in the bunzip_data struct.

Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agokfifo: add example files to the kernel sample directory
Stefani Seibold [Wed, 11 Aug 2010 01:03:39 +0000 (18:03 -0700)]
kfifo: add example files to the kernel sample directory

Add four examples to the kernel sample directory.

It shows how to handle:
- a byte stream fifo
- a integer type fifo
- a dynamic record sized fifo
- the fifo DMA functions

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Stefani Seibold <stefani@seibold.net>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agokfifo: replace the old non generic API
Stefani Seibold [Wed, 11 Aug 2010 01:03:38 +0000 (18:03 -0700)]
kfifo: replace the old non generic API

Simply replace the whole kfifo.c and kfifo.h files with the new generic
version and fix the kerneldoc API template file.

Signed-off-by: Stefani Seibold <stefani@seibold.net>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agokfifo: add the new generic kfifo API
Stefani Seibold [Wed, 11 Aug 2010 01:03:38 +0000 (18:03 -0700)]
kfifo: add the new generic kfifo API

Add the new version of the kfifo API files kfifo.c and kfifo.h.

Signed-off-by: Stefani Seibold <stefani@seibold.net>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agokfifo: fix kfifo miss use of nozami.c
Stefani Seibold [Wed, 11 Aug 2010 01:03:37 +0000 (18:03 -0700)]
kfifo: fix kfifo miss use of nozami.c

There are different types of a fifo which can not handled in C without a
lot of overhead.  So i decided to write the API as a set of macros, which
is the only way to do a kind of template meta programming without C++.
This macros handles the different types of fifos in a transparent way.

There are a lot of benefits:

- Compile time handling of the different fifo types
- Better performance (a save put or get of an integer does only generate
  9 assembly instructions on a x86)
- Type save
- Cleaner interface, the additional kfifo_..._rec() functions are gone
- Easier to use
- Less error prone
- Different types of fifos: it is now possible to define a int fifo or
  any other type. See below for an example.
- Smaller footprint for none byte type fifos
- No need of creating a second hidden variable, like in the old DEFINE_KFIFO

The API was not changed.

There are now real in place fifos where the data space is a part of the
structure.  The fifo needs now 20 byte plus the fifo space.  Dynamic
assigned or allocated create a little bit more code.

Most of the macros code will be optimized away and simple generate a
function call.  Only the really small one generates inline code.

Additionally you can now create fifos for any data type, not only the
"unsigned char" byte streamed fifos.

There is also a new kfifo_put and kfifo_get function, to handle a single
element in a fifo.  This macros generates inline code, which is lit bit
larger but faster.

I know that this kind of macros are very sophisticated and not easy to
maintain.  But i have all tested and it works as expected.  I analyzed the
output of the compiler and for the x86 the code is as good as hand written
assembler code.  For the byte stream fifo the generate code is exact the
same as with the current kfifo implementation.  For all other types of
fifos the code is smaller before, because the interface is easier to use.

The main goal was to provide an API which is very intuitive, save and easy
to use.  So linux will get now a powerful fifo API which provides all what
a developer needs.  This will save in the future a lot of kernel space,
since there is no need to write an own implementation.  Most of the device
driver developers need a fifo, and also deep kernel development will gain
benefit from this API.

Here are the results of the text section usage:

Example 1:
                        kfifo_put/_get  kfifo_in/out    current kfifo
dynamic allocated       0x000002a8      0x00000291      0x00000299
in place                0x00000291      0x0000026e      0x00000273

kfifo.c                 new             old
text section size       0x00000be5      0x000008b2

As you can see, kfifo_put/kfifo_get creates a little bit more code than
kfifo_in/kfifo_out, but it is much faster (the code is inline).

The code is complete hand crafted and optimized.  The text section size is
as small as possible.  You get all the fifo handling in only 3 kb.  This
includes type safe fix size records, dynamic records and DMA handling.

This should be the final version. All requested features are implemented.

Note: Most features of this API doesn't have any users.  All functions
which are not used in the next 9 months will be removed.  So, please adapt
your drivers and other sources as soon as possible to the new API and post
it.

This are the features which are currently not used in the kernel:

kfifo_to_user()
kfifo_from_user()
kfifo_dma_....() macros
kfifo_esize()
kfifo_recsize()
kfifo_put()
kfifo_get()

The fixed size record elements, exclude "unsigned char" fifo's and the
variable size records fifo's

This patch:

User of the kernel fifo should never bypass the API and directly access
the fifo structure.  Otherwise it will be very hard to maintain the API.

Signed-off-by: Stefani Seibold <stefani@seibold.net>
Cc: Greg KH <greg@kroah.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agokfifo: kfifo_is_{full,empty} should return bools, not ints
Robert P. J. Day [Wed, 11 Aug 2010 01:03:34 +0000 (18:03 -0700)]
kfifo: kfifo_is_{full,empty} should return bools, not ints

For consistency with other kfifo routines, return bool, not int.

Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Cc: Stefani Seibold <stefani@seibold.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agofs/sysv/super.c: add support for non-PDP11 v7 filesystems
Lubomir Rintel [Wed, 11 Aug 2010 01:03:34 +0000 (18:03 -0700)]
fs/sysv/super.c: add support for non-PDP11 v7 filesystems

This adds byte order autodetection (of PDP-11 and LE filesystems).  No
attempt is made to detect big-endian filesystems -- were there any?
Tested with PDP-11 v7 filesystems and PC-IX maintenance floppy.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Cc: Christoph Hellwig <hch@lst.de>
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>
14 years agofs/sysv: v7: adjust sanity checks for some volumes
Lubomir Rintel [Wed, 11 Aug 2010 01:03:33 +0000 (18:03 -0700)]
fs/sysv: v7: adjust sanity checks for some volumes

Newly mkfs-ed filesystems from Seventh Edition have last modification time
set to zero, but are otherwise perfectly valid.

