drivers/rtc/rtc-s3c.c: fix return value in s3c_rtc_probe()

Return the value returned by platform_get_irq() instead of -ENOENT;

Signed-off-by: Sachin Kamat <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc_sysfs_show_hctosys(): display 0 if resume failed

Without this patch /sys/class/rtc/$CONFIG_RTC_HCTOSYS_DEVICE/hctosys
contains a 1 (meaning "This rtc was used to initialize the system
clock") even if setting the time by do_settimeofday() at bootup failed.
The RTC can also be used to set the clock on resume, if it did 1,
otherwise 0.  Previously there was no indication if the RTC was used
to set the clock in resume.

This uses only CONFIG_RTC_HCTOSYS_DEVICE for conditional compilation
instead of it and CONFIG_RTC_HCTOSYS to be more consistent.
rtc_hctosys_ret was moved to class.c so class.c no longer depends on
hctosys.c.

[[email protected]: fix build]
Signed-off-by: David Fries <[email protected]>
Cc: Matthew Garrett <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Cc: Uwe Kleine-König <[email protected]>
Signed-off-by: Stephen Rothwell <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: rtc-mxc: convert to module_platform_driver

Converting to module_platform_driver can make the code smaller and cleaner.

Signed-off-by: Fabio Estevam <[email protected]>
Cc: Sascha Hauer <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: rtc-mxc: adapt to the new i.mx clock framework

i.mx drivers should use clk_prepare_enable/clk_disable_unprepare() in
order to avoid clk warnings.

While at it, convert to devm_clk_get() since other devm_ functions are
used in this driver and it can also save some clk_put() calls.

Signed-off-by: Fabio Estevam <[email protected]>
Acked-by: Sascha Hauer <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/rtc/rtc-coh901331.c: use clk_prepare_enable() and clk_disable_unprepare()

clk_prepare_enable and clk_disable_unprepare combine clk_prepare and
clk_enable, and clk_disable and clk_unprepare.  They make the code more
concise, and ensure that clk_unprepare is called when clk_enable fails.

A simplified version of the semantic patch that introduces calls to these
functions is as follows: (http://coccinelle.lip6.fr/)

// <smpl>
@@
expression e;
@@

- clk_prepare(e);
- clk_enable(e);
+ clk_prepare_enable(e);

@@
expression e;
@@

- clk_disable(e);
- clk_unprepare(e);
+ clk_disable_unprepare(e);
// </smpl>

Signed-off-by: Julia Lawall <[email protected]>
Acked-by: Linus Walleij <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: rc5t583: add ricoh rc5t583 RTC driver

Add an RTC driver for the RTC device on Ricoh MFD Rc5t583.  Ricoh RTC has
3 types of alarms.  The current patch adds support for the Y-Alarm of
RC5t583 RTC.

Signed-off-by: Venu Byravarasu <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Cc: Mark Brown <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/rtc/rtc-spear.c: fix several error checks

There are several comparisons of a unsigned int to less than zero int
spear RTC driver.  Such a check will always be true.  In all these cases a
signed int is assigned to the unsigned variable, which is checked, before.
 So the right fix is to make the checked variable signed as well.  In one
case the check can be dropped completely, because all it does it returns
'err' if 'err' is less than zero, otherwise it returns 0.  Since in this
particular case 'err' is always either 0 or less this is the same as just
returning 'err'.

The issue has been found using the following coccinelle semantic patch:

//<smpl>
@@
type T;
unsigned T i;
@@
(
*i < 0
|
*i >= 0
)
//</smpl>

Signed-off-by: Lars-Peter Clausen <[email protected]>
Cc: Viresh Kumar <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/rtc/rtc-jz4740.c: fix IRQ error check

The irq field of the jz4740_irc struct is unsigned.  Yet we assign the
result of platform_get_irq() to it.  platform_get_irq() may return a
negative error code and the code checks for this condition by checking if
'irq' is less than zero.  But since 'irq' is unsigned this test will
always be false.  Fix it by making 'irq' signed.

The issue was found using the following coccinelle semantic patch:

//<smpl>
@@
type T;
unsigned T i;
@@
(
*i < 0
|
*i >= 0
)
//</smpl>

Signed-off-by: Lars-Peter Clausen <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: kconfig: remove unnecessary dependencies.

Remove "depends on RTC_CLASS = y" for multiple Kconfig definitions, as all
of them are already placed under "if RTC_CLASS".

Signed-off-by: Venu Byravarasu <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: add MAX8907 RTC driver

The MAX8907 is an I2C-based power-management IC containing voltage
regulators, a reset controller, a real-time clock, and a touch-screen
controller.

The driver is based on an original by or fixed by:
* Tom Cherry
* Prashant Gaikwad
* Joseph Yoon

During upstreaming, I (swarren):
* Converted to regmap.
* Fixed handling of RTC_HOUR register containing 12.
* Fixed handling of RTC_WEEKDAY register.
* General cleanup.

Signed-off-by: Stephen Warren <[email protected]>
Cc: Tom Cherry <[email protected]>
Cc: Prashant Gaikwad <[email protected]>
Cc: Joseph Yoon <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: tps65910: add RTC driver for TPS65910 PMIC RTC

TPS65910 PMIC is a MFD with RTC as one of the device.  Adding RTC driver
for supporting RTC device present inside TPS65910 PMIC.

Only support for RTC alarm is implemented as part of this patch.

