13 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Sat, 14 Nov 2009 21:00:17 +0000 (13:00 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

* 'for-linus' of git://
  Input: psmouse - remove unneeded '\n' from psmouse.proto parameter
  Input: atkbd - restore LED state at reconnect
  Input: force LED reset on resume
  Input: fix locking in memoryless force-feedback devices

13 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Sat, 14 Nov 2009 20:59:32 +0000 (12:59 -0800)]
Merge branch 'for-linus' of git://

* 'for-linus' of git://
  md/raid5: Allow dirty-degraded arrays to be assembled when only party is degraded.
  Don't unconditionally set in_sync on newly added device in raid5_reshape
  md: allow v0.91 metadata to record devices as being active but not in-sync.
  md: factor out updating of 'recovery_offset'.

13 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Sat, 14 Nov 2009 20:59:06 +0000 (12:59 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tj/percpu

* 'for-linus' of git://
  percpu: restructure pcpu_extend_area_map() to fix bugs and improve readability

13 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Sat, 14 Nov 2009 20:57:39 +0000 (12:57 -0800)]
Merge branch 'for-linus' of git://

* 'for-linus' of git://
  [S390] s390: fix single stepping on svc0
  [S390] sclp: undo quiesce handler override on resume
  [S390] reset cputime accounting after IPL from NSS
  [S390] monreader: fix use after free bug with suspend/resume

13 years agofix memory leak in fixed btusb_close
Oliver Neukum [Fri, 13 Nov 2009 13:26:23 +0000 (14:26 +0100)]
fix memory leak in fixed btusb_close

If the waker is killed before it can replay outstanding URBs, these URBs
won't be freed or will be replayed at the next open.  This patch closes
the window by explicitely discarding outstanding URBs.

Signed-off-by: Oliver Neukum <>
Acked-by: Marcel Holtmann <>
Signed-off-by: Linus Torvalds <>
13 years agoFix memory corruption caused by nfsd readdir+
Petr Vandrovec [Sat, 14 Nov 2009 09:47:07 +0000 (10:47 +0100)]
Fix memory corruption caused by nfsd readdir+

Commit 8177e6d6dfb9cd03d9bdeb647c32161f8f58f686 ("nfsd: clean up
readdirplus encoding") introduced single character typo in nfs3 readdir+
implementation.  Unfortunately that typo has quite bad side effects:
random memory corruption, followed (on my box) with immediate
spontaneous box reboot.

Using 'p1' instead of 'p' fixes my Linux box rebooting whenever VMware
ESXi box tries to list contents of my home directory.

Signed-off-by: Petr Vandrovec <>
Cc: "J. Bruce Fields" <>
Cc: Neil Brown <>
Signed-off-by: Linus Torvalds <>
13 years ago[S390] s390: fix single stepping on svc0
Christian Borntraeger [Fri, 13 Nov 2009 14:43:54 +0000 (15:43 +0100)]
[S390] s390: fix single stepping on svc0

On s390 there are two ways of specifying the system call number for
the svc instruction. The standard way is to use the immediate field
in the instruction (or to use EXecute for values unknown during
assemble time). This can encode 256 system calls.
The kernel ABI also allows to put the system call number in r1 and
then execute svc 0 to enable system call numbers > 255.

It turns out that single stepping svc 0 is broken, since the PER
program check handler uses r1. We have to use a different register.

Signed-off-by: Christian Borntraeger <>
Signed-off-by: Martin Schwidefsky <>
13 years ago[S390] sclp: undo quiesce handler override on resume
Martin Schwidefsky [Fri, 13 Nov 2009 14:43:53 +0000 (15:43 +0100)]
[S390] sclp: undo quiesce handler override on resume

In a system where the ctrl-alt-del init action initiated by signal
quiesce suspends the machine the quiesce handler override for
_machine_restart, _machine_halt and _machine_power_off needs to be
undone, otherwise the override is still present in the resumed
system. The next shutdown would then load the quiesce state psw
instead of performing the correct shutdown action.

Signed-off-by: Martin Schwidefsky <>
13 years ago[S390] reset cputime accounting after IPL from NSS
Martin Schwidefsky [Fri, 13 Nov 2009 14:43:52 +0000 (15:43 +0100)]
[S390] reset cputime accounting after IPL from NSS

After an IPL from NSS the uptime of the system is incorrect. The reason
is that the startup code in head.S is not executed in case of an IPL
from NSS. Due to that sched_clock_base_cc which is used to initialze
wall_to_monotonic contains the time stamp when the NSS has been created
instead of the time stamp of the system start.

Reinitialize the cputime accounting values in create_kernel_nss after
the SAVESYS CP command that created the NSS segment.

Signed-off-by: Martin Schwidefsky <>
13 years ago[S390] monreader: fix use after free bug with suspend/resume
Gerald Schaefer [Fri, 13 Nov 2009 14:43:51 +0000 (15:43 +0100)]
[S390] monreader: fix use after free bug with suspend/resume

The monreader device driver doesn't set dev->driver_data to NULL after
freeing the corresponding data structure. This leads to a use after
free bug in the freeze/thaw suspend/resume functions after the device
has been opened and closed once. Fix this by clearing dev->driver_data
in the close() function.

Signed-off-by: Gerald Schaefer <>
Signed-off-by: Martin Schwidefsky <>
13 years agoInput: psmouse - remove unneeded '\n' from psmouse.proto parameter
Takashi Iwai [Fri, 13 Nov 2009 07:30:52 +0000 (23:30 -0800)]
Input: psmouse - remove unneeded '\n' from psmouse.proto parameter

The get parameter function should return a string without a life-feed.
Otherwise you'll see additional empty line in sysfs parameters file.

Signed-off-by: Takashi Iwai <>
Signed-off-by: Dmitry Torokhov <>
13 years agoInput: atkbd - restore LED state at reconnect
Dmitry Torokhov [Fri, 13 Nov 2009 07:19:05 +0000 (23:19 -0800)]
Input: atkbd - restore LED state at reconnect

Even though input core tells us to restore LED state and repeat rate
at resume keyboard may be reconnected either by request from userspace
(via sysfs) or just by pulling it from the box and plugging it back in.
In these cases we still need to restore state ourselves.

Signed-off-by: Dmitry Torokhov <>
13 years agoInput: force LED reset on resume
Dmitry Torokhov [Fri, 13 Nov 2009 07:19:05 +0000 (23:19 -0800)]
Input: force LED reset on resume

We should be sending EV_LED event down to drivers upon resume even in cases
when in-kernel state of the LED is off since device could come up with some
leds turned on.

Reported-and-tested-by: Mikko Vinni <>
Signed-off-by: Dmitry Torokhov <>
13 years agomd/raid5: Allow dirty-degraded arrays to be assembled when only party is degraded.
NeilBrown [Fri, 13 Nov 2009 06:47:00 +0000 (17:47 +1100)]
md/raid5: Allow dirty-degraded arrays to be assembled when only party is degraded.

Normally is it not safe to allow a raid5 that is both dirty and
degraded to be assembled without explicit request from that admin, as
it can cause hidden data corruption.
This is because 'dirty' means that the parity cannot be trusted, and
'degraded' means that the parity needs to be used.

However, if the device that is missing contains only parity, then
there is no issue and assembly can continue.
This particularly applies when a RAID5 is being converted to a RAID6
and there is an unclean shutdown while the conversion is happening.

So check for whether the degraded space only contains parity, and
in that case, allow the assembly.

Signed-off-by: NeilBrown <>
13 years agoDon't unconditionally set in_sync on newly added device in raid5_reshape
NeilBrown [Fri, 13 Nov 2009 06:40:51 +0000 (17:40 +1100)]
Don't unconditionally set in_sync on newly added device in raid5_reshape

When a reshape finds that it can add spare devices into the array,
those devices might already be 'in_sync' if they are beyond the old
size of the array, or they might not if they are within the array.