Also, tighten up other sanity checks to filter out most filesystems with

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agofs/sysv: add v7 alias
Lubomir Rintel [Wed, 11 Aug 2010 01:03:32 +0000 (18:03 -0700)]
fs/sysv: add v7 alias

So that the module gets autoloaded when a v7 filesystem is mounted.

Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agokexec: return -EFAULT on copy_to_user() failures
Dan Carpenter [Wed, 11 Aug 2010 01:03:31 +0000 (18:03 -0700)]
kexec: return -EFAULT on copy_to_user() failures

copy_to/from_user() returns the number of bytes remaining to be copied.
It never returns a negative value.  The correct return code is -EFAULT and
not -EIO.

All the callers check for non-zero returns so that's Ok, but the return
code is passed to the user so we should fix this.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Cc: Hidetoshi Seto <seto.hidetoshi@jp.fujitsu.com>
Cc: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Simon Kagstrom <simon.kagstrom@netinsight.net>
Acked-by: WANG Cong <xiyou.wangcong@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoparport_serial: use the PCI IRQ if offered
Fr?d?ric Bri?re [Wed, 11 Aug 2010 01:03:30 +0000 (18:03 -0700)]
parport_serial: use the PCI IRQ if offered

Commit 51dcdfe ("parport: Use the PCI IRQ if offered") added IRQ support
for PCI parallel port devices handled by parport_pc, but turned it off for
parport_serial, despite a printk() message to the contrary.

Signed-off-by: Fr?d?ric Bri?re <fbriere@fbriere.net>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agolib/bug.c: add oops end marker to WARN implementation
Anton Blanchard [Wed, 11 Aug 2010 01:03:30 +0000 (18:03 -0700)]
lib/bug.c: add oops end marker to WARN implementation

We are missing the oops end marker for the exception based WARN implementation
in lib/bug.c. This is useful for logfile analysis tools.

Signed-off-by: Anton Blanchard <anton@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Arjan van de Ven <arjan@infradead.org>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agolib/bug.c: make WARN implementation match the kernel/panic.c one
Anton Blanchard [Wed, 11 Aug 2010 01:03:29 +0000 (18:03 -0700)]
lib/bug.c: make WARN implementation match the kernel/panic.c one

There are a few issues with the exception based WARN implementation in
lib/bug.c:

- Inconsistent printk flags. The "cut here" line is printed at KERN_EMERG, so
  the console and all logged in users see the single line:

------------[ cut here ]------------

  for each WARN. Fix this so we print everything at KERN_WARNING to match the
  kernel/panic.c version.

- The lib/bug.c WARN would print "Badness at". Change it to match the
  kernel/panic.c version which prints "WARNING: at".

- Print the list of modules, similar to kernel/panic.c of modules, similar to
  kernel/panic.c

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Anton Blanchard <anton@samba.org>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Arjan van de Ven <arjan@infradead.org>
Cc: "Kirill A. Shutemov" <kirill@shutemov.name>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agopanic: keep blinking in spite of long spin timer mode
TAMUKI Shoichi [Wed, 11 Aug 2010 01:03:28 +0000 (18:03 -0700)]
panic: keep blinking in spite of long spin timer mode

To keep panic_timeout accuracy when running under a hypervisor, the
current implementation only spins on long time (1 second) calls to mdelay.
 That brings a good effect, but the problem is the keyboard LEDs don't
blink at all on that situation.

This patch changes to call to panic_blink_enter() between every mdelay and
keeps blinking in spite of long spin timer mode.

The time to call to mdelay is now 100ms.  Even this change will keep
panic_timeout accuracy enough when running under a hypervisor.

Signed-off-by: TAMUKI Shoichi <tamuki@linet.gr.jp>
Cc: Ben Dooks <ben-linux@fluff.org>
Cc: Russell King <linux@arm.linux.org.uk>
Acked-by: Dmitry Torokhov <dtor@mail.ru>
Cc: Anton Blanchard <anton@samba.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoafs: destroy work queue on init failure
Dan Carpenter [Wed, 11 Aug 2010 01:03:27 +0000 (18:03 -0700)]
afs: destroy work queue on init failure

We can clean up the work queue on this error path.  This function is
called from afs_init().

Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodma-mapping: add DMA_xxBIT_MASK to feature-removal-schedule.txt
FUJITA Tomonori [Wed, 11 Aug 2010 01:03:26 +0000 (18:03 -0700)]
dma-mapping: add DMA_xxBIT_MASK to feature-removal-schedule.txt

DMA_xxBIT_MASK macros were marked as deprecated in June 2009.  One more
year is long enough, I think.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agopci: add PCI DMA unamp state API to feature-removal-schedule.txt
FUJITA Tomonori [Wed, 11 Aug 2010 01:03:26 +0000 (18:03 -0700)]
pci: add PCI DMA unamp state API to feature-removal-schedule.txt

It was replaced with the DMA unamp state API (which can be used for
any bus).

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation: DMA-API-HOWTO.txt: add multiple types of IOMMUs support
FUJITA Tomonori [Wed, 11 Aug 2010 01:03:25 +0000 (18:03 -0700)]
Documentation: DMA-API-HOWTO.txt: add multiple types of IOMMUs support

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodma-mapping: remove dma_is_consistent API
FUJITA Tomonori [Wed, 11 Aug 2010 01:03:25 +0000 (18:03 -0700)]
dma-mapping: remove dma_is_consistent API

Architectures implement dma_is_consistent() in different ways (some
misinterpret the definition of API in DMA-API.txt).  So it hasn't been so
useful for drivers.  We have only one user of the API in tree.  Unlikely
out-of-tree drivers use the API.