Signed-off-by: Venu Byravarasu <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Cc: Mark Brown <[email protected]>
Cc: Samuel Ortiz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/rtc/rtc-at91sam9.c: use module_platform_driver() macro

This driver does seems to do only platform_driver_register in the init
function and platform_driver_unregister in the exit function,

so replace all this code including the module_init and module_exit with
module_platform_driver macro...

Signed-off-by: Devendra Naga <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: recycle id when unloading a rtc driver

When calling rtc_device_unregister, we are not freeing the id used by the
driver.  So when doing a unload/load cycle for a RTC driver (e.g.  rmmod
rtc_cmos && modprobe rtc_cmos), its id is incremented by one.  As a
consequence, we no longer have neither an rtc0 driver nor a
/proc/driver/rtc (as it only exists for the first driver).

Signed-off-by: Vincent Palatin <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: snvs: add Freescale rtc-snvs driver

Add an RTC driver for Freescale Secure Non-Volatile Storage (SNVS)
Low Power (LP) RTC.

Signed-off-by: Shawn Guo <[email protected]>
Reviewed-by: Stephen Boyd <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Cc: Kim Phillips <[email protected]>
Cc: Sascha Hauer <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc: add Dallas DS2404 driver

[[email protected]: fix warning]
Signed-off-by: Sven Schnelle <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

rtc-proc: permit the /proc/driver/rtc device to use other devices

To get time information via /proc/driver/rtc, only the first device (rtc0)
is used.  If the rtcN (eg.  rtc1 or rtc2) is used for the system clock,
there is no way to get information of rtcN via /proc/driver/rtc.  With
this patch, the time data can be retrieved from the system clock RTC.

If the RTC_HCTOSYS_DEVICE is not defined, then rtc0 is used by default.

Signed-off-by: Milo(Woogyom) Kim <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/rtc/rtc-isl1208.c: add support for the ISL1218

The ISL1218 chip is identical to the ISL1208, except that it has 6
additional user-storage registers.  This patch does not enable access to
those additional registers, but only adds the chip name to the list.

Signed-off-by: Ben Gardner <[email protected]>
Cc: Alessandro Zummo <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

binfmt_elf: Uninitialized variable

load_elf_interp() has interp_map_addr carefully described as
"uninitialized_var" and marked so as to avoid a warning.  However if you
trace the code it is passed into load_elf_interp and then this value is
checked against NULL.

As this return value isn't used this is actually safe but it freaks
various analysis tools that see un-initialized memory addresses being read
before their value is ever defined.

Set it to NULL as a matter of programming good taste if nothing else

Signed-off-by: Alan Cox <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

epoll: support for disabling items, and a self-test app

Enhanced epoll_ctl to support EPOLL_CTL_DISABLE, which disables an epoll
item.  If epoll_ctl doesn't return -EBUSY in this case, it is then safe to
delete the epoll item in a multi-threaded environment.  Also added a new
test_epoll self- test app to both demonstrate the need for this feature
and test it.

Signed-off-by: Paton J. Lewis <[email protected]>
Cc: Alexander Viro <[email protected]>
Cc: Jason Baron <[email protected]>
Cc: Paul Holland <[email protected]>
Cc: Davide Libenzi <[email protected]>
Cc: Michael Kerrisk <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

checkpatch: fix name of "MODULE_PARM_DESC"

Fix macro name in checkpatch: s/PARAM/PARM/.

Signed-off-by: Kees Cook <[email protected]>
Cc: Joe Perches <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

CodingStyle: add networking specific block comment style

The block comment style in net/ and drivers/net is non-standard.
Document it.

Signed-off-by: Joe Perches <[email protected]>
Cc: "Allan, Bruce W" <[email protected]>
Cc: Andy Whitcroft <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

checkpatch: check networking specific block comment style

In an effort to get fewer checkpatch reviewer corrections, add a
networking specific style test for the preferred networking comment style.

/* The preferred style for block comments in
 * drivers/net/... and net/... is like this
 */

These tests are only used in net/ and drivers/net/

Tested with:

$ cat drivers/net/t.c

/* foo */

/*
 * foo
 */

/* foo
 */

/* foo
 * bar */
$ ./scripts/checkpatch.pl -f drivers/net/t.c
WARNING: networking block comments don't use an empty /* line, use /* Comment...
#4: FILE: net/t.c:4:
+
+/*

WARNING: networking block comments put the trailing */ on a separate line
#12: FILE: net/t.c:12:
+ * bar */

total: 0 errors, 2 warnings, 12 lines checked

Signed-off-by: Joe Perches <[email protected]>
Cc: "Allan, Bruce W" <[email protected]>
Cc: Andy Whitcroft <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

checkpatch: update suggested printk conversions

Direct conversion of printk(KERN_<LEVEL>...  to pr_<level> isn't the
preferred conversion when a struct net_device or struct device is
available.

Hint that using netdev_<level> or dev_<level> is preferred to using
pr_<level>.  Add netdev_dbg and dev_dbg variants too.

Miscellaneous whitespace neatening of a misplaced close brace.

Signed-off-by: Joe Perches <[email protected]>
Cc: Andy Whitcroft <[email protected]>
Cc: Greg KH <[email protected]>
Cc: Dan Carpenter <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

checkpatch: check utf-8 content from a commit log when it's missing from charset

Check that a commit log doesn't contain UTF-8 when a mail header
explicitly defines a different charset, like

'Content-Type: text/plain; charset="us-ascii"'

Signed-off-by: Pasi Savanainen <[email protected]>
Cc: Joe Perches <[email protected]>
Cc: Andy Whitcroft <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

scatterlist: atomic sg_mapping_iter() no longer needs disabled IRQs

SG mapping iterator w/ SG_MITER_ATOMIC set required IRQ disabled because
it originally used KM_BIO_SRC_IRQ to allow use from IRQ handlers.
kmap_atomic() has long been updated to handle stacking atomic mapping
requests on per-cpu basis and only requires not sleeping while mapped.