The first case happens when we change an N-drive RAID5 to an
N+1-drive RAID5.
The second happens when we convert an N-drive RAID5 to an
N+1-drive RAID6.

So set the flag more carefully.
Also, ->recovery_offset is only meaningful when the flag is clear,
so only set it in that case.

This change needs the preceding two to ensure that the non-in_sync
device doesn't get evicted from the array when it is stopped, in the
case where v0.90 metadata is used.

Signed-off-by: NeilBrown <>
13 years agomd: allow v0.91 metadata to record devices as being active but not in-sync.
NeilBrown [Fri, 13 Nov 2009 06:40:48 +0000 (17:40 +1100)]
md: allow v0.91 metadata to record devices as being active but not in-sync.

This is a combination that didn't really make sense before.
However when a reshape is converting e.g. raid5 -> raid6, the extra
device is not fully in-sync, but is certainly active and contains
important data.
So allow that start to be meaningful and in particular get
the 'recovery_offset' value (which is needed for any non-in-sync
active device) from the reshape_position.

Signed-off-by: NeilBrown <>
13 years agoLinux 2.6.32-rc7 v2.6.32-rc7
Linus Torvalds [Fri, 13 Nov 2009 00:46:07 +0000 (16:46 -0800)]
Linux 2.6.32-rc7

13 years agoMerge branch 'omap-fixes-for-linus' of git://
Linus Torvalds [Thu, 12 Nov 2009 22:57:15 +0000 (14:57 -0800)]
Merge branch 'omap-fixes-for-linus' of git://git./linux/kernel/git/tmlind/linux-omap-2.6

* 'omap-fixes-for-linus' of git://
  omap3: Decrease cpufreq transition latency
  omap3: update Pandora defconfig
  omap3: 3430sdp: Enable Linux Regulator framework
  omap3: beagle: Fix USB host port power control
  omap3: pandora: Fix keypad keymap
  omap1: Amstrad Delta defconfig fixes
  omap: Fix omapfb/lcdc on OMAP1510 broken when PM set
  omap: Use resource_size
  omap: Fix race condition in omap dma driver

13 years agopercpu: restructure pcpu_extend_area_map() to fix bugs and improve readability
Tejun Heo [Wed, 11 Nov 2009 06:35:18 +0000 (15:35 +0900)]
percpu: restructure pcpu_extend_area_map() to fix bugs and improve readability

pcpu_extend_area_map() had the following two bugs.

* It should return 1 if pcpu_lock was dropped and reacquired but it
  returned 0.  This could lead to oops if free_percpu() races with
  area map extension.

* pcpu_mem_free() was called under pcpu_lock.  pcpu_mem_free() might
  end up calling vfree() which isn't IRQ safe.  This could lead to
  deadlock through lock order inversion via IRQ.

In addition, Linus pointed out that the temporary lock dropping and
subtle three-way return value of pcpu_extend_area_map() was very ugly
and suggested to split the function into two - pcpu_need_to_extend()
and pcpu_extend_area_map().

This patch restructures pcpu_extend_area_map() as suggested and fixes
the two bugs.

Signed-off-by: Tejun Heo <>
Acked-by: Linus Torvalds <>
Cc: Ingo Molnar <>
13 years ago__generic_block_fiemap(): fix for files bigger than 4GB
Mike Hommey [Wed, 11 Nov 2009 22:26:55 +0000 (14:26 -0800)]
__generic_block_fiemap(): fix for files bigger than 4GB

Because of an integer overflow on start_blk, various kind of wrong results
would be returned by the generic_block_fiemap() handler, such as no
extents when there is a 4GB+ hole at the beginning of the file, or wrong
fe_logical when an extent starts after the first 4GB.

Signed-off-by: Mike Hommey <>
Cc: Alexander Viro <>
Cc: Steven Whitehouse <>
Cc: Theodore Ts'o <>
Cc: Eric Sandeen <>
Cc: Josef Bacik <>
Cc: Mark Fasheh <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agopps: events reporting fix up
Rodolfo Giometti [Wed, 11 Nov 2009 22:26:54 +0000 (14:26 -0800)]
pps: events reporting fix up

PPS events must be recorded according to PPS's mode settings.

If a process asks for (i.e.) capture-assert events only, when the PPS
client calls the pps_event() function to save the current PPS event, we
should verify the event type and then discard unwanted ones.

Also, without this patch userland processes waiting for a specific PPS
event (assert or clear but not both) may be awakened at wrong time.

Signed-off-by: Rodolfo Giometti <>
Tested-by: William S. Brasher <>
Tested-by: Reg Clemens <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agopps: locking scheme fix up for PPS_GETPARAMS
Rodolfo Giometti [Wed, 11 Nov 2009 22:26:52 +0000 (14:26 -0800)]
pps: locking scheme fix up for PPS_GETPARAMS

Userland programs may read/write PPS parameters at same time and these
operations may corrupt PPS data.

Signed-off-by: Rodolfo Giometti <>
Tested-by: Reg Clemens <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agodrivers/video/msm: update to new kernel
Pavel Machek [Wed, 11 Nov 2009 22:26:51 +0000 (14:26 -0800)]
drivers/video/msm: update to new kernel

TASK_INTERRUPTIBLE and friends are now only available after including
<linux/sched.h>, so include it when needed.

bus_id is no longer available/necessary, so remove that.

Android pmem driver is not available in mainline, so remove its hooks
from drivers/video.

Signed-off-by: Pavel Machek <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agogpiolib: fix device_create() result check
Sergei Shtylyov [Wed, 11 Nov 2009 22:26:50 +0000 (14:26 -0800)]
gpiolib: fix device_create() result check

In case of failure, device_create() returns not NULL but the error code.
The current code checks for non-NULL though which causes kernel oops in
sysfs_create_group() when device_create() fails.  Check for error using
IS_ERR() and propagate the error value using PTR_ERR() instead of fixed
-ENODEV code returned now...

Signed-off-by: Sergei Shtylyov <>
Cc: David Brownell <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agortc: v3020: fix v3020_mmio_read_bit()
Scott Valentine [Wed, 11 Nov 2009 22:26:49 +0000 (14:26 -0800)]
rtc: v3020: fix v3020_mmio_read_bit()

v3020_mmio_read_bit() always returns 0 when left_shift > 7.

v3020_mmio_read_bit()'s return type is (unsigned char).  The code returns
a value masked by (1 << left_shift) that is casted to the return type.  If
left_shift is larger than 7, the cast will always result in a 0 return
value.  The problem was discovered with left_shift = 16, and the included
patch corrects the problem.

The bug was introduced in the last (Apr 3 2009) commit of the file, kernel
versions 2.6.30 and later.

Cc: Alessandro Zummo <>
Cc: Paul Gortmaker <>
Cc: Raphael Assenat <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agortc-vr41xx: fix do_div() warning
Yoichi Yuasa [Wed, 11 Nov 2009 22:26:48 +0000 (14:26 -0800)]
rtc-vr41xx: fix do_div() warning

drivers/rtc/rtc-vr41xx.c: In function 'vr41xx_rtc_irq_set_freq':
drivers/rtc/rtc-vr41xx.c:217: warning: comparison of distinct pointer types lacks a cast
drivers/rtc/rtc-vr41xx.c:217: warning: right shift count >= width of type
drivers/rtc/rtc-vr41xx.c:217: warning: passing argument 1 of '__div64_32' from incompatible pointer type
include/asm-generic/div64.h:35: note: expected 'uint64_t *' but argument is of type 'long unsigned int *'

Signed-off-by: Yoichi Yuasa <>
Cc: Alessandro Zummo <>
Cc: Paul Gortmaker <>
Cc: Paul Mundt <>
Cc: David Brownell <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoexec: setup_arg_pages() fails to return errors
Anton Blanchard [Wed, 11 Nov 2009 22:26:48 +0000 (14:26 -0800)]
exec: setup_arg_pages() fails to return errors

In setup_arg_pages we work hard to assign a value to ret, but on exit we
always return 0.