Even if we fix dma_is_consistent() in some architectures, it doesn't look
useful at all.  It was invented long ago for some old systems that can't
allocate coherent memory at all.  It's better to export only APIs that are
definitely necessary for drivers.

Let's remove this API.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoscsi: 53c700: remove dma_is_consistent usage
FUJITA Tomonori [Wed, 11 Aug 2010 01:03:24 +0000 (18:03 -0700)]
scsi: 53c700: remove dma_is_consistent usage

This driver is the only user of dma_is_consistent().  We plan to remove this
API.

The driver uses the API in the following way:

BUG_ON(!dma_is_consistent(hostdata->dev, pScript) && L1_CACHE_BYTES < dma_get_cache_alignment());

The above code tries to see if L1_CACHE_BYTES is greater than
dma_get_cache_alignment() on sysmtes that can not allocate coherent memory
(some old systems can't).

James Bottomley exmplained that this is necesary because the driver packs the
set of mailboxes into a single coherent area and separates the different
usages by a L1 cache stride.  So it's fatal if the dma

He also pointed out that we can kill this checking because we don't hit this
BUG_ON on all architectures that actually use the driver.

(akpm: stolen from the scsi tree because
dma-mapping-remove-dma_is_consistent-api.patch needs it)

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodma-mapping: parisc: set ARCH_DMA_MINALIGN
FUJITA Tomonori [Wed, 11 Aug 2010 01:03:23 +0000 (18:03 -0700)]
dma-mapping: parisc: set ARCH_DMA_MINALIGN

Architectures that handle DMA-non-coherent memory need to set
ARCH_DMA_MINALIGN to make sure that kmalloc'ed buffer is DMA-safe: the
buffer doesn't share a cache with the others.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Kyle McMartin <kyle@mcmartin.ca>
Cc: Helge Deller <deller@gmx.de>
Cc: James E.J. Bottomley <jejb@parisc-linux.org>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodma-mapping: unify dma_get_cache_alignment implementations
FUJITA Tomonori [Wed, 11 Aug 2010 01:03:22 +0000 (18:03 -0700)]
dma-mapping: unify dma_get_cache_alignment implementations

dma_get_cache_alignment returns the minimum DMA alignment.  Architectures
defines it as ARCH_DMA_MINALIGN (formally ARCH_KMALLOC_MINALIGN).  So we
can unify dma_get_cache_alignment implementations.

Note that some architectures implement dma_get_cache_alignment wrongly.
dma_get_cache_alignment() should return the minimum DMA alignment.  So
fully-coherent architectures should return 1.  This patch also fixes this
issue.

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodma-mapping: rename ARCH_KMALLOC_MINALIGN to ARCH_DMA_MINALIGN
FUJITA Tomonori [Wed, 11 Aug 2010 01:03:22 +0000 (18:03 -0700)]
dma-mapping: rename ARCH_KMALLOC_MINALIGN to ARCH_DMA_MINALIGN

Now each architecture has the own dma_get_cache_alignment implementation.

dma_get_cache_alignment returns the minimum DMA alignment.  Architectures
define it as ARCH_KMALLOC_MINALIGN (it's used to make sure that malloc'ed
buffer is DMA-safe; the buffer doesn't share a cache with the others).  So
we can unify dma_get_cache_alignment implementations.

This patch:

dma_get_cache_alignment() needs to know if an architecture defines
ARCH_KMALLOC_MINALIGN or not (needs to know if architecture has DMA
alignment restriction).  However, slab.h define ARCH_KMALLOC_MINALIGN if
architectures doesn't define it.

Let's rename ARCH_KMALLOC_MINALIGN to ARCH_DMA_MINALIGN.
ARCH_KMALLOC_MINALIGN is used only in the internals of slab/slob/slub
(except for crypto).

Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoedac: mpc85xx: add support for new MPCxxx/Pxxxx EDAC controllers
Anton Vorontsov [Wed, 11 Aug 2010 01:03:21 +0000 (18:03 -0700)]
edac: mpc85xx: add support for new MPCxxx/Pxxxx EDAC controllers

Simply add proper IDs into the device table.

Signed-off-by: Anton Vorontsov <avorontsov@mvista.com>
Cc: Scott Wood <scottwood@freescale.com>
Cc: Peter Tyser <ptyser@xes-inc.com>
Cc: Dave Jiang <djiang@mvista.com>
Cc: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoedac: i5400: improve handling of pci_enable_device() return value
Kulikov Vasiliy [Wed, 11 Aug 2010 01:03:20 +0000 (18:03 -0700)]
edac: i5400: improve handling of pci_enable_device() return value

-EIO is not the only error code that pci_enable_device() may return, also
the set of errors can be enhanced in future.  We should compare return
code with zero, not with concrete error value.

Signed-off-by: Kulikov Vasiliy <segooon@gmail.com>
Acked-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Cc: Jeff Roberson <jroberson@jroberson.net>
Cc: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoedac: i5000: improve handling of pci_enable_device() return value
Kulikov Vasiliy [Wed, 11 Aug 2010 01:03:19 +0000 (18:03 -0700)]
edac: i5000: improve handling of pci_enable_device() return value

-EIO is not the only error code that pci_enable_device() may return, also
the set of errors can be enhanced in future.  We should compare return
code with zero, not with concrete error value.

Signed-off-by: Kulikov Vasiliy <segooon@gmail.com>
Acked-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Cc: Jeff Roberson <jroberson@jroberson.net>
Cc: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoedac: add wissing pieces from MPC85xx -> FSL_SOC_BOOKE
Christoph Egger [Wed, 11 Aug 2010 01:03:18 +0000 (18:03 -0700)]
edac: add wissing pieces from MPC85xx -> FSL_SOC_BOOKE

In 5753c082f66eca5be81f6bda85c1718c5eea6ada ("powerpc/85xx: Kconfig
cleanup") menuconfig MPC85xx was replaced by FSL_SOC_BOOKE but some
references insider the code were not adjusted accordingly.  This patch
adresses these missing pieces.