Update sg_mapping_iter such that atomic iterators only require disabling
preemption instead of disabling IRQ.

While at it, convert wte weird @ARG@ notations to @ARG in the comment of
sg_miter_start().

Signed-off-by: Tejun Heo <[email protected]>
Cc: Maxim Levitsky <[email protected]>
Cc: Alex Dubov <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/plist.c: make plist test announcements KERN_DEBUG

They show up in dmesg

[    4.041094] start plist test
[    4.045804] end plist test

without a lot of meaning so hide them behind debug loglevel.

Signed-off-by: Borislav Petkov <[email protected]>
Cc: Lai Jiangshan <[email protected]>
Acked-by: Steven Rostedt <[email protected]>
Cc: Paul Gortmaker <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/vsprintf.c: improve standard conformance of sscanf()

Xen's pciback points out a couple of deficiencies with vsscanf()'s
standard conformance:

- Trailing character matching cannot be checked by the caller: With a
  format string of "(%x:%x.%x) %n" absence of the closing parenthesis
  cannot be checked, as input of "(00:00.0)" doesn't cause the %n to be
  evaluated (because of the code not skipping white space before the
  trailing %n).

- The parameter corresponding to a trailing %n could get filled even if
  there was a matching error: With a format string of "(%x:%x.%x)%n",
  input of "(00:00.0]" would still fill the respective variable pointed to
  (and hence again make the mismatch non-detectable by the caller).

This patch aims at fixing those, but leaves other non-conforming aspects
of it untouched, among them these possibly relevant ones:

- improper handling of the assignment suppression character '*' (blindly
  discarding all succeeding non-white space from the format and input
  strings),

- not honoring conversion specifiers for %n, - not recognizing the C99
  conversion specifier 't' (recognized by vsprintf()).

Signed-off-by: Jan Beulich <[email protected]>
Cc: Konrad Rzeszutek Wilk <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/spinlock_debug: avoid livelock in do_raw_spin_lock()

The logic in do_raw_spin_lock() attempts to acquire a spinlock by invoking
arch_spin_trylock() in a loop with a delay between each attempt.  Now
consider the following situation in a 2 CPU system:

1. CPU-0 continually acquires and releases a spinlock in a
   tight loop; it stays in this loop until some condition X
   is satisfied. X can only be satisfied by another CPU.

2. CPU-1 tries to acquire the same spinlock, in an attempt
   to satisfy the aforementioned condition X. However, it
   never sees the unlocked value of the lock because the
   debug spinlock code uses trylock instead of just lock;
   it checks at all the wrong moments - whenever CPU-0 has
   locked the lock.

Now in the absence of debug spinlocks, the architecture specific spinlock
code can correctly allow CPU-1 to wait in a "queue" (e.g., ticket
spinlocks), ensuring that it acquires the lock at some point.  However,
with the debug spinlock code, livelock can easily occur due to the use of
try_lock, which obviously cannot put the CPU in that "queue".  This
queueing mechanism is implemented in both x86 and ARM spinlock code.

Note that the situation mentioned above is not hypothetical.  A real
problem was encountered where CPU-0 was running hrtimer_cancel with
interrupts disabled, and CPU-1 was attempting to run the hrtimer that
CPU-0 was trying to cancel.

Address this by actually attempting arch_spin_lock once it is suspected
that there is a spinlock lockup.  If we're in a situation that is
described above, the arch_spin_lock should succeed; otherwise other
timeout mechanisms (e.g., watchdog) should alert the system of a lockup.
Therefore, if there is a genuine system problem and the spinlock can't be
acquired, the end result (irrespective of this change being present) is
the same.  If there is a livelock caused by the debug code, this change
will allow the lock to be acquired, depending on the implementation of the
lower level arch specific spinlock code.

[[email protected]: tweak comment]
Signed-off-by: Vikram Mulukutla <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

genalloc: make it possible to use a custom allocation algorithm

Premit use of another algorithm than the default first-fit one.  For
example a custom algorithm could be used to manage alignment requirements.

As I can't predict all the possible requirements/needs for all allocation
uses cases, I add a "free" field 'void *data' to pass any needed
information to the allocation function.  For example 'data' could be used
to handle a structure where you store the alignment, the expected memory
bank, the requester device, or any information that could influence the
allocation algorithm.

An usage example may look like this:
struct my_pool_constraints {
int align;
int bank;
...
};

unsigned long my_custom_algo(unsigned long *map, unsigned long size,
unsigned long start, unsigned int nr, void *data)
{
struct my_pool_constraints *constraints = data;
...
deal with allocation contraints
...
return the index in bitmap where perform the allocation
}

void create_my_pool()
{
struct my_pool_constraints c;
struct gen_pool *pool = gen_pool_create(...);
gen_pool_add(pool, ...);
gen_pool_set_algo(pool, my_custom_algo, &c);
}

Add of best-fit algorithm function:
most of the time best-fit is slower then first-fit but memory fragmentation
is lower. The random buffer allocation/free tests don't show any arithmetic
relation between the allocation time and fragmentation but the
best-fit algorithm
is sometime able to perform the allocation when the first-fit can't.

