In userspace, the .lzma format has become mostly a legacy file format that
got superseded by the .xz format.  Similarly, LZMA Utils was superseded by
XZ Utils.

These patches add support for XZ decompression into the kernel.  Most of
the code is as is from XZ Embedded <http://tukaani.org/xz/embedded.html>.
It was written for the Linux kernel but is usable in other projects too.

Advantages of XZ over the current LZMA code in the kernel:
  - Nice API that can be used by other kernel modules; it's
    not limited to kernel, initramfs, and initrd decompression.
  - Integrity check support (CRC32)
  - BCJ filters improve compression of executable code on
    certain architectures. These together with LZMA2 can
    produce a few percent smaller kernel or Squashfs images
    than plain LZMA without making the decompression slower.

This patch: Add the main decompression code (xz_dec), testing module
(xz_dec_test), wrapper script (xz_wrap.sh) for the xz command line tool,
and documentation.  The xz_dec module is enough to have a usable XZ
decompressor e.g.  for Squashfs.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Callback-to-callback decompression mode is used for initrd (not
initramfs).  The LZO wrapper is broken for this use case for two reasons:

  - The argument validation is needlessly too strict by
    requiring that "posp" is non-NULL when "fill" is non-NULL.

  - The buffer handling code didn't work at all for this
    use case.

I tested with LZO-compressed kernel, initramfs, initrd, and corrupt
(truncated) initramfs and initrd images.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The code assumes that the input is valid and not truncated.  Add checks to
avoid reading past the end of the input buffer.  Change the type of "skip"
from u8 to int to fix a possible integer overflow.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The return value of flush() is not checked in unlzo().  This means that
the decompressor won't stop even if the caller doesn't want more data.
This can happen e.g.  with a corrupt LZO-compressed initramfs image.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Validate the newly decoded distance (rep0) in process_bit1().  This is to
detect corrupt LZMA data quickly.  The old code can run for long time
producing garbage until it hits the end of the input.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The return value of wr->flush() is not checked in write_byte().  This
means that the decompressor won't stop even if the caller doesn't want
more data.  This can happen e.g.  with corrupt LZMA-compressed initramfs.
Returning the error quickly allows the user to see the error message
quicker.

There is a similar missing check for wr.flush() near the end of unlzma().
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Return value of rc->fill() is checked in rc_read() and error() is called
when needed, but then the code continues as if nothing had happened.

rc_read() is a void function and it's on the top of performance critical
call stacks, so propagating the error code via return values doesn't sound
like the best fix.  It seems better to check rc->buffer_size (which holds
the return value of rc->fill()) in the main loop.  It does nothing bad
that the code runs a little with unknown data after a failed rc->fill().

This fixes an infinite loop in initramfs decompression if the
LZMA-compressed initramfs image is corrupt.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Validation of header.pos calls error() but doesn't make the function
return to indicate an error to the caller.  Instead the decoding is
attempted with invalid header.pos.  This fixes it.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Currently users of mm.h need to include <linux/slab.h> to use the macros
malloc() and free() provided by mm.h.  This fixes it.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

set_error_fn() has become a useless complication after c1e7c3ae
("bzip2/lzma/gzip: pre-boot malloc doesn't return NULL on failure") fixed
the use of error() in malloc().  Only decompress_unlzma.c had some use for
it and that was easy to change too.

This also gets rid of the static function pointer "error", which
should have been marked as __initdata.
Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NLasse Collin <lasse.collin@tukaani.org>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Alain Knaff <alain@knaff.lu>
Cc: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This header uses things like __be32, so pull in linux/types.h.

Further, it uses BLOCK_SIZE, so pull in linux/fs.h.
Signed-off-by: NMike Frysinger <vapier@gentoo.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Cc: Ahmed S. Darwish <darwish.07@gmail.com>
Cc: Kyungmin Park <kyungmin.park@samsung.com>
Cc: Marco Stornelli <marco.stornelli@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Generate a unique inode numbers for any entries in the cram file system.
For files which did not contain data's (device nodes, fifos and sockets)
the offset of the directory entry inside the cramfs plus 1 will be used as
inode number.

The + 1 for the inode will it make possible to distinguish between a file
which contains no data and files which has data, the later one has a inode
value where the lower two bits are always 0.