Signed-off-by: Christoph Egger <siccegge@cs.fau.de>
Cc: Doug Thompson <dougthompson@xmission.com>
Cc: Peter Tyser <ptyser@xes-inc.com>
Cc: Kumar Gala <galak@kernel.crashing.org>
Cc: Scott Wood <scottwood@freescale.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agopids: alloc_pidmap: remove the unnecessary boundary checks
Oleg Nesterov [Wed, 11 Aug 2010 01:03:17 +0000 (18:03 -0700)]
pids: alloc_pidmap: remove the unnecessary boundary checks

alloc_pidmap() calculates max_scan so that if the initial offset != 0 we
inspect the first map->page twice.  This is correct, we want to find the
unused bits < offset in this bitmap block.  Add the comment.

But it doesn't make any sense to stop the find_next_offset() loop when we
are looking into this map->page for the second time.  We have already
already checked the bits >= offset during the first attempt, it is fine to
do this again, no matter if we succeed this time or not.

Remove this hard-to-understand code.  It optimizes the very unlikely case
when we are going to fail, but slows down the more likely case.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Salman Qazi <sqazi@google.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agopids: fix a race in pid generation that causes pids to be reused immediately
Salman [Wed, 11 Aug 2010 01:03:16 +0000 (18:03 -0700)]
pids: fix a race in pid generation that causes pids to be reused immediately

A program that repeatedly forks and waits is susceptible to having the
same pid repeated, especially when it competes with another instance of
the same program.  This is really bad for bash implementation.
Furthermore, many shell scripts assume that pid numbers will not be used
for some length of time.

Race Description:

A                                    B

// pid == offset == n                // pid == offset == n + 1
test_and_set_bit(offset, map->page)
                                     test_and_set_bit(offset, map->page);
                                     pid_ns->last_pid = pid;
pid_ns->last_pid = pid;
                                     // pid == n + 1 is freed (wait())

                                     // Next fork()...
                                     last = pid_ns->last_pid; // == n
                                     pid = last + 1;

Code to reproduce it (Running multiple instances is more effective):

#include <errno.h>
#include <sys/types.h>
#include <sys/wait.h>
#include <unistd.h>
#include <stdio.h>
#include <stdlib.h>

// The distance mod 32768 between two pids, where the first pid is expected
// to be smaller than the second.
int PidDistance(pid_t first, pid_t second) {
  return (second + 32768 - first) % 32768;
}

int main(int argc, char* argv[]) {
  int failed = 0;
  pid_t last_pid = 0;
  int i;
  printf("%d\n", sizeof(pid_t));
  for (i = 0; i < 10000000; ++i) {
    if (i % 32786 == 0)
      printf("Iter: %d\n", i/32768);
    int child_exit_code = i % 256;
    pid_t pid = fork();
    if (pid == -1) {
      fprintf(stderr, "fork failed, iteration %d, errno=%d", i, errno);
      exit(1);
    }
    if (pid == 0) {
      // Child
      exit(child_exit_code);
    } else {
      // Parent
      if (i > 0) {
        int distance = PidDistance(last_pid, pid);
        if (distance == 0 || distance > 30000) {
          fprintf(stderr,
                  "Unexpected pid sequence: previous fork: pid=%d, "
                  "current fork: pid=%d for iteration=%d.\n",
                  last_pid, pid, i);
          failed = 1;
        }
      }
      last_pid = pid;
      int status;
      int reaped = wait(&status);
      if (reaped != pid) {
        fprintf(stderr,
                "Wait return value: expected pid=%d, "
                "got %d, iteration %d\n",
                pid, reaped, i);
        failed = 1;
      } else if (WEXITSTATUS(status) != child_exit_code) {
        fprintf(stderr,
                "Unexpected exit status %x, iteration %d\n",
                WEXITSTATUS(status), i);
        failed = 1;
      }
    }
  }
  exit(failed);
}

Thanks to Ted Tso for the key ideas of this implementation.

Signed-off-by: Salman Qazi <sqazi@google.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Theodore Ts'o <tytso@mit.edu>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agopartitions: fix sometimes unreadable partition strings
Alexey Dobriyan [Wed, 11 Aug 2010 01:03:14 +0000 (18:03 -0700)]
partitions: fix sometimes unreadable partition strings

Fix this garbage happening quite often:

==>  sda:
scsi 3:0:0:0: CD-ROM            TOSHIBA
==>  sda1 sda2 sda3 sda4 <sr0: scsi3-mmc drive: 24x/24x writer dvd-ram cd/rw xa/form2 cdda tray
    ^^^
Uniform CD-ROM driver Revision: 3.20
sr 3:0:0:0: Attached scsi CD-ROM sr0
==>  sda5 sda6 sda7 >

Make "sda: sda1 ..." lines actually lines.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agocs5535-mfgpt: reuse timers that have never been set up
Jens Rottmann [Wed, 11 Aug 2010 01:03:12 +0000 (18:03 -0700)]
cs5535-mfgpt: reuse timers that have never been set up

The MFGPT hardware may be set up only once, therefore
cs5535_mfgpt_free_timer() didn't re-set the timer's "avail" bit.  However
if a timer is freed before it has actually been in use then it may be made
available again.