This new algorithm help to remove static allocations on ESRAM, a small but
fast on-chip RAM of few KB, used for high-performance uses cases like DMA
linked lists, graphic accelerators, encoders/decoders. On the Ux500
(in the ARM tree) we have define 5 ESRAM banks of 128 KB each and use of
static allocations becomes unmaintainable:
cd arch/arm/mach-ux500 && grep -r ESRAM .
./include/mach/db8500-regs.h:/* Base address and bank offsets for ESRAM */
./include/mach/db8500-regs.h:#define U8500_ESRAM_BASE   0x40000000
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK_SIZE      0x00020000
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK0  U8500_ESRAM_BASE
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK1       (U8500_ESRAM_BASE + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK2       (U8500_ESRAM_BANK1 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK3       (U8500_ESRAM_BANK2 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_BANK4       (U8500_ESRAM_BANK3 + U8500_ESRAM_BANK_SIZE)
./include/mach/db8500-regs.h:#define U8500_ESRAM_DMA_LCPA_OFFSET     0x10000
./include/mach/db8500-regs.h:#define U8500_DMA_LCPA_BASE
(U8500_ESRAM_BANK0 + U8500_ESRAM_DMA_LCPA_OFFSET)
./include/mach/db8500-regs.h:#define U8500_DMA_LCLA_BASE U8500_ESRAM_BANK4

I want to use genalloc to do dynamic allocations but I need to be able to
fine tune the allocation algorithm. I my case best-fit algorithm give
better results than first-fit, but it will not be true for every use case.

Signed-off-by: Benjamin Gaignard <[email protected]>
Cc: Huang Ying <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/gcd.c: prevent possible div by 0

Account for all properties when a and/or b are 0:
gcd(0, 0) = 0
gcd(a, 0) = a
gcd(0, b) = b

Fixes no known problems in current kernels.

Signed-off-by: Davidlohr Bueso <[email protected]>
Cc: Eric Dumazet <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/Kconfig.debug: adjust hard-lockup related Kconfig options

The main option should not appear in the resulting .config when the
dependencies aren't met (i.e.  use "depends on" rather than directly
setting the default from the combined dependency values).

The sub-options should depend on the main option rather than a more
generic higher level one.

Signed-off-by: Jan Beulich <[email protected]>
Acked-by: Don Zickus <[email protected]>
Cc: Peter Zijlstra <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/parser.c: avoid overflow in match_number()

The result of converting an integer value to another signed integer type
that's unable to represent the original value is implementation defined.
(See notes in section 6.3.1.3 of the C standard.)

In match_number(), the result of simple_strtol() (which returns type long)
is assigned to a value of type int.

Instead, handle the result of simple_strtol() in a well-defined way, and
return -ERANGE if the result won't fit in the int variable used to hold
the parsed result.

No current callers pay attention to the particular error value returned,
so this additional return code shouldn't do any harm.

[[email protected]: coding-style tweaks]
Signed-off-by: Alex Elder <[email protected]>
Cc: Randy Dunlap <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

idr: rename MAX_LEVEL to MAX_IDR_LEVEL

To avoid name conflicts:

  drivers/video/riva/fbdev.c:281:9: sparse: preprocessor token MAX_LEVEL redefined

While at it, also make the other names more consistent and add
parentheses.

[[email protected]: repair fallout]
[[email protected]: IB/mlx4: fix for MAX_ID_MASK to MAX_IDR_MASK name change]
Signed-off-by: Fengguang Wu <[email protected]>
Cc: Bernd Petrovitsch <[email protected]>
Cc: walter harms <[email protected]>
Cc: Glauber Costa <[email protected]>
Signed-off-by: Stephen Rothwell <[email protected]>
Cc: Roland Dreier <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

kvm: replace test_and_set_bit_le() in mark_page_dirty_in_slot() with set_bit_le()

Now that we have defined generic set_bit_le() we do not need to use
test_and_set_bit_le() for atomically setting a bit.

Signed-off-by: Takuya Yoshikawa <[email protected]>
Cc: Avi Kivity <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

powerpc: bitops: introduce {clear,set}_bit_le()

Needed to replace test_and_set_bit_le() in virt/kvm/kvm_main.c which is
being used for this missing function.

Signed-off-by: Takuya Yoshikawa <[email protected]>
Acked-by: Benjamin Herrenschmidt <[email protected]>
Cc: Avi Kivity <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

bitops: introduce generic {clear,set}_bit_le()

Needed to replace test_and_set_bit_le() in virt/kvm/kvm_main.c which is
being used for this missing function.

Signed-off-by: Takuya Yoshikawa <[email protected]>
Acked-by: Arnd Bergmann <[email protected]>
Cc: Marcelo Tosatti <[email protected]>
Cc: Avi Kivity <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/net/ethernet/dec/tulip: Use standard __set_bit_le() function

To introduce generic set_bit_le() later, we remove our own definition
and use a proper non-atomic bitops function: __set_bit_le().

Signed-off-by: Takuya Yoshikawa <[email protected]>
Acked-by: Grant Grundler <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/net/ethernet/sfc: use standard __{clear,set}_bit_le() functions

There are now standard functions for dealing with little-endian bit
arrays, so use them instead of our own implementations.