Also remove a now duplicated exit path and branch to out_unlock instead.

Signed-off-by: Anton Blanchard <>
Acked-by: Serge Hallyn <>
Reviewed-by: WANG Cong <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agofb: put framebuffer drivers back under the "frame buffer devices" menu
Guennadi Liakhovetski [Wed, 11 Nov 2009 22:26:47 +0000 (14:26 -0800)]
fb: put framebuffer drivers back under the "frame buffer devices" menu

The config FB_PRE_INIT_FB entry in drivers/video/Kconfig pushes all entries
below it out of the menuconfig selection. Fix this.

Signed-off-by: Guennadi Liakhovetski <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoMAINTAINERS: correct SECURITY SUBSYSTEM git entry
Joe Perches [Wed, 11 Nov 2009 22:26:46 +0000 (14:26 -0800)]

Use not

Signed-off-by: Joe Perches <>
Cc: James Morris <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoMAINTAINERS: correct NETFILTER git entry format
Joe Perches [Wed, 11 Nov 2009 22:26:45 +0000 (14:26 -0800)]
MAINTAINERS: correct NETFILTER git entry format

Signed-off-by: Joe Perches <>
Cc: Patrick McHardy <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoMAINTAINERS: correct 9P FILE SYSTEM git entry
Joe Perches [Wed, 11 Nov 2009 22:26:44 +0000 (14:26 -0800)]
MAINTAINERS: correct 9P FILE SYSTEM git entry

Signed-off-by: Joe Perches <>
Cc: Latchesar Ionkov <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agommci-omap: free irq resource
Ladislav Michl [Wed, 11 Nov 2009 22:26:43 +0000 (14:26 -0800)]
mmci-omap: free irq resource

Free IRQ on remove.

Signed-off-by: Ladislav Michl <>
Acked-by: Tony Lindgren <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoserial: add support for the Lava Quattro PCI quad-port 16550A card
Lennert Buytenhek [Wed, 11 Nov 2009 22:26:42 +0000 (14:26 -0800)]
serial: add support for the Lava Quattro PCI quad-port 16550A card

This seems to be a different model (with a different PCI ID) than the
"Quatro" card that is also in the list.

Signed-off-by: Lennert Buytenhek <>
Cc: Alan Cox <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoserial: fix printk format specifiers for struct uart_port::iobase
Lennert Buytenhek [Wed, 11 Nov 2009 22:26:41 +0000 (14:26 -0800)]
serial: fix printk format specifiers for struct uart_port::iobase

struct uart_port::iobase is unsigned long, so use %lx as printk format

Signed-off-by: Lennert Buytenhek <>
Cc: Alan Cox <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agortc: pcf50633: consider alrm->enable in pcf50633_rtc_set_alarm
Werner Almesberger [Wed, 11 Nov 2009 22:26:40 +0000 (14:26 -0800)]
rtc: pcf50633: consider alrm->enable in pcf50633_rtc_set_alarm

According to Documentation/rtc.txt, RTC_WKALM_SET sets the alarm time and
enables/disables the alarm.  We implement RTC_WKALM_SET through
pcf50633_rtc_set_alarm.  The enabling/disabling part was missing.

Signed-off-by: Werner Almesberger <>
Reported-by: Michael 'Mickey' Lauer <>
Signed-off-by: Paul Fertser <>
Cc: Paul Gortmaker <>
Cc: Balaji Rao <>
Cc: Alessandro Zummo <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agortc: pcf50633: fix month off-by-one error
Rask Ingemann Lambertsen [Wed, 11 Nov 2009 22:26:36 +0000 (14:26 -0800)]
rtc: pcf50633: fix month off-by-one error

The PCF50633 stores a month value of 1-12, but the kernel wants 0-11.

Signed-off-by: Rask Ingemann Lambertsen <>
Signed-off-by: Paul Fertser <>
Cc: Paul Gortmaker <>
Cc: Balaji Rao <>
Cc: Alessandro Zummo <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoatmel_lcdfb: new alternate pixel clock formula
Nicolas Ferre [Wed, 11 Nov 2009 22:26:35 +0000 (14:26 -0800)]
atmel_lcdfb: new alternate pixel clock formula

at91sam9g45 non ES lots have an alternate pixel clock calculation formula.
Introduce this one with condition on the cpu_is_xxxxx() macros.

Newer 9g45 SOC will not have good pixel clock calculation without this

Signed-off-by: Nicolas Ferre <>
Cc: Krzysztof Helt <>
Cc: Haavard Skinnemoen <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agofs: add missing compat_ptr handling for FS_IOC_RESVSP ioctl
Heiko Carstens [Wed, 11 Nov 2009 22:26:34 +0000 (14:26 -0800)]
fs: add missing compat_ptr handling for FS_IOC_RESVSP ioctl

For FS_IOC_RESVSP and FS_IOC_RESVSP64 compat_sys_ioctl() uses its
arg argument as a pointer to userspace. However it is missing a
a call to compat_ptr() which will do a proper pointer conversion.

This was introduced with 3e63cbb1 "fs: Add new pre-allocation ioctls
to vfs for compatibility with legacy xfs ioctls".

Signed-off-by: Heiko Carstens <>
Cc: Ankit Jain <>
Acked-by: Christoph Hellwig <>
Cc: Al Viro <>
Acked-by: Arnd Bergmann <>
Acked-by: David S. Miller <>
Cc: <> [2.6.31.x]
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agopidns: fix a leak in /proc dentries and inodes with pid namespaces.
Sukadev Bhattiprolu [Wed, 11 Nov 2009 22:26:32 +0000 (14:26 -0800)]
pidns: fix a leak in /proc dentries and inodes with pid namespaces.

Daniel Lezcano reported a leak in 'struct pid' and 'struct pid_namespace'
that is discussed in:

To summarize the thread, when container-init is terminated, it sets the
PF_EXITING flag, zaps other processes in the container and waits to reap
them.  As a part of reaping, the container-init should flush any /proc
dentries associated with the processes.  But because the container-init is
itself exiting and the following PF_EXITING check, the dentries are not
flushed, resulting in leak in /proc inodes and dentries.

This fix reverts the commit 7766755a2f249e7e0 ("Fix /proc dcache deadlock
in do_exit") which introduced the check for PF_EXITING.  At the time of
the commit, shrink_dcache_parent() flushed dentries from other filesystems
also and could have caused a deadlock which the commit fixed.  But as
pointed out by Eric Biederman, after commit 0feae5c47aabdde59,
shrink_dcache_parent() no longer affects other filesystems.  So reverting
the commit is now safe.

As pointed out by Jan Kara, the leak is not as critical since the
unclaimed space will be reclaimed under memory pressure or by:

echo 3 > /proc/sys/vm/drop_caches

But since this check is no longer required, its best to remove it.