It also reimplements the behavior to set the size and the number of block
to 0 for special file, which is the right value for empty files, devices,
fifos and sockets

As a little benefit it will be also more compatible which older mkcramfs,
because it will never use the cramfs_inode->offset for creating a inode
number for special files.

[akpm@linux-foundation.org: trivial comment fix]
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: NStefani Seibold <stefani@seibold.net>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Pekka Enberg <penberg@kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Currently, 3 kernel function prototypes are present in a header
file exported to userland. This patch fixes it.
Signed-off-by: NAlexander Shishkin <virtuoso@slind.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

aio_run_iocbs() is not used at all, so get rid of it.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: NJeff Moyer <jmoyer@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

'nr >= min_nr >= 0' always satisfies 'nr >= 0' so the check is unnecesary.
Signed-off-by: NNamhyung Kim <namhyung@gmail.com>
Acked-by: NJeff Moyer <jmoyer@redhat.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

When hypervisor decides to decrease target balloon size while the balloon
driver tries to lock pages hypervisor may respond with
VMW_BALLOON_PPN_NOTNEEDED.  Use this data and immediately stop reserving
pages and wait for the next update cycle to fetch new target instead of
continuing trying to lock pages until size of refused list grows above
VMW_BALLOON_MAX_REFUSED (16) pages.

As a result the driver stops bothering the hypervisor with its attempts to
lock more pages that are not needed anymore.  Most likely next order from
hypervisor will be to reduce ballon size anyway.

It is a small optimization.
Signed-off-by: NDmitry Torokhov <dtor@vmware.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This is a 1-wire/w1 DS2423 slave driver for reading the values from all 4
counters available DS2423 devices by using standard w1_slave file.  In
ds2423 the counters are tied to ram pages 12-15 in and each of those
ram-pages.  Each of these counter values (and asoociated ram page values)
are represented as a own line in w1_slave file.  Driver has been tested on
mips and x86.

usage example:
cat /sys/bus/w1/devices/1d-00000009b964/w1_slave

00 02 00 00 00 00 00 00 00 6d 38 00 ff ff 00 00 fe ff 00 00 ff ff 00 00 ff
ff 00 00 ff ff 00 00 ff ff 00 00 ff ff 00 00 ff ff crc=YES c=2
00 02 00 00 00 00 00 00 00 e0 1f 00 ff ff 00 00 ff ff 00 00 ff ff 00 00 ff
ff 00 00 ff ff 00 00 ff ff 00 00 ff ff 00 00 ff ff crc=YES c=2
00 5a 0e 5f 18 00 00 00 00 0b 28 00 ff ff 00 00 ff ff 00 00 ff ff 00 00 ff
ff 00 00 ff ff 00 00 ff ff 00 00 ff ff 00 00 ff ff crc=YES c=408882778
00 05 00 00 00 00 00 00 00 8d 39 ff ff ff ff ff ff ff ff ff ff ff ff ff ff
ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff ff crc=YES c=5

Patch includes also the documentation.

[randy.dunlap@oracle.com: fix ds2423 build, needs to select CRC16]
Signed-off-by: NMika Laitio <lamikr@pilppa.org>
Signed-off-by: NRandy Dunlap <randy.dunlap@oracle.com>
Cc: Evgeniy Polyakov <johnpol@2ka.mipt.ru>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Apart from currently used standard memstick data transfer method, Sony
introduced several newer ones, to uncover full bandwidth/capacity of its
Pro, HG and XC media formats. This patch lays a foundation to enable
those methods as made possible by host/media capabilities.

As a side effect of this patch, mspro_block_read_attributes became more
streamlined and readable.

[akpm@linux-foundation.org: fix printk warning]
Signed-off-by: NAlex Dubov <oakad@yahoo.com>
Reported-by: NMaxim Levitsky <maximlevitsky@gmail.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

mspro_block_mutex is identical in scope to mspro_block_disk_lock and therefore
unnecessary.
Signed-off-by: NAlex Dubov <oakad@yahoo.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Implement the usual pattern around idr_pre_get() and idr_get_new() to
handlethe situation where another thread concurrently steals this thread's
idr_pre_get() preallocation.
Signed-off-by: NAlex Dubov <oakad@yahoo.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NAries Lee <arieslee@jmicron.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>
Cc: Alex Dubov <oakad@yahoo.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Add a function jmb38x_ms_pmos() to enable / disable PMOS setups for
JMicron 38x controllers.
Signed-off-by: NAries Lee <arieslee@jmicron.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>
Cc: Alex Dubov <oakad@yahoo.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This patch corrects the definition of clock values for JMicron 38x
controllers and sets the value properly per interface type.
Also, it adds a check for TPC errors in the interrupt handler.
Signed-off-by: NAries Lee <arieslee@jmicron.com>
Signed-off-by: NTakashi Iwai <tiwai@suse.de>
Cc: Alex Dubov <oakad@yahoo.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