Signed-off-by: Ben Hutchings <[email protected]>
Signed-off-by: Takuya Yoshikawa <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/video/backlight/platform_lcd.c: add support for device tree based probe

This patch adds the of_match_table to platform-lcd driver to be
probed when platform-lcd device node is found in the device tree.

[[email protected]: include of.h]
Signed-off-by: Jingoo Han <[email protected]>
Cc: Richard Purdie <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/video/backlight/da9052_bl.c: drop devm_kfree of devm_kzalloc'd data

devm_kfree should not have to be explicitly used.

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

// <smpl>
@@
expression x,d;
@@

x = devm_kzalloc(...)
...
?-devm_kfree(d,x);
// </smpl>

Signed-off-by: Julia Lawall <[email protected]>
Acked-by: Jingoo Han <[email protected]>
Signed-off-by: Florian Tobias Schandinat <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

backlight: remove ProGear driver

This driver was for the ProGear webpad device which was produced in
2000/2001 and is not available on a market.  I no longer have this
hardware so can not even check how Linux works on it.

Signed-off-by: Marcin Juszkiewicz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

backlight: add new lm3639 backlight driver

This driver is a general version for LM3639 backlgiht + flash driver chip
of TI.

LM3639:
The LM3639 is a single chip LCD Display Backlight driver + white LED
Camera driver.  Programming is done over an I2C compatible interface.
www.ti.com

[[email protected]: code layout tweaks]
Signed-off-by: G.Shark Jeong <[email protected]>
Cc: Richard Purdie <[email protected]>
Cc: Daniel Jeong <[email protected]>
Cc: Randy Dunlap <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

backlight: add Backlight driver for lm3630 chip

This driver is a general version for LM3630 backlgiht driver chip of TI.

LM3630 :
The LM3630 is a current mode boost converter which supplies the power
and controls the current in two strings of up to 10 LEDs per string.
Programming is done over an I2C compatible interface.
www.ti.com

[[email protected]: make bled_name[] static, a few coding style tuneups, create new set_intensity(), partly to avoid awkward layout gymnastics]
Signed-off-by: G.Shark Jeong <[email protected]>
Cc: Richard Purdie <[email protected]>
Cc: Daniel Jeong <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

backlight: lp855x: add FAST bit description for LP8556

LP8556 backlight driver supports fast refresh mode when exiting the low
power mode.  This bit can be configurable in the platform side.

Signed-off-by: Milo(Woogyom) Kim <[email protected]>
Cc: Richard Purdie <[email protected]>
Cc: Bryan Wu <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/video/backlight/kb3886_bl.c: use usleep_range() instead of msleep() for small sleeps

Since msleep() might not sleep for the desired amount when less than 20ms,
use usleep_range().

Signed-off-by: Jingoo Han <[email protected]>
Cc: Claudio Nieder <[email protected]>
Cc: Richard Purdie <[email protected]>
Cc: Sachin Kamat <[email protected]>
Cc: Mark Brown <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/video/backlight/ltv350qv.c: use usleep_range() instead of msleep() for small sleeps

Since msleep() might not sleep for the desired amount when less than 20ms,
use usleep_range().

Signed-off-by: Jingoo Han <[email protected]>
Cc: Haavard Skinnemoen <[email protected]>
Cc: Richard Purdie <[email protected]>
Cc: Sachin Kamat <[email protected]>
Cc: Mark Brown <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/video/backlight/da9052_bl.c: use usleep_range() instead of msleep() for small sleeps

Since msleep() might not sleep for the desired amount when less than 20ms,
use usleep_range().

Signed-off-by: Jingoo Han <[email protected]>
Cc: Ashish Jangam <[email protected]>
Cc: Richard Purdie <[email protected]>
Cc: Sachin Kamat <[email protected]>
Cc: Mark Brown <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

MAINTAINERS: update email address for Khalid Aziz

Signed-off-by: Khalid Aziz <[email protected]>
Cc: Khalid Aziz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

MAINTAINERS: fix indentation for Viresh Kumar

Signed-off-by: Andy Shevchenko <[email protected]>
Acked-by: Viresh Kumar <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

MAINTAINERS: Update gianfar_ptp after renaming

Commit ec21e2ec3676 ("freescale: Move the Freescale drivers") moved the
files, update the pattern.

Signed-off-by: Joe Perches <[email protected]>
Acked-by: Richard Cochran <[email protected]>
Acked-by: Jeff Kirsher <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

MAINTAINERS: add defconfig file to IMX section

Signed-off-by: Uwe Kleine-König <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

MAINTAINERS: update gpio subsystem file list

Signed-off-by: Yang Bai <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib/vsprintf: update documentation to cover all of %p[Mm][FR]

Acked-by: Andrei Emeltchenko <[email protected]>
Signed-off-by: Andy Shevchenko <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib: vsprintf: fix broken comments

Numbering the 8 potential digits 2 though 9 never did make a lot of sense.

Signed-off-by: George Spelvin <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: Michal Nazarewicz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib: vsprintf: optimize put_dec_trunc8()

If you're going to have a conditional branch after each 32x32->64-bit
multiply, might as well shrink the code and make it a loop.

This also avoids using the long multiply for small integers.

(This leaves the comments in a confusing state, but that's a separate
patch to make review easier.)

Signed-off-by: George Spelvin <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: Michal Nazarewicz <[email protected]>
Cc: Rabin Vincent <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib: vsprintf: optimize division by 10000

The same multiply-by-inverse technique can be used to convert division by
10000 to a 32x32->64-bit multiply.