Signed-off-by: Sukadev Bhattiprolu <>
Reported-by: Daniel Lezcano <>
Acked-by: Eric W. Biederman <>
Acked-by: Jan Kara <>
Cc: Andrea Arcangeli <>
Cc: Serge Hallyn <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoalpha: move THREAD_SIZE definition outside #ifndef ASSEMBLY
Andrew Morton [Wed, 11 Nov 2009 22:26:30 +0000 (14:26 -0800)]
alpha: move THREAD_SIZE definition outside #ifndef ASSEMBLY

arch/alpha/kernel/ uses it:

arch/alpha/kernel/ undefined symbol `THREAD_SIZE' referenced in expression

Seems to have been caused by

commit 9d93f00580243cc059510d9d6ac4d2f5e97e5b83
Author:     Geoffrey Thomas <>
AuthorDate: Thu Sep 24 10:36:26 2009 -0400
Commit:     Linus Torvalds <>
CommitDate: Thu Sep 24 17:16:22 2009 -0700

    alpha: Clean up linker script using new linker script macros.

    Note that .data.page_aligned and .data.cacheline_aligned are now after
    _data; it was probably a bug that they were before it.

    Also, some explicit ALIGN(8)'s between various initcall sections were
    removed; this should be harmless as the implicit alignment of
    initcall_t was already 8.

Cc: Geoffrey Thomas <>
Cc: Tim Abbott <>
Cc: Ivan Kokshaysky <>
Cc: Sam Ravnborg <>
Cc: Richard Henderson <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoalpha: fix F_SETOWN_EX and F_GETLK64 conflict
Peter Zijlstra [Wed, 11 Nov 2009 22:26:27 +0000 (14:26 -0800)]
alpha: fix F_SETOWN_EX and F_GETLK64 conflict

Fix a bug in

    commit ba0a6c9f6fceed11c6a99e8326f0477fe383e6b5
    Author:     Peter Zijlstra <>
    AuthorDate: Wed Sep 23 15:57:03 2009 -0700
    Commit:     Linus Torvalds <>
    CommitDate: Thu Sep 24 07:21:01 2009 -0700

        fcntl: add F_[SG]ETOWN_EX

In asm-generic/fcntl.h, F_SETOWN_EX and F_GETLK64 both have value 12, and
F_GETOWN_EX and F_SETLK64 both have value 13.

Reported-by: "Joseph S. Myers" <>
Cc: Richard Henderson <>
Cc: Ivan Kokshaysky <>
Cc: Andreas Schwab <>
Signed-off-by: Peter Zijlstra <>
Acked-by: Ulrich Drepper <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agomemcg: fix wrong pointer initialization at page migration when memcg is disabled.
KAMEZAWA Hiroyuki [Wed, 11 Nov 2009 22:26:26 +0000 (14:26 -0800)]
memcg: fix wrong pointer initialization at page migration when memcg is disabled.

Lee Schermerhorn reported that he saw bad pointer dereference in
mem_cgroup_end_migration() when he disabled memcg by boot option.

memcg's page migration logic works as

mem_cgroup_prepare_migration(page, &ptr);
do page migration
mem_cgroup_end_migration(page, ptr);

Now, ptr is not initialized in prepare_migration when memcg is disabled
by boot option. This causes panic in end_migration. This patch fixes it.

Reported-by: Lee Schermerhorn <>
Cc: Balbir Singh <>
Signed-off-by: KAMEZAWA Hiroyuki <>
Reviewed-by: Daisuke Nishimura <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agosavagefb: fix blanking mode on CRT display
Krzysztof Helt [Wed, 11 Nov 2009 22:26:25 +0000 (14:26 -0800)]
savagefb: fix blanking mode on CRT display

Fix wrong bit mask for blanking register.  Due to the error a CRT monitor
blanks off due to wrong frequency (out of range) instead of PM signal
(vertical and horizontal frequencies cut off).

Just compare the mask with bits set in the switch(blank) clause below the
changed line.

Signed-off-by: Krzysztof Helt <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agofb: remove fb_save_state() and fb_restore_state operations
Krzysztof Helt [Wed, 11 Nov 2009 22:26:22 +0000 (14:26 -0800)]
fb: remove fb_save_state() and fb_restore_state operations

Remove fb_save_state() and fb_restore_state operations from frame buffer layer.
They are used only in two drivers:
1. savagefb  - and cause bug #11248
2. uvesafb

Usage of these operations is misunderstood in both drivers so kill these
operations, fix the bug #11248 and avoid confusion in the future.

Tested on Savage 3D/MV card and the patch fixes the bug #11248.

The frame buffer layer uses these funtions during switch between graphics
and text mode of the console, but these drivers saves state before
switching of the frame buffer (in the fb_open) and after releasing it (in
the fb_release).  This defeats the purpose of these operations.


Signed-off-by: Krzysztof Helt <>
Reported-by: Jochen Hein <>
Tested-by: Jochen Hein <>
Cc: Geert Uytterhoeven <>
Cc: Michal Januszewski <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agopage allocator: Do not allow interrupts to use ALLOC_HARDER
Mel Gorman [Wed, 11 Nov 2009 22:26:17 +0000 (14:26 -0800)]
page allocator: Do not allow interrupts to use ALLOC_HARDER

Commit 341ce06f69abfafa31b9468410a13dbd60e2b237 ("page allocator:
calculate the alloc_flags for allocation only once") altered watermark
logic slightly by allowing rt_tasks that are handling an interrupt to set
ALLOC_HARDER.  This patch brings the watermark logic more in line with

This change results in a reduction of the number high-order GFP_ATOMIC
allocation failures reported.  See

[ Spotted the problem]
Signed-off-by: Mel Gorman <>
Reviewed-by: Pekka Enberg <>
Reviewed-by: Rik van Riel <>
Reviewed-by: KOSAKI Motohiro <>
Cc: David Rientjes <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agopage allocator: always wake kswapd when restarting an allocation attempt after direct...
Mel Gorman [Wed, 11 Nov 2009 22:26:14 +0000 (14:26 -0800)]
page allocator: always wake kswapd when restarting an allocation attempt after direct reclaim failed

If a direct reclaim makes no forward progress, it considers whether it
should go OOM or not.  Whether OOM is triggered or not, it may retry the
allocation afterwards.  In times past, this would always wake kswapd as
well but currently, kswapd is not woken up after direct reclaim fails.
For order-0 allocations, this makes little difference but if there is a
heavy mix of higher-order allocations that direct reclaim is failing for,
it might mean that kswapd is not rewoken for higher orders as much as it
did previously.

This patch wakes up kswapd when an allocation is being retried after a
direct reclaim failure.  It would be expected that kswapd is already
awake, but this has the effect of telling kswapd to reclaim at the higher
order as well.

Signed-off-by: Mel Gorman <>
Reviewed-by: Christoph Lameter <>
Reviewed-by: Pekka Enberg <>
Reviewed-by: KOSAKI Motohiro <>
Cc: <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoMAINTAINERS: openipmi list is moderated
Randy Dunlap [Wed, 11 Nov 2009 22:26:13 +0000 (14:26 -0800)]
MAINTAINERS: openipmi list is moderated

openipmi list is moderated.

Signed-off-by: Randy Dunlap <>
Acked-by: Corey Minyard <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoget_maintainer: fix usage comment
Roel Kluin [Wed, 11 Nov 2009 22:26:13 +0000 (14:26 -0800)]
get_maintainer: fix usage comment

With the `s' it just won't work.