Signed-off-by: Jens Rottmann <JRottmann@LiPPERTEmbedded.de>
Acked-by: Andres Salomon <dilinger@queued.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Jordan Crouse <jordan@cosmicpenguin.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodrivers/char/n_gsm.c: add missing spin_unlock_irqrestore
Julia Lawall [Wed, 11 Aug 2010 01:03:12 +0000 (18:03 -0700)]
drivers/char/n_gsm.c: add missing spin_unlock_irqrestore

Add a spin_unlock_irqrestore missing on the error path.  Converting the
return to break leads to the spin_unlock_irqrestore at the end of the
function.

The semantic match that finds this problem is as follows:
(http://coccinelle.lip6.fr/)

// <smpl>
@@
expression E1;
@@

* spin_lock_irqsave(E1,...);
  <+... when != E1
  if (...) {
    ... when != E1
*   return ...;
  }
  ...+>
* spin_unlock_irqrestore(E1,...);
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Cc: Alan Cox <alan@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoipmi: print info for spmi and smbios paths like acpi and pci
Yinghai Lu [Wed, 11 Aug 2010 01:03:10 +0000 (18:03 -0700)]
ipmi: print info for spmi and smbios paths like acpi and pci

Print out the reg spacing and size for spmi and smbios so BIOS developers
can make them consistent.

Also remove extra PFX on the duplicating path.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Corey Minyard <minyard@acm.org>
Cc: Matthew Garrett <mjg@redhat.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Myron Stowe <myron.stowe@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoipmi: fix memleaking for add_smi when duplicating happen
Yinghai Lu [Wed, 11 Aug 2010 01:03:10 +0000 (18:03 -0700)]
ipmi: fix memleaking for add_smi when duplicating happen

Free the temporary info struct when we have duplicated ones.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>
Cc: Corey Minyard <minyard@acm.org>
Cc: Matthew Garrett <mjg@redhat.com>
Cc: Len Brown <len.brown@intel.com>
Cc: Myron Stowe <myron.stowe@hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodrivers/char/ipmi/ipmi_si_intf.c: fix warning: variable 'addr_space' set but not...
Justin P. Mattock [Wed, 11 Aug 2010 01:03:09 +0000 (18:03 -0700)]
drivers/char/ipmi/ipmi_si_intf.c: fix warning: variable 'addr_space' set but not used

Fix a warning message generated by GCC, and also updates a web address
pointing to a pdf containing information.

CC [M]  drivers/char/ipmi/ipmi_si_intf.o
drivers/char/ipmi/ipmi_si_intf.c: In function 'try_init_spmi':
drivers/char/ipmi/ipmi_si_intf.c:2016:8: warning: variable 'addr_space' set but not used

Signed-off-by: Sergey V. <sftp.mtuci@gmail.com>
Signed-off-by: Justin P. Mattock <justinmattock@gmail.com>
Acked-by: Corey Minyard <minyard@acm.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoprocfs: simplify conditional processing of fs/proc.o.
Robert P. J. Day [Wed, 11 Aug 2010 01:03:08 +0000 (18:03 -0700)]
procfs: simplify conditional processing of fs/proc.o.

Since the entire fs/proc directory is conditionally included based on
CONFIG_PROC_FS, it's redundant to check that same variable within that
directory.

Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agosignalfd: fill in ssi_int for posix timers and message queues
Nathan Lynch [Wed, 11 Aug 2010 01:03:08 +0000 (18:03 -0700)]
signalfd: fill in ssi_int for posix timers and message queues

If signalfd is used to consume a signal generated by a POSIX interval
timer or POSIX message queue, the ssi_int field does not reflect the data
(sigevent->sigev_value) supplied to timer_create(2) or mq_notify(3).  (The
ssi_ptr field, however, is filled in.)

This behavior differs from signalfd's treatment of sigqueue-generated
signals -- see the default case in signalfd_copyinfo.  It also gives
results that differ from the case when a signal is handled conventionally
via a sigaction-registered handler.

So, set signalfd_siginfo->ssi_int in the remaining cases (__SI_TIMER,
__SI_MESGQ) where ssi_ptr is set.

akpm: a non-back-compatible change.  Merge into -stable to minimise the
number of kernels which are in the field and which miss this feature.

Signed-off-by: Nathan Lynch <ntl@pobox.com>
Acked-by: Davide Libenzi <davidel@xmailserver.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoptrace: optimize exit_ptrace() for the likely case
Oleg Nesterov [Wed, 11 Aug 2010 01:03:07 +0000 (18:03 -0700)]
ptrace: optimize exit_ptrace() for the likely case

exit_ptrace() takes tasklist_lock unconditionally.  We need this lock to
avoid the race with ptrace_traceme(), it acts as a barrier.

Change its caller, forget_original_parent(), to call exit_ptrace() under
tasklist_lock.  Change exit_ptrace() to drop and reacquire this lock if
needed.

This allows us to add the fastpath list_empty(ptraced) check.  In the
likely no-tracees case exit_ptrace() just returns and we avoid the lock()
+ unlock() sequence.

"Zhang, Yanmin" <yanmin_zhang@linux.intel.com> suggested to add this
check, and he reports that this change adds about 11% improvement in some
tests.