Signed-off-by: George Spelvin <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: Michal Nazarewicz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

lib: vsprintf: optimize division by 10 for small integers

Shrink the reciprocal approximations used in put_dec_full4() based on the
comments in put_dec_full9().

Signed-off-by: George Spelvin <[email protected]>
Cc: Denys Vlasenko <[email protected]>
Cc: Michal Nazarewicz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

poweroff: fix bug in orderly_poweroff()

orderly_poweroff is trying to poweroff platform in two steps:

step 1: Call user space application to poweroff
step 2: If user space poweroff fail, then do a force power off if force param
        is set.

The bug here is, step 1 is always successful with param UMH_NO_WAIT, which obey
the design goal of orderly_poweroff.

We have two choices here:
UMH_WAIT_EXEC which means wait for the exec, but not the process;
UMH_WAIT_PROC which means wait for the process to complete.
we need to trade off the two choices:

If using UMH_WAIT_EXEC, there is potential issue comments by Serge E.
Hallyn: The exec will have started, but may for whatever (very unlikely)
reason fail.

If using UMH_WAIT_PROC, there is potential issue comments by Eric W.
Biederman: If the caller is not running in a kernel thread then we can
easily get into a case where the user space caller will block waiting for
us when we are waiting for the user space caller.

Thanks for their excellent ideas, based on the above discussion, we
finally choose UMH_WAIT_EXEC, which is much more safe, if the user
application really fails, we just complain the application itself, it
seems a better choice here.

Signed-off-by: Feng Hong <[email protected]>
Acked-by: Kees Cook <[email protected]>
Acked-by: Serge Hallyn <[email protected]>
Cc: "Eric W. Biederman" <[email protected]>
Acked-by: "Rafael J. Wysocki" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

kernel/sys.c: call disable_nonboot_cpus() in kernel_restart()

As kernel_power_off() calls disable_nonboot_cpus(), we may also want to
have kernel_restart() call disable_nonboot_cpus().  Doing so can help
machines that require boot cpu be the last alive cpu during reboot to
survive with kernel restart.

This fixes one reboot issue seen on imx6q (Cortex-A9 Quad).  The machine
requires that the restart routine be run on the primary cpu rather than
secondary ones.  Otherwise, the secondary core running the restart
routine will fail to come to online after reboot.

Signed-off-by: Shawn Guo <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

tile: fix personality bits handling upon exec()

Historically, the top three bytes of personality have been used for
things such as ADDR_NO_RANDOMIZE, which made sense only for specific
architectures.

We now however have a flag there that is general no matter the
architecture (UNAME26); generally we have to be careful to preserve the
personality flags across exec().

This patch fixes tile architecture not to forcefully overwrite
personality flags during exec().

In addition to that, we fix two other things along the way:

- exec_domain switching is fixed -- set_personality() should always
  be used instead of directly assigning to current->personality.
- as pointed out by Arnd Bergmann, PER_LINUX_32BIT is not used anywhere
  by tile, so let's just drop that in favor of PER_LINUX

Signed-off-by: Jiri Kosina <[email protected]>
Acked-by: Chris Metcalf <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

cross-arch: don't corrupt personality flags upon exec()

Historically, the top three bytes of personality have been used for
things such as ADDR_NO_RANDOMIZE, which made sense only for specific
architectures.

We now however have a flag there that is general no matter the
architecture (UNAME26); generally we have to be careful to preserve the
personality flags across exec().

This patch tries to fix all architectures that forcefully overwrite
personality flags during exec() (ppc32 and s390 have been fixed recently
by commits f9783ec862ea ("[S390] Do not clobber personality flags on
exec") and 59e4c3a2fe9c ("powerpc/32: Don't clobber personality flags on
exec") in a similar way already).

Signed-off-by: Jiri Kosina <[email protected]>
Cc: Haavard Skinnemoen <[email protected]>
Cc: Hans-Christian Egtvedt <[email protected]>
Cc: Mike Frysinger <[email protected]>
Cc: Mark Salter <[email protected]>
Cc: Mikael Starvik <[email protected]>
Cc: Jesper Nilsson <[email protected]>
Cc: David Howells <[email protected]>
Cc: Yoshinori Sato <[email protected]>
Cc: Richard Kuo <[email protected]>
Cc: Hirokazu Takata <[email protected]>
Cc: Geert Uytterhoeven <[email protected]>
Cc: Michal Simek <[email protected]>
Cc: Koichi Yasutake <[email protected]>
Cc: Jonas Bonn <[email protected]>
Cc: Chen Liqin <[email protected]>
Cc: Lennox Wu <[email protected]>
Cc: Paul Mundt <[email protected]>
Cc: "David S. Miller" <[email protected]>
Cc: Chris Zankel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

score: select generic atomic64_t support

It's required for the core fs/namespace.c and many other basic features.

Signed-off-by: Fengguang Wu <[email protected]>
Acked-by: Lennox Wu <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

frv: kill used but uninitialized variable

Commit 6afe1a1fe8ff ("PM: Remove legacy PM") removed the initialization
of retval, causing:

  arch/frv/kernel/pm.c: In function 'sysctl_pm_do_suspend':
  arch/frv/kernel/pm.c:165:5: warning: 'retval' may be used uninitialized in this function [-Wuninitialized]

Remove the variable completely to fix this, and convert to a proper
switch (...) { ... } construct to improve readability.