Signed-off-by: Roel Kluin <>
Acked-by: Joe Perches <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoMAINTAINERS: VMWARE VMXNET3 - Quote name with comma and period, use tabs
Joe Perches [Wed, 11 Nov 2009 22:26:12 +0000 (14:26 -0800)]
MAINTAINERS: VMWARE VMXNET3 - Quote name with comma and period, use tabs

Names with periods or commas need to be quoted
Use tab not spaces

Signed-off-by: Joe Perches <>
Cc: Shreyas Bhatewara <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoMAINTAINERS: SERVER ENGINES 10Gbps iSCSI - Use single line M:
Joe Perches [Wed, 11 Nov 2009 22:26:11 +0000 (14:26 -0800)]

Integrate P:/M: lines to single M:
Use tabs not spaces

Signed-off-by: Joe Perches <>
Cc: Jayamohan Kallickal <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoMAINTAINERS: BROCADE BFA - Use single line M: and tabs
Joe Perches [Wed, 11 Nov 2009 22:26:10 +0000 (14:26 -0800)]
MAINTAINERS: BROCADE BFA - Use single line M: and tabs

Integrate P:/M: to single M:
Use tab for spacing

Signed-off-by: Joe Perches <>
Cc: Jing Huang <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agoMAINTAINERS: ASUS ACPI EXTRAS - remove F:arch/x86/kernel/acpi/boot.c
Joe Perches [Wed, 11 Nov 2009 22:26:09 +0000 (14:26 -0800)]
MAINTAINERS: ASUS ACPI EXTRAS - remove F:arch/x86/kernel/acpi/boot.c

Oops.  How did that get there?
(Don't look, it's my original pattern commit...)

Signed-off-by: Joe Perches <>
Signed-off-by: Andrew Morton <>
Signed-off-by: Linus Torvalds <>
13 years agofs/jbd: Export log_start_commit to fix ext3 build.
Stefan Schmidt [Thu, 12 Nov 2009 08:53:50 +0000 (09:53 +0100)]
fs/jbd: Export log_start_commit to fix ext3 build.

This fixes:
ERROR: "log_start_commit" [fs/ext3/ext3.ko] undefined!

Signed-off-by: Stefan Schmidt <>
13 years agomd: factor out updating of 'recovery_offset'.
NeilBrown [Thu, 12 Nov 2009 01:08:04 +0000 (12:08 +1100)]
md: factor out updating of 'recovery_offset'.

Each device has its own 'recovery_offset' showing how far
recovery has progressed on the device.
As the only real significance of this is that fact that it can
be stored in the metadata and recovered at restart, and as
only 1.x metadata can do this, we were only updating
'recovery_offset' to 'curr_resync_completed' when updating
v1.x metadata.
But this is wrong, and we will shortly make limited use of this
field in v0.90 metadata.

So move the update into common code.

Signed-off-by: NeilBrown <>
13 years agoomap3: Decrease cpufreq transition latency
Mike Turquette [Wed, 11 Nov 2009 19:00:38 +0000 (11:00 -0800)]
omap3: Decrease cpufreq transition latency

Adjust OMAP3 frequency transition latency from 10,000,000uS to a more
reasonable 300,000uS.  This causes ondemand and conservative governors to
sample CPU load more often resulting in more responsive behavior.

Tested on Android 2.6.29; using this value and conservative governor, CORE
power consumption on Zoom2 was comparable to the old and unresponsive
10,000,000uS value while UI responsiveness was greatly improved.

Signed-off-by: Mike Turquette <>
Signed-off-by: Kevin Hilman <>
Signed-off-by: Tony Lindgren <>
13 years agoMerge git://
Linus Torvalds [Wed, 11 Nov 2009 21:38:59 +0000 (13:38 -0800)]
Merge git://git./linux/kernel/git/mason/btrfs-unstable

* git://
  Btrfs: fix panic when trying to destroy a newly allocated
  Btrfs: allow more metadata chunk preallocation
  Btrfs: fallback on uncompressed io if compressed io fails
  Btrfs: find ideal block group for caching
  Btrfs: avoid null deref in unpin_extent_cache()
  Btrfs: skip btrfs_release_path in btrfs_update_root and btrfs_del_root
  Btrfs: fix some metadata enospc issues
  Btrfs: fix how we set max_size for free space clusters
  Btrfs: cleanup transaction starting and fix journal_info usage
  Btrfs: fix data allocation hint start

13 years agobtusb bluetooth driver: wait for 'waker' work too before closing
Linus Torvalds [Wed, 11 Nov 2009 21:32:29 +0000 (13:32 -0800)]
btusb bluetooth driver: wait for 'waker' work too before closing

Rafael debugged a resume-time hang (with oopses in workqueue handling)
on his laptop that was due to the 'waker' workqueue entry being
disconnected and then released without the workqueue entry having been

Several people were involved, with Oleg Nesterov doing a debugging patch
showing what workqueue entry was corrupt etc.

This was a regression introduced by commit 7bee549e19 ("Bluetooth: Add
USB autosuspend support to btusb driver") as Rafael points out (not
actually bisected, but it became clear once the bug was found).

Tested-and-reported-by: Rafael J. Wysocki <>
Acked-by: Oliver Neukum <>
Cc: Oleg Nesterov <>
Cc: Tejun Heo <>
Cc: Marcel Holtmann <>
Signed-off-by: Linus Torvalds <>
13 years agoBtrfs: fix panic when trying to destroy a newly allocated
Josef Bacik [Wed, 11 Nov 2009 20:53:34 +0000 (15:53 -0500)]
Btrfs: fix panic when trying to destroy a newly allocated

There is a problem where iget5_locked will look for an inode, not find it, and
then subsequently try to allocate it.  Another CPU will have raced in and
allocated the inode instead, so when iget5_locked gets the inode spin lock again
and does a search, it finds the new inode.  So it goes ahead and calls
destroy_inode on the inode it just allocated.  The problem is we don't set
BTRFS_I(inode)->root until the new inode is completely initialized.  This patch
makes us set root to NULL when alloc'ing a new inode, so when we get to
btrfs_destroy_inode and we see that root is NULL we can just free up the memory
and continue on.  This fixes the panic


Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
13 years agoMerge branch 'for_linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:52:22 +0000 (11:52 -0800)]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs-2.6

* 'for_linus' of git://
  JBD/JBD2: free j_wbuf if journal init fails.
  ext3: Wait for proper transaction commit on fsync
  ext3: retry failed direct IO allocations

13 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:35:04 +0000 (11:35 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound-2.6

* 'for-linus' of git://
  ALSA: hda - possible read past array alc88[02]_parse_auto_config()
  sound: rawmidi: fix MIDI device O_APPEND error handling
  sound: rawmidi: fix checking of O_APPEND when opening MIDI device
  sound: rawmidi: fix double init when opening MIDI device with O_APPEND
  ALSA: hda - Avoid quirk for HP dc5750
  ALSA: hda - Tweak OLPC XO-1.5 microphone bias
  ALSA: hda: Use model=auto quirk for Sony VAIO VGN-FW170J using ALC262

13 years agoMerge branch 'for-linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:34:14 +0000 (11:34 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/jbarnes/pci-2.6

* 'for-linus' of git://
  x86/PCI: Adjust GFP mask handling for coherent allocations
  PCI ASPM: fix oops on root port removal

13 years agoMerge branch 'merge' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:33:08 +0000 (11:33 -0800)]
Merge branch 'merge' of git://git./linux/kernel/git/benh/powerpc

* 'merge' of git://
  powerpc: pasemi_defconfig update
  powerpc: 2.6.32 update of defconfigs for embedded 6xx/7xxx, 8xx, 8{3,5,6}xxx
  powerpc/8xxx: enable IPsec ESP by default on mpc83xx/mpc85xx
  powerpc/83xx: Fix u-boot partion size for MPC8377E-WLAN boards
  powerpc/85xx: Fix USB GPIOs for MPC8569E-MDS boards
  powerpc/82xx: kmalloc failure ignored in ep8248e_mdio_probe()
  powerpc/85xx: sbc8548 - fixup of PCI-e related DTS fields

13 years agoMerge
Linus Torvalds [Wed, 11 Nov 2009 19:32:42 +0000 (11:32 -0800)]