Suggested-and-tested-by: "Zhang, Yanmin" <yanmin_zhang@linux.intel.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: convert to use zone_to_nid() from bare zone->zone_pgdat->node_id
KOSAKI Motohiro [Wed, 11 Aug 2010 01:03:06 +0000 (18:03 -0700)]
memcg: convert to use zone_to_nid() from bare zone->zone_pgdat->node_id

We have zone_to_nid().  this patch convert all existing users of
zone->zone_pgdat->node_id.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Nishimura Daisuke <d-nishimura@mtf.biglobe.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: remove nid and zid argument from mem_cgroup_soft_limit_reclaim()
KOSAKI Motohiro [Wed, 11 Aug 2010 01:03:05 +0000 (18:03 -0700)]
memcg: remove nid and zid argument from mem_cgroup_soft_limit_reclaim()

mem_cgroup_soft_limit_reclaim() has zone, nid and zid argument.  but nid
and zid can be calculated from zone.  So remove it.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Nishimura Daisuke <d-nishimura@mtf.biglobe.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: mem_cgroup_shrink_node_zone() doesn't need sc.nodemask
KOSAKI Motohiro [Wed, 11 Aug 2010 01:03:05 +0000 (18:03 -0700)]
memcg: mem_cgroup_shrink_node_zone() doesn't need sc.nodemask

Currently mem_cgroup_shrink_node_zone() call shrink_zone() directly.  thus
it doesn't need to initialize sc.nodemask because shrink_zone() doesn't
use it at all.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Nishimura Daisuke <d-nishimura@mtf.biglobe.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: kill unnecessary initialization in mem_cgroup_shrink_node_zone()
KOSAKI Motohiro [Wed, 11 Aug 2010 01:03:04 +0000 (18:03 -0700)]
memcg: kill unnecessary initialization in mem_cgroup_shrink_node_zone()

sc.nr_reclaimed and sc.nr_scanned have already been initialized few lines
above "struct scan_control sc = {}" statement.

So, This patch remove this unnecessary code.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Mel Gorman <mel@csn.ul.ie>
Cc: Nishimura Daisuke <d-nishimura@mtf.biglobe.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: sc.nr_to_reclaim should be initialized
KOSAKI Motohiro [Wed, 11 Aug 2010 01:03:02 +0000 (18:03 -0700)]
memcg: sc.nr_to_reclaim should be initialized

Currently, mem_cgroup_shrink_node_zone() initialize sc.nr_to_reclaim as 0.
 It mean shrink_zone() only scan 32 pages and immediately return even if
it doesn't reclaim any pages.

This patch fixes it.

Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: Mel Gorman <mel@csn.ul.ie>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Nishimura Daisuke <d-nishimura@mtf.biglobe.ne.jp>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: avoid css_get()
KAMEZAWA Hiroyuki [Wed, 11 Aug 2010 01:03:02 +0000 (18:03 -0700)]
memcg: avoid css_get()

Now, memory cgroup increments css(cgroup subsys state)'s reference count
per a charged page.  And the reference count is kept until the page is
uncharged.  But this has 2 bad effect.

 1. Because css_get/put calls atomic_inc()/dec, heavy call of them
    on large smp will not scale well.
 2. Because css's refcnt cannot be in a state as "ready-to-release",
    cgroup's notify_on_release handler can't work with memcg.
 3. css's refcnt is atomic_t, it means smaller than 32bit. Maybe too small.

This has been a problem since the 1st merge of memcg.

This is a trial to remove css's refcnt per a page. Even if we remove
refcnt, pre_destroy() does enough synchronization as
  - check res->usage == 0.
  - check no pages on LRU.

This patch removes css's refcnt per page.  Even after this patch, at the
1st look, it seems css_get() is still called in try_charge().

But the logic is.

  - If a memcg of mm->owner is cached one, consume_stock() will work.
    At success, return immediately.
  - If consume_stock returns false, css_get() is called and go to
    slow path which may be blocked. At the end of slow path,
    css_put() is called and restart from the start if necessary.

So, in the fast path, we don't call css_get() and can avoid access to
shared counter. This patch can make the most possible case fast.

Here is a result of multi-threaded page fault benchmark.

[Before]
    25.32%  multi-fault-all  [kernel.kallsyms]      [k] clear_page_c
     9.30%  multi-fault-all  [kernel.kallsyms]      [k] _raw_spin_lock_irqsave
     8.02%  multi-fault-all  [kernel.kallsyms]      [k] try_get_mem_cgroup_from_mm <=====(*)
     7.83%  multi-fault-all  [kernel.kallsyms]      [k] down_read_trylock
     5.38%  multi-fault-all  [kernel.kallsyms]      [k] __css_put
     5.29%  multi-fault-all  [kernel.kallsyms]      [k] __alloc_pages_nodemask
     4.92%  multi-fault-all  [kernel.kallsyms]      [k] _raw_spin_lock_irq
     4.24%  multi-fault-all  [kernel.kallsyms]      [k] up_read
     3.53%  multi-fault-all  [kernel.kallsyms]      [k] css_put
     2.11%  multi-fault-all  [kernel.kallsyms]      [k] handle_mm_fault
     1.76%  multi-fault-all  [kernel.kallsyms]      [k] __rmqueue
     1.64%  multi-fault-all  [kernel.kallsyms]      [k] __mem_cgroup_commit_charge

[After]
    28.41%  multi-fault-all  [kernel.kallsyms]      [k] clear_page_c
    10.08%  multi-fault-all  [kernel.kallsyms]      [k] _raw_spin_lock_irq
     9.58%  multi-fault-all  [kernel.kallsyms]      [k] down_read_trylock
     9.38%  multi-fault-all  [kernel.kallsyms]      [k] _raw_spin_lock_irqsave
     5.86%  multi-fault-all  [kernel.kallsyms]      [k] __alloc_pages_nodemask
     5.65%  multi-fault-all  [kernel.kallsyms]      [k] up_read
     2.82%  multi-fault-all  [kernel.kallsyms]      [k] handle_mm_fault
     2.64%  multi-fault-all  [kernel.kallsyms]      [k] mem_cgroup_add_lru_list
     2.48%  multi-fault-all  [kernel.kallsyms]      [k] __mem_cgroup_commit_charge

Then, 8.02% of try_get_mem_cgroup_from_mm() disappears because this patch
removes css_tryget() in it. (But yes, this is an extreme case.)

Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: use find_lock_task_mm() in memory cgroups oom
KAMEZAWA Hiroyuki [Wed, 11 Aug 2010 01:03:00 +0000 (18:03 -0700)]
memcg: use find_lock_task_mm() in memory cgroups oom

When the OOM killer scans task, it check a task is under memcg or
not when it's called via memcg's context.

But, as Oleg pointed out, a thread group leader may have NULL ->mm
and task_in_mem_cgroup() may do wrong decision. We have to use
find_lock_task_mm() in memcg as generic OOM-Killer does.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Reviewed-by: Minchan Kim <minchan.kim@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: remove mem from arg of charge_common
Daisuke Nishimura [Wed, 11 Aug 2010 01:02:59 +0000 (18:02 -0700)]
memcg: remove mem from arg of charge_common

mem_cgroup_charge_common() is always called with @mem = NULL, so it's
meaningless.  This patch removes it.

Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: remove redundant code
Daisuke Nishimura [Wed, 11 Aug 2010 01:02:59 +0000 (18:02 -0700)]
memcg: remove redundant code

- try_get_mem_cgroup_from_mm() calls rcu_read_lock/unlock by itself, so we
  don't have to call them in task_in_mem_cgroup().
- *mz is not used in __mem_cgroup_uncharge_common().
- we don't have to call lookup_page_cgroup() in mem_cgroup_end_migration()
  after we've cleared PCG_MIGRATION of @oldpage.
- remove empty comment.
- remove redundant empty line in mem_cgroup_cache_charge().

Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: clean up waiting move acct
KAMEZAWA Hiroyuki [Wed, 11 Aug 2010 01:02:58 +0000 (18:02 -0700)]
memcg: clean up waiting move acct

Now, for checking a memcg is under task-account-moving, we do css_tryget()
against mc.to and mc.from.  But this is just complicating things.  This
patch makes the check easier.

This patch adds a spinlock to move_charge_struct and guard modification of
mc.to and mc.from.  By this, we don't have to think about complicated
races arount this not-critical path.

[balbir@linux.vnet.ibm.com: don't crash on a null memcg being passed]
Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: clean up try_charge main loop
KAMEZAWA Hiroyuki [Wed, 11 Aug 2010 01:02:57 +0000 (18:02 -0700)]
memcg: clean up try_charge main loop

mem_cgroup_try_charge() has a big loop in it and seems to be hard to read.
 Most of routines are for slow path.  This patch moves codes out from the
loop and make it clear what's done.

Summary:
 - refactoring a function to detect a memcg is under acccount move or not.
 - refactoring a function to wait for the end of moving task acct.
 - refactoring a main loop('s slow path) as a function and make it clear
   why we retry or quit by return code.
 - add fatal_signal_pending() check for bypassing charge loops.

Signed-off-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: Balbir Singh <balbir@in.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomemcg: remove experimental from swap account config
KAMEZAWA Hiroyuki [Wed, 11 Aug 2010 01:02:56 +0000 (18:02 -0700)]
memcg: remove experimental from swap account config

It's 11 months since we changed swap_map[] to indicates SWAP_HAS_CACHE.
Since that, memcg's swap accounting has been very stable and it seems
it can be maintained.

So, I'd like to remove EXPERIMENTAL from the config.

Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Acked-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
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>
14 years agoblkdev: cgroup whitelist permission fix
Chris Wright [Wed, 11 Aug 2010 01:02:55 +0000 (18:02 -0700)]
blkdev: cgroup whitelist permission fix

The cgroup device whitelist code gets confused when trying to grant
permission to a disk partition that is not currently open.  Part of
blkdev_open() includes __blkdev_get() on the whole disk.

Basically, the only ways to reliably allow a cgroup access to a partition
on a block device when using the whitelist are to 1) also give it access
to the whole block device or 2) make sure the partition is already open in
a different context.

The patch avoids the cgroup check for the whole disk case when opening a
partition.

Addresses https://bugzilla.redhat.com/show_bug.cgi?id=589662

Signed-off-by: Chris Wright <chrisw@sous-sol.org>
Acked-by: Serge E. Hallyn <serue@us.ibm.com>
Tested-by: Serge E. Hallyn <serue@us.ibm.com>
Reported-by: Vivek Goyal <vgoyal@redhat.com>
Cc: Al Viro <viro@ZenIV.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: "Daniel P. Berrange" <berrange@redhat.com>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agocgroups: save space for the terminator
Dan Carpenter [Wed, 11 Aug 2010 01:02:54 +0000 (18:02 -0700)]
cgroups: save space for the terminator

The original code didn't leave enough space for a NULL terminator.  These
strings are copied with strcpy() into fixed length buffers in
cgroup_root_from_opts().

Signed-off-by: Dan Carpenter <error27@gmail.com>
Acked-by: Serge E. Hallyn <serge@hallyn.com>
Reviewd-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Ben Blum <bblum@andrew.cmu.edu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation/padata.txt: fix typos etc.
Randy Dunlap [Wed, 11 Aug 2010 01:02:53 +0000 (18:02 -0700)]
Documentation/padata.txt: fix typos etc.

Fix typos & grammar.
Use CPU instead of cpu in text.

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Acked-by: Steffen Klassert <steffen.klassert@secunet.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation/00-INDEX: remove reference to exception.txt
Huang Shijie [Wed, 11 Aug 2010 01:02:52 +0000 (18:02 -0700)]
Documentation/00-INDEX: remove reference to exception.txt

The exception.txt has been removed from the Documentation directory.  So
update the index file for it.