Signed-off-by: Geert Uytterhoeven <[email protected]>
Cc: David Howells <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix const sections for crc32 table

Fix the const sections for the code generated by crc32 table.  There's
no ro version of the cacheline aligned section, so we cannot put in
const data without a conflict Just don't make the crc tables const for
now.

[[email protected]: some fixes and new description]
[[email protected]: checkpatch fixes]
Signed-off-by: Joe Mario <[email protected]>
Signed-off-by: Andi Kleen <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in sound

Signed-off-by: Andi Kleen <[email protected]>
Cc: Takashi Iwai <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in net

Signed-off-by: Andi Kleen <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in net/can

Signed-off-by: Andi Kleen <[email protected]>
Cc: Oliver Hartkopp <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in mm/percpu.c

Signed-off-by: Andi Kleen <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/video

Signed-off-by: Andi Kleen <[email protected]>
Cc: Florian Tobias Schandinat <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/scsi

Signed-off-by: Andi Kleen <[email protected]>
Cc: James Bottomley <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/platform/x86

Signed-off-by: Andi Kleen <[email protected]>
Cc: Matthew Garrett <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/net/wan

Signed-off-by: Andi Kleen <[email protected]>
Cc: Krzysztof Halasa <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/net/hamradio

Signed-off-by: Andi Kleen <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/net

Signed-off-by: Andi Kleen <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/mmc

Signed-off-by: Andi Kleen <[email protected]>
Cc: Chris Ball <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/mfd

Signed-off-by: Andi Kleen <[email protected]>
Cc: Samuel Ortiz <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/macintosh

[[email protected]: checkpatch fixes]
Signed-off-by: Andi Kleen <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/ide

Signed-off-by: Andi Kleen <[email protected]>
Cc: David Miller <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/char

Signed-off-by: Andi Kleen <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in drivers/atm

Signed-off-by: Andi Kleen <[email protected]>
Cc: Chas Williams <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/x86

Signed-off-by: Andi Kleen <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Thomas Gleixner <[email protected]>
Cc: "H. Peter Anvin" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/sh

Signed-off-by: Andi Kleen <[email protected]>
Cc: Paul Mundt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/score

Signed-off-by: Andi Kleen <[email protected]>
Cc: Chen Liqin <[email protected]>
Cc: Lennox Wu <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/powerpc

Signed-off-by: Andi Kleen <[email protected]>
Cc: Benjamin Herrenschmidt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/mips

Signed-off-by: Andi Kleen <[email protected]>
Cc: Ralf Baechle <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/ia64

Signed-off-by: Andi Kleen <[email protected]>
Cc: "Luck, Tony" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/h8300

[[email protected]: checkpatch fixes]
Signed-off-by: Andi Kleen <[email protected]>
Cc: Yoshinori Sato <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/frv

Signed-off-by: Andi Kleen <[email protected]>
Cc: David Howells <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: fix section conflicts in arch/arm/

Signed-off-by: Andi Kleen <[email protected]>
Cc: Russell King <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sections: disable const sections for PA-RISC v2

The PA-RISC tool chain seems to have some problem with correct
read/write attributes on sections.  This causes problems when the const
sections are fixed up for other architecture to only contain truly
read-only data.

Disable const sections for PA-RISC

This can cause a bit of noise with modpost.

Signed-off-by: Andi Kleen <[email protected]>
Cc: "James E.J. Bottomley" <[email protected]>
Cc: Helge Deller <[email protected]>
Acked-by: Sam Ravnborg <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/scsi/atp870u.c: fix bad use of udelay

The ACARD driver calls udelay() with a value > 2000, which leads to to
the following compilation error on ARM:

  ERROR: "__bad_udelay" [drivers/scsi/atp870u.ko] undefined!
  make[1]: *** [__modpost] Error 1

This is because udelay is defined on ARM, roughly speaking, as

#define udelay(n) ((n) > 2000 ? __bad_udelay() : \
__const_udelay((n) * ((2199023U*HZ)>>11)))

The argument to __const_udelay is the number of jiffies to wait divided
by 4, but this does not work unless the multiplication does not
overflow, and that is what the build error is designed to prevent.  The
intended behavior can be achieved by using mdelay to call udelay
multiple times in a loop.

[[email protected]: adding context]
Signed-off-by: Martin Michlmayr <[email protected]>
Signed-off-by: Jonathan Nieder <[email protected]>
Cc: James Bottomley <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

unicore32: select generic atomic64_t support

It's required for the core fs/namespace.c and many other basic features.

Signed-off-by: Guan Xuetao <[email protected]>
Signed-off-by: Fengguang Wu <[email protected]>
Cc: "Eric W. Biederman" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

kbuild: make: fix if_changed when command contains backslashes

The call if_changed mechanism does not work when the command contains
backslashes.  This basically is an issue with lzo and bzip2 compressed
kernels.  The compressed binaries do not contain the uncompressed image
size, so these use size_append to append the size.  This results in
backslashes in the executed command.  With this if_changed always
detects a change in the command and rebuilds the compressed image even
if nothing has changed.

Fix this by escaping backslashes in make-cmd

Signed-off-by: Sascha Hauer <[email protected]>
Signed-off-by: Jan Luebbe <[email protected]>
Cc: Sam Ravnborg <[email protected]>
Cc: Bernhard Walle <[email protected]>
Cc: Michal Marek <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/dma/dmaengine.c: lower the priority of 'failed to get' dma channel message

Do the same as commit a03a202e95fd ("dmaengine: failure to get a
specific DMA channel is not critical") to get rid of the following
messages during kernel boot:

  dmaengine_get: failed to get dma1chan0: (-22)
  dmaengine_get: failed to get dma1chan1: (-22)
  dmaengine_get: failed to get dma1chan2: (-22)
  dmaengine_get: failed to get dma1chan3: (-22)
  ..