  ARM: Use a definition for the userspace cmpxchg emulation syscall
  [ARM] Fix test for unimplemented ARM syscalls
  ARM: 5784/1: fix early boot machine ID mismatch error display
  [ARM] orion5x: update defconfig
  [ARM] Kirkwood: update defconfig
  [ARM] Kirkwood: clarify PCIe MEM bus/physical address distinction
  [ARM] kirkwood: fix PCI I/O port assignment
  [ARM] kirkwood: fix section mismatch
  [ARM] OpenRD base: Initialize PCI express and i2c
  [ARM] properly report mv78100 stepping A1
  ARM: 5780/1: KS8695: Fix macro definition bug in regs-switch.c
  ARM: 5779/1: ep93xx/micro9.c: fix implicit declaration of function __raw_readl and IOMEM
  RealView: Add sparsemem support for the RealView PBX platform
  RealView: Remove duplicated #define REALVIEW_SYS_FLAGS* statements
  RealView: Add default memory configuration
  Check whether the SCU was already initialised
  ARMv7: Check whether the SMP/nAMP mode was already enabled
  [ARM] pxa: fix resume failure by saving/restoring IPRx registers
  [ARM] pxa/palm: fix incorrect initialization of Palm Tungsten C keyboard
  [ARM] pxa/zaurus: fix NAND flash OOB layout for Borzoi

13 years agoMerge branch 'drm-linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:32:04 +0000 (11:32 -0800)]
Merge branch 'drm-linus' of git://git./linux/kernel/git/airlied/drm-2.6

* 'drm-linus' of git:// (52 commits)
  drm/kms: Init the CRTC info fields for modes forced from the command line.
  drm/radeon/r600: CS parser updates
  drm/radeon/kms: add debugfs for power management for AtomBIOS devices
  drm/radeon/kms: initial mode validation support
  drm/radeon/kms/atom/dce3: call transmitter init on mode set
  drm/radeon/kms: store detailed connector info
  drm/radeon/kms/atom/dce3: fix up usPixelClock calculation for Transmitter tables
  drm/radeon/kms/r600: fix rs880 support v2
  drm/radeon/kms/r700: fix some typos in chip init
  drm/radeon/kms: remove some misleading debugging output
  drm/radeon/kms: stop putting VRAM at 0 in MC space on r600s.
  drm/radeon/kms: disable D1VGA and D2VGA if enabled
  drm/radeon/kms: Don't RMW CP_RB_CNTL
  drm/radeon/kms: fix coherency issues on AGP cards.
  drm/radeon/kms: fix rc410 suspend/resume.
  drm/radeon/kms: add quirk for hp dc5750
  drm/radeon/kms/atom: fix potential oops in spread spectrum code
  drm/kms: typo fix
  drm/radeon/kms/atom: Make card_info per device
  drm/radeon/kms/atom: Fix DVO support

13 years agoMerge branch 'core-fixes-for-linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:30:15 +0000 (11:30 -0800)]
Merge branch 'core-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-fixes-for-linus' of git://
  highmem: Fix debug_kmap_atomic() to also handle KM_IRQ_PTE, KM_NMI, and KM_NMI_PTE
  highmem: Fix race in debug_kmap_atomic() which could cause warn_count to underflow
  rcu: Fix long-grace-period race between forcing and initialization
  uids: Prevent tear down race

13 years agoMerge branch 'irq-fixes-for-linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:29:58 +0000 (11:29 -0800)]
Merge branch 'irq-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'irq-fixes-for-linus' of git://
  genirq: try_one_irq() must be called with irq disabled

13 years agoMerge branch 'perf-fixes-for-linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:29:34 +0000 (11:29 -0800)]
Merge branch 'perf-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'perf-fixes-for-linus' of git://
  perf tools: Fix permission checks
  perf_events: Fix some typo in the perf events config description

13 years agoMerge branch 'sched-fixes-for-linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:29:24 +0000 (11:29 -0800)]
Merge branch 'sched-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'sched-fixes-for-linus' of git://
  sched: Use root_task_group_empty only with FAIR_GROUP_SCHED
  sched: Fix kernel-doc function parameter name

13 years agoMerge branch 'x86-fixes-for-linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:29:10 +0000 (11:29 -0800)]
Merge branch 'x86-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'x86-fixes-for-linus' of git://
  x86, amd-ucode: Check UCODE_MAGIC before loading the container file
  x86: Fix error return sequence in __ioremap_caller()
  x86: Add Phoenix/MSC BIOSes to lowmem corruption list

13 years agoMerge branch 'for_linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:28:11 +0000 (11:28 -0800)]
Merge branch 'for_linus' of git://git./linux/kernel/git/tytso/ext4

* 'for_linus' of git://
  ext4: partial revert to fix double brelse WARNING()
  ext4: Fix return value of ext4_split_unwritten_extents() to fix direct I/O
  ext4: code clean up for dio fallocate handling
  ext4: skip conversion of uninit extents after direct IO if there isn't any
  ext4: fix ext4_ext_direct_IO()'s return value after converting uninit extents
  ext4: discard preallocation when restarting a transaction during truncate

13 years agoMerge git://
Linus Torvalds [Wed, 11 Nov 2009 19:27:49 +0000 (11:27 -0800)]
Merge git://git./linux/kernel/git/wim/linux-2.6-watchdog

* git://
  [WATCHDOG] SBC-FITPC2 watchdog driver registration fix

13 years agoMerge branch 'fixes-s3c-2632-rc6' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:27:26 +0000 (11:27 -0800)]
Merge branch 'fixes-s3c-2632-rc6' of git://

* 'fixes-s3c-2632-rc6' of git://
  ARM: S3C64XX: DMA: Free node for non-circular queues
  ARM: S3C64XX: DMA: Callback with correct buffer pointer
  ARM: S3C64XX: DMA: Make src and dst transfer size same
  ARM: S3C64XX: DMA: Unify callback functions for success/failure
  ARM: S3C64XX: DMA: Protect buffer pointers while manipulation
  ARM: S3C64XX: Tidy definition and comments in s3c_dma_has_circular()
  ARM: S3C64XX: Remove duplicate s3c_dma_has_circular() definition for S3C64xx.
  ARM: SMDK6410: Allocate more GPIO space for WM1190-EV1
  ARM: SMDK6410: Configure GPIO pull up for WM835x IRQ line

13 years agoMerge branch 'for_linus' of git://
Linus Torvalds [Wed, 11 Nov 2009 19:26:42 +0000 (11:26 -0800)]
Merge branch 'for_linus' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'for_linus' of git:// (27 commits)
  V4L/DVB (13314): saa7134: set ts_force_val for the Hauppauge WinTV HVR-1150
  V4L/DVB (13313): saa7134: add support for FORCE_TS_VALID mode for mpeg ts input
  V4L/DVB (13311): uvcvideo: Fix compilation warning with 2.6.32 due to type mismatch with abs()
  V4L/DVB (13309): uvcvideo: Ignore the FIX_BANDWIDTH for compressed video
  V4L/DVB (13287): ce6230 - saa7164-cmd: Fix wrong sizeof
  V4L/DVB (13286): pxa-camera: Fix missing sched.h
  V4L/DVB (13264): gspca_mr97310a: Change vstart for CIF sensor type 1 cams
  V4L/DVB (13257): gspca - m5602-s5k4aa: Add vflip for Fujitsu Amilo Xi 2528
  V4L/DVB (13256): gspca - m5602-s5k4aa: Add another MSI GX700 vflip quirk
  V4L/DVB (13255): gspca - m5602-s5k4aa: Add vflip quirk for the Bruneinit laptop
  V4L/DVB (13240): firedtv: fix regression: tuning fails due to bogus error return
  V4L/DVB (13237): firedtv: length field corrupt in ca2host if length>127
  V4L/DVB (13230): s2255drv: Don't conditionalize video buffer completion on waiting processes
  V4L/DVB (13202): smsusb: add autodetection support for three additional Hauppauge USB IDs
  V4L/DVB (13190): em28xx: fix panic that can occur when starting audio streaming
  V4L/DVB (13170): bttv: Fix reversed polarity error when switching video standard
  V4L/DVB (13169): bttv: Fix potential out-of-order field processing
  V4L/DVB (13167): pt1: Fix a compile error on arm
  V4L/DVB (13132): fix use-after-free Oops, resulting from a driver-core API change
  V4L/DVB (13131): pxa_camera: fix camera pixel format configuration

13 years agoBtrfs: allow more metadata chunk preallocation
Chris Mason [Wed, 11 Nov 2009 15:16:57 +0000 (10:16 -0500)]
Btrfs: allow more metadata chunk preallocation

On an FS where all of the space has not been allocated into chunks yet,
the enospc can return enospc just because the existing metadata chunks
are full.