If device_register() fails then call put_device().  See comment to
device_register.
Signed-off-by: NVasiliy Kulikov <segooon@gmail.com>
Cc: Maxim Levitsky <maximlevitsky@gmail.com>
Cc: Alex Dubov <oakad@yahoo.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Add PPS signal generator which utilizes STROBE pin of a parallel port to
send PPS signals.  It uses parport abstraction layer and hrtimers to
precisely control the signal.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Cc: Rodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Add PPS signal generator which utilizes STROBE pin of a parallel port to
send PPS signals.  It uses parport abstraction layer and hrtimers to
precisely control the signal.

[akpm@linux-foundation.org: fix build]
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Cc: john stultz <johnstul@us.ibm.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Add parallel port PPS client.  It uses a standard method for capturing
timestamps for assert edge transitions: getting a timestamp soon after an
interrupt has happened.

This is not a very precise source of time information due to interrupt
handling delays.  However, timestamps for clear edge transitions are much
more precise because the interrupt handler continuously polls hardware
port until the transition is done.

Hardware port operations require only about 1us so the maximum error
should not exceed this value.  This was my primary goal when developing
this client.

Clear edge capture could be disabled using clear_wait parameter.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Add an optional feature of PPSAPI, kernel consumer support, which uses the
added hardpps() function.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

MONOTONIC_RAW clock timestamps are ideally suited for frequency
calculation and also fit well into the original NTP hardpps design.  Now
phase and frequency can be adjusted separately: the former based on
REALTIME clock and the latter based on MONOTONIC_RAW clock.

A new function getnstime_raw_and_real is added to timekeeping subsystem to
capture both timestamps at the same time and atomically.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NJohn Stultz <johnstul@us.ibm.com>
Cc: Rodolfo Giometti <giometti@enneenne.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This commit adds hardpps() implementation based upon the original one from
the NTPv4 reference kernel code from David Mills.  However, it is highly
optimized towards very fast syncronization and maximum stickness to PPS
signal.  The typical error is less then a microsecond.

To make it sync faster I had to throw away exponential phase filter so
that the full phase offset is corrected immediately.  Then I also had to
throw away median phase filter because it gives a bigger error itself if
used without exponential filter.

Maybe we will find an appropriate filtering scheme in the future but it's
not necessary if the signal quality is ok.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NJohn Stultz <johnstul@us.ibm.com>
Cc: Rodolfo Giometti <giometti@enneenne.com>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Remove the code that gatheres timestamp in pps_tty_dcd_change() in case
passed ts parameter is NULL because it never happens in the current code.
Fix comments as well.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Bitwise conjunction is distributive so we can simplify some conditions.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This way less overhead is involved when running production kernel.  If you
want to debug a pps client module please define DEBUG to enable the
checks.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Now pps_idr_lock is never used in interrupt context so we can replace
spin_lock_irq/spin_unlock_irq with plain spin_lock/spin_unlock. But
there is also a potential race condition when someone can steal an id
which was allocated by idr_pre_get before it is used. So convert spin
lock to mutex and protect the whole id generation process.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Cc: Rodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Since now idr is only used to manage char device id's and not used in
kernel API anymore it should be moved to pps.c.  This also makes it
possible to release id only at actual device freeing so nobody can
register a pps device with the same id while our device is not freed yet.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Since we now have direct pointers to struct pps_device everywhere it's
easy to use dev_* functions to print messages instead of plain printks.
Where dev_* cannot be used printks are converted to pr_*.
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Using device index as a pointer needs some unnecessary work to be done
every time the pointer is needed (in irq handler for example).  Using a
direct pointer is much more easy (and safe as well).
Signed-off-by: NAlexander Gordeev <lasaine@lvk.cs.msu.su>
Acked-by: NRodolfo Giometti <giometti@linux.it>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>