Signed-off-by: Fabio Estevam <[email protected]>
Cc: Vinod Koul <[email protected]>
Cc: Dan Williams <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mn10300: only add -mmem-funcs to KBUILD_CFLAGS if gcc supports it

It seems the current (gcc 4.6.3) no longer provides this so make it
conditional.

As reported by Tony before, the mn10300 architecture cross-compiles with
gcc-4.6.3 if -mmem-funcs is not added to KBUILD_CFLAGS.

Reported-by: Tony Breeds <[email protected]>
Signed-off-by: Geert Uytterhoeven <[email protected]>
Cc: David Howells <[email protected]>
Cc: Koichi Yasutake <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

audit.h: replace defines with C stubs

Replace the #defines used when CONFIG_AUDIT or CONFIG_AUDIT_SYSCALLS are
disabled so we get type checking during those builds.

Suggested-by: Andrew Morton <[email protected]>
Signed-off-by: Kees Cook <[email protected]>
Cc: Al Viro <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Merge tag 'kvm-3.7-1' of git://git./virt/kvm/kvm

Pull KVM updates from Avi Kivity:
 "Highlights of the changes for this release include support for vfio
  level triggered interrupts, improved big real mode support on older
  Intels, a streamlines guest page table walker, guest APIC speedups,
  PIO optimizations, better overcommit handling, and read-only memory."

* tag 'kvm-3.7-1' of git://git.kernel.org/pub/scm/virt/kvm/kvm: (138 commits)
  KVM: s390: Fix vcpu_load handling in interrupt code
  KVM: x86: Fix guest debug across vcpu INIT reset
  KVM: Add resampling irqfds for level triggered interrupts
  KVM: optimize apic interrupt delivery
  KVM: MMU: Eliminate pointless temporary 'ac'
  KVM: MMU: Avoid access/dirty update loop if all is well
  KVM: MMU: Eliminate eperm temporary
  KVM: MMU: Optimize is_last_gpte()
  KVM: MMU: Simplify walk_addr_generic() loop
  KVM: MMU: Optimize pte permission checks
  KVM: MMU: Update accessed and dirty bits after guest pagetable walk
  KVM: MMU: Move gpte_access() out of paging_tmpl.h
  KVM: MMU: Optimize gpte_access() slightly
  KVM: MMU: Push clean gpte write protection out of gpte_access()
  KVM: clarify kvmclock documentation
  KVM: make processes waiting on vcpu mutex killable
  KVM: SVM: Make use of asm.h
  KVM: VMX: Make use of asm.h
  KVM: VMX: Make lto-friendly
  KVM: x86: lapic: Clean up find_highest_vector() and count_vectors()
  ...

Conflicts:
arch/s390/include/asm/processor.h
arch/x86/kvm/i8259.c

firmware: use 'kernel_read()' to read firmware into kernel buffer

Fengguang correctly points out that the firmware reading should not use
vfs_read(), since the buffer is in kernel space.

The vfs_read() just happened to work for kernel threads, but sparse
warns about the incorrect address spaces, and it's definitely incorrect
and could fail for other users of the firmware loading.

Reported-by: Fengguang Wu <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs

Pull ext3 & udf fixes from Jan Kara:
 "Shortlog pretty much says it all.

  The interesting bits are UDF support for direct IO and ext3 fix for a
  long standing oops in data=journal mode."

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  jbd: Fix assertion failure in commit code due to lacking transaction credits
  UDF: Add support for O_DIRECT
  ext3: Replace 0 with NULL for pointer in super.c file
  udf: add writepages support for udf
  ext3: don't clear orphan list on ro mount with errors
  reiserfs: Make reiserfs_xattr_handlers static

Merge tag 'remoteproc-for-3.7' of git://git./linux/kernel/git/ohad/remoteproc

Pull remoteproc update from Ohad Ben-Cohen:

 - Remoteproc Recovery - by Fernando Guzman Lugo

   When a remote processor crash is detected, this mechanism will remove
   all virtio children devices, wait until their drivers let go, hard
   reset the remote processor and reload the firmware (resulting in the
   relevant virtio children devices re-added).  Essentially the entire
   software stack is reset, together with the relevant hardware, so
   users don't have to reset the entire phone.

 - STE Modem driver is added - by Sjur Brændeland

 - OMAP DSP boot address support is added - by Juan Gutierrez

 - A handful of fixes/cleanups - Sjur Brændeland, Dan Carpenter, Emil
   Goode

* tag 'remoteproc-for-3.7' of git://git.kernel.org/pub/scm/linux/kernel/git/ohad/remoteproc:
  remoteproc: Fix use of format specifyer
  remoteproc: fix a potential NULL-dereference on cleanup
  remoteproc: select VIRTIO to avoid build breakage
  remoteproc: return -EFAULT on copy_from_user failure
  remoteproc: snprintf() can return more than was printed
  remoteproc: Add STE modem driver
  remtoteproc: maintain max notifyid
  remoteproc: create a 'recovery' debugfs entry
  remoteproc: add actual recovery implementation
  remoteproc: add rproc_report_crash function to notify rproc crashes
  remoteproc: Add dependency to HAS_DMA
  remoteproc/omap: set bootaddr support