We get around this by allowing more metadata chunks to be allocated up
to a certain limit, and finding the right limit is a little fuzzy.  The
problem is the reservations for delalloc would preallocate way too much
of the FS as metadata.  We need to start saying no and just force some
IO to happen.

But we also need to let a reasonable amount of the FS become metadata.
This bumps the hard limit up, later releases will have a better system.

Signed-off-by: Chris Mason <>
13 years agoBtrfs: fallback on uncompressed io if compressed io fails
Josef Bacik [Wed, 11 Nov 2009 02:23:48 +0000 (21:23 -0500)]
Btrfs: fallback on uncompressed io if compressed io fails

Currently compressed IO does not deal with not having its entire extent able to
be allocated.  So if we have enough free space to allocate for the extent, but
its not contiguous, it will fail spectacularly.  This patch fixes this by
falling back on uncompressed IO which lets us spread the delalloc extent across
multiple extents.  I tested this by making us randomly think the reservation had
failed to make it fallback on the uncompressed io way and it seemed to work
fine.  Thanks,

Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
13 years agoBtrfs: find ideal block group for caching
Josef Bacik [Wed, 11 Nov 2009 02:23:48 +0000 (21:23 -0500)]
Btrfs: find ideal block group for caching

This patch changes a few things.  Hopefully the comments are helpfull, but
I'll try and be as verbose here.


My fedora box was taking 1 minute and 21 seconds to boot with btrfs as root.
Part of this problem was we pick the first block group we can find and start
caching it, even if it may not have enough free space.  The other problem is
we only search for cached block groups the first time around, which we won't
find any cached block groups because this is a newly mounted fs, so we end up
caching several block groups during bootup, which with alot of fragmentation
takes around 30-45 seconds to complete, which bogs down the system.  So


1) Don't cache block groups willy-nilly at first.  Instead try and figure out
which block group has the most free, and therefore will take the least amount
of time to cache.

2) Don't be so picky about cached block groups.  The other problem is once
we've filled up a cluster, if the block group isn't finished caching the next
time we try and do the allocation we'll completely ignore the cluster and
start searching from the beginning of the space, which makes us cache more
block groups, which slows us down even more.  So instead of skipping block
groups that are not finished caching when we have a hint, only skip the block
group if it hasn't started caching yet.

There is one other tweak in here.  Before if we allocated a chunk and still
couldn't find new space, we'd end up switching the space info to force another
chunk allocation.  This could make us end up with way too many chunks, so keep
track of this particular case.

With this patch and my previous cluster fixes my fedora box now boots in 43
seconds, and according to the bootchart is not held up by our block group
caching at all.

Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
13 years agoBtrfs: avoid null deref in unpin_extent_cache()
Dan Carpenter [Tue, 10 Nov 2009 09:01:43 +0000 (09:01 +0000)]
Btrfs: avoid null deref in unpin_extent_cache()

I re-orderred the checks to avoid dereferencing "em" if it was null.

Found by smatch static checker.

Signed-off-by: Dan Carpenter <>
Signed-off-by: Chris Mason <>
13 years agoBtrfs: skip btrfs_release_path in btrfs_update_root and btrfs_del_root
Li Dongyang [Fri, 6 Nov 2009 14:33:01 +0000 (14:33 +0000)]
Btrfs: skip btrfs_release_path in btrfs_update_root and btrfs_del_root

We don't need to call btrfs_release_path because btrfs_free_path will do
that for us.

Signed-off-by: Li Dongyang <>
Signed-off-by: Chris Mason <>
13 years agoBtrfs: fix some metadata enospc issues
Josef Bacik [Wed, 11 Nov 2009 02:23:48 +0000 (21:23 -0500)]
Btrfs: fix some metadata enospc issues

We weren't reserving metadata space for rename, rmdir and unlink, which could
cause problems.

Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
13 years agoBtrfs: fix how we set max_size for free space clusters
Josef Bacik [Wed, 11 Nov 2009 02:23:48 +0000 (21:23 -0500)]
Btrfs: fix how we set max_size for free space clusters

This patch fixes a problem where max_size can be set to 0 even though we
filled the cluster properly.  We set max_size to 0 if we restart the cluster
window, but if the new start entry is big enough to be our new cluster then we
could return with a max_size set to 0, which will mean the next time we try to
allocate from this cluster it will fail.  So set max_extent to the entry's
size.  Tested this on my box and now we actually allocate from the cluster
after we fill it.  Thanks,

Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
13 years agoBtrfs: cleanup transaction starting and fix journal_info usage
Josef Bacik [Wed, 11 Nov 2009 02:23:48 +0000 (21:23 -0500)]
Btrfs: cleanup transaction starting and fix journal_info usage

We use journal_info to tell if we're in a nested transaction to make sure we
don't commit the transaction within a nested transaction.  We use another
method to see if there are any outstanding ioctl trans handles, so if we're
starting one do not set current->journal_info, since it will screw with other
filesystems.  This patch also cleans up the starting stuff so there aren't any
magic numbers.

Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
13 years agoBtrfs: fix data allocation hint start
Josef Bacik [Wed, 11 Nov 2009 02:23:47 +0000 (21:23 -0500)]
Btrfs: fix data allocation hint start

Sometimes our start allocation hint when we cow a file can be either
EXTENT_HOLE or some other such place holder, which is not optimal.  So if we
find that our em->block_start is one of these special values, check to see
where the first block of the inode is stored, and use that as a hint.  If that
block is also a special value, just fallback on a hint of 0 and let the
allocator figure out a good place to put the data.

Signed-off-by: Josef Bacik <>
Signed-off-by: Chris Mason <>
13 years agoomap3: update Pandora defconfig
Grazvydas Ignotas [Wed, 11 Nov 2009 19:00:38 +0000 (11:00 -0800)]
omap3: update Pandora defconfig

This patch updates defconfig to enable options needed to properly
boot OMAP3 pandora board. It also enables MMC, OTG, GPIO LEDs,
TWL4030 GPIO and sound drivers.

Signed-off-by: Grazvydas Ignotas <>
Signed-off-by: Tony Lindgren <>
13 years agoomap3: 3430sdp: Enable Linux Regulator framework
Sergio Aguirre [Wed, 11 Nov 2009 19:00:37 +0000 (11:00 -0800)]
omap3: 3430sdp: Enable Linux Regulator framework

Some drivers have dependencies on this, and therefore should be

Signed-off-by: Sergio Aguirre <>
Signed-off-by: Tony Lindgren <>
13 years agoomap3: beagle: Fix USB host port power control
Jarkko Nikula [Wed, 11 Nov 2009 19:00:36 +0000 (11:00 -0800)]
omap3: beagle: Fix USB host port power control

The host port power is enabled by driving the nEN_USB_PWR low as stated in
the comment. This fix is originally from Steve Sakoman <>.