Signed-off-by: Huang Shijie <shijie8@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodocbook: need xmldoclinks for all doc types
Ben Hutchings [Wed, 11 Aug 2010 01:02:51 +0000 (18:02 -0700)]
docbook: need xmldoclinks for all doc types

$ rm -rf build
$ mkdir build
$ cp .config build
$ make O=build htmldocs
...
xmlto: linux-2.6/build/Documentation/DocBook/media.xml
does not validate (status 3)
xmlto: Fix document syntax or use --skip-validation option
linux-2.6/build/Documentation/DocBook/media.xml:4:
warning: failed to load external entity
"linux-2.6/build/Documentation/DocBook/media-entities.tmpl"

We need the xmldoclinks built for any document types built from the
XML sources.

Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Acked-by: Andy Whitcroft <apw@canonical.com>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoDocumentation/networking/wavelan.txt: deleted, not in tree
Joe Perches [Wed, 11 Aug 2010 01:02:50 +0000 (18:02 -0700)]
Documentation/networking/wavelan.txt: deleted, not in tree

Commit 1d794e3b353b ("Staging: wavelan: delete the driver") removed the
source, so remove the documentation as well.

Signed-off-by: Joe Perches <joe@perches.com>
Cc: Jean Tourrilhes <jt@hpl.hp.com>
Acked-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agomtd/nand_base: fix kernel-doc warnings & typos
Randy Dunlap [Wed, 11 Aug 2010 01:02:50 +0000 (18:02 -0700)]
mtd/nand_base: fix kernel-doc warnings & typos

Fix mtd/nand_base.c kernel-doc warnings and typos.

Warning(drivers/mtd/nand/nand_base.c:893): No description found for parameter 'mtd'
Warning(drivers/mtd/nand/nand_base.c:893): No description found for parameter 'ofs'
Warning(drivers/mtd/nand/nand_base.c:893): No description found for parameter 'len'
Warning(drivers/mtd/nand/nand_base.c:893): No description found for parameter 'invert'
Warning(drivers/mtd/nand/nand_base.c:930): No description found for parameter 'mtd'
Warning(drivers/mtd/nand/nand_base.c:930): No description found for parameter 'ofs'
Warning(drivers/mtd/nand/nand_base.c:930): No description found for parameter 'len'
Warning(drivers/mtd/nand/nand_base.c:987): No description found for parameter 'mtd'
Warning(drivers/mtd/nand/nand_base.c:987): No description found for parameter 'ofs'
Warning(drivers/mtd/nand/nand_base.c:987): No description found for parameter 'len'
Warning(drivers/mtd/nand/nand_base.c:2087): No description found for parameter 'len'

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: David Woodhouse <dwmw2@infradead.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agofusion: fix kernel-doc warnings
Randy Dunlap [Wed, 11 Aug 2010 01:02:48 +0000 (18:02 -0700)]
fusion: fix kernel-doc warnings

Fix (delete) empty kernel-doc lines/warnings:
Warning(drivers/message/fusion/mptbase.c:6916): bad line:
Warning(drivers/message/fusion/mptbase.c:7060): bad line:

Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Cc: Eric Moore <Eric.Moore@lsi.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoreiserfs: remove unused local `wait'
Changli Gao [Wed, 11 Aug 2010 01:02:48 +0000 (18:02 -0700)]
reiserfs: remove unused local `wait'

Signed-off-by: Changli Gao <xiaosuo@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodrivers/video/igafb.c: make igafb_setup() and igafb_init() static
Andrew Morton [Wed, 11 Aug 2010 01:02:47 +0000 (18:02 -0700)]
drivers/video/igafb.c: make igafb_setup() and igafb_init() static

Cc: Kulikov Vasiliy <segooon@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agodrivers/video/igafb.c: introduce lost 'return'
Kulikov Vasiliy [Wed, 11 Aug 2010 01:02:47 +0000 (18:02 -0700)]
drivers/video/igafb.c: introduce lost 'return'

If iga_init() fails, code releases resources and continues to use it.  It
seems that after releasing resources 'return' should be.

Signed-off-by: Kulikov Vasiliy <segooon@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agofbmem: VM_IO set, but not propagated
Daniel De Graaf [Wed, 11 Aug 2010 01:02:45 +0000 (18:02 -0700)]
fbmem: VM_IO set, but not propagated

When we setup up the VMA flags for the mmap flag and we end up using the
fallback mmap functionality we set the vma->vm_flags |= VM_IO.  However we
neglect to propagate the flag to the vma->vm_page_prot.

This bug was found when Linux kernel was running under Xen.  In that
scenario, any page that has VM_IO flag to it, means that it MUST be a
MMIO/VRAM backend memory , _not_ System RAM.  That is what the fbmem.c
does: sets VM_IO, ioremaps the region - everything is peachy.

Well, not exactly.  The vm_page_prot does not get the relevant PTE flags
set (_PAGE_IOMAP) which under Xen is a death-kneel to pages that are
referencing real physical devices but don't have that flag set.

This patch fixes this.

Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Tested-by: Eamon Walsh <ewalsh@tycho.nsa.gov>
Cc: Florian Tobias Schandinat <FlorianSchandinat@gmx.de>
Cc: Dave Airlie <airlied@linux.ie>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
14 years agoarm: samsung: remove pixclock from several boards
Maurus Cuelenaere [Wed, 11 Aug 2010 01:02:44 +0000 (18:02 -0700)]
arm: samsung: remove pixclock from several boards

Since "s3c-fb: Automatically calculate pixel clock when none is given",
there's no need for manually calculating the pixel clock anymore so remove
these lines and add the correct refresh rate where appropriately.

Signed-off-by: Maurus Cuelenaere <mcuelenaere@gmail.com>
Cc: Pawel Osciak <p.osciak@samsung.com>
Cc: Marek Szyprowski <m.szyprowski@samsung.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Cc: InKi Dae <inki.dae@samsung.com>
Cc: Ben Dooks <ben-linux@fluff.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>