Signed-off-by: Jarkko Nikula <>
Cc: Steve Sakoman <>
Signed-off-by: Tony Lindgren <>
13 years agoomap3: pandora: Fix keypad keymap
Grazvydas Ignotas [Wed, 11 Nov 2009 19:00:35 +0000 (11:00 -0800)]
omap3: pandora: Fix keypad keymap

The original TWL4030 keypad driver from linux-omap used KEY()
macro defined as (col, row), but while it was merged upstream
it was changed to use matrix keypad infrastructure, which uses
(row, col) format. Update the keymap in board file to match
layout of mainline driver.

Signed-off-by: Grazvydas Ignotas <>
Signed-off-by: Tony Lindgren <>
13 years agoomap1: Amstrad Delta defconfig fixes
Janusz Krzysztofik [Wed, 11 Nov 2009 19:00:34 +0000 (11:00 -0800)]
omap1: Amstrad Delta defconfig fixes

The patch provides the following fixes:

- keep kernel small enough to boot with standard tools,
- ensure compatibility with both new and legacy distros,
- turn on support for recently added or fixed hardware features.

Created and tested against linux-2.6.32-rc5.

Signed-off-by: Janusz Krzysztofik <>
Signed-off-by: Tony Lindgren <>
13 years agoomap: Fix omapfb/lcdc on OMAP1510 broken when PM set
Janusz Krzysztofik [Wed, 11 Nov 2009 19:00:03 +0000 (11:00 -0800)]
omap: Fix omapfb/lcdc on OMAP1510 broken when PM set

With CONFIG_PM=y, the omapfb/lcdc device on Amstrad Delta, after initially
starting correctly, breaks with the following error messages:

omapfb omapfb: resetting (status 0xffffff96,reset count 1)
omapfb omapfb: resetting (status 0xffffff96,reset count 100)
omapfb omapfb: too many reset attempts, giving up.

Looking closer at this I have found that it had been broken almost 2 years ago
with commit 2418996e3b100114edb2ae110d5d4acb928909d2, PM fixes for OMAP1.

The definite reason for broken omapfb/lcdc behavoiur in PM mode
appeared to be ARM_IDLECT1:IDLIF_ARM (bit 6) put into idle regardless of LCD
DMA possibly running. The bit were set based on return value of the
omap_dma_running() function that did not check for dedicated LCD DMA
channel status. The patch below fixes this.

Note that the hardcoded register value will be fixed during the next merge
cycle to use OMAP_LCDC_ defines. Currently the OMAP_LCDC_ defines are local
to drivers/video/omap/lcdc.c, so let's not start moving those right now.

Created against linux-2.6.32-rc6

Tested on Amstrad Delta

Signed-off-by: Janusz Krzysztofik <>
Signed-off-by: Tony Lindgren <>
13 years agoJBD/JBD2: free j_wbuf if journal init fails.
Tao Ma [Tue, 10 Nov 2009 09:13:22 +0000 (17:13 +0800)]
JBD/JBD2: free j_wbuf if journal init fails.

If journal init fails, we need to free j_wbuf.

Cc: Andrew Morton <>
Cc: Jan Kara <>
Signed-off-by: Tao Ma <>
Signed-off-by: Jan Kara <>
13 years agoext3: Wait for proper transaction commit on fsync
Jan Kara [Fri, 16 Oct 2009 17:26:15 +0000 (19:26 +0200)]
ext3: Wait for proper transaction commit on fsync

We cannot rely on buffer dirty bits during fsync because pdflush can come
before fsync is called and clear dirty bits without forcing a transaction
commit. What we do is that we track which transaction has last changed
the inode and which transaction last changed allocation and force it to
disk on fsync.

Signed-off-by: Jan Kara <>
Reviewed-by: Aneesh Kumar K.V <>
13 years agoext3: retry failed direct IO allocations
Eric Sandeen [Tue, 13 Oct 2009 02:34:27 +0000 (21:34 -0500)]
ext3: retry failed direct IO allocations

On a 256M 4k block filesystem, doing this in a loop:

    dd if=/dev/zero of=test oflag=direct bs=1M count=64
    rm -f test

eventually leads to spurious ENOSPC:

    dd: writing `test': No space left on device

As with other block allocation callers, it looks like we need to
potentially retry the allocations on the initial ENOSPC.

A similar patch went into ext4 (commit

Signed-off-by: Eric Sandeen <>
Signed-off-by: Jan Kara <>
13 years agoMerge branch 'fix/hda' into for-linus
Takashi Iwai [Wed, 11 Nov 2009 07:10:31 +0000 (08:10 +0100)]
Merge branch 'fix/hda' into for-linus

13 years agoALSA: hda - possible read past array alc88[02]_parse_auto_config()
Roel Kluin [Tue, 10 Nov 2009 19:11:55 +0000 (20:11 +0100)]
ALSA: hda - possible read past array alc88[02]_parse_auto_config()

The test of index `i' is after the read - too late - and
unsafe: if snd_hda_get_connections() fails in the last
iteration a read beyond the array is possible.

Signed-off-by: Roel Kluin <>
Signed-off-by: Takashi Iwai <>
13 years agoInput: fix locking in memoryless force-feedback devices
Dmitry Torokhov [Sat, 7 Nov 2009 05:39:07 +0000 (21:39 -0800)]
Input: fix locking in memoryless force-feedback devices

Now that input core acquires dev->event_lock spinlock and disables
interrupts when propagating input events, using spin_lock_bh() in
ff-memless driver is not allowed. Actually, the timer_lock itself
is not needed anymore, we should simply use dev->event_lock
as well.

Also do a small cleanup in force-feedback core.

Reported-by: Arjan van de Ven <>
Signed-off-by: Dmitry Torokhov <>
13 years agopowerpc: pasemi_defconfig update
Olof Johansson [Tue, 6 Oct 2009 15:52:50 +0000 (10:52 -0500)]
powerpc: pasemi_defconfig update

pasemi_defconfig hasn't been updated for a year.

Mostly a refresh of defaults, but this also disables 64K pages.

Signed-off-by: Olof Johansson <>
Signed-off-by: Benjamin Herrenschmidt <>
13 years agoMerge commit 'kumar/merge' into merge
Benjamin Herrenschmidt [Wed, 11 Nov 2009 04:43:25 +0000 (15:43 +1100)]
Merge commit 'kumar/merge' into merge

13 years agoomap: Use resource_size
Tobias Klauser [Wed, 11 Nov 2009 02:55:19 +0000 (18:55 -0800)]
omap: Use resource_size

Use the resource_size function instead of manually calculating the
resource size. This reduces the chance of introducing off-by-one errors
and actually fixes one in mailbox.c.

Signed-off-by: Tobias Klauser <>
Signed-off-by: Tony Lindgren <>
13 years agoomap: Fix race condition in omap dma driver
Tao Hu [Wed, 11 Nov 2009 02:55:17 +0000 (18:55 -0800)]
omap: Fix race condition in omap dma driver

The bug could cause irq enable bit of one DMA channel is
cleared/set unexpectedly when 2 (or more) drivers are calling
omap_request_dma()/omap_free_dma() simultaneously

Signed-off-by: Fei Yang <>
Signed-off-by: Tao Hu <>
Signed-off-by: Tony Lindgren <>
13 years agoMerge branch 'for-rmk' of git://
Russell King [Tue, 10 Nov 2009 23:37:21 +0000 (23:37 +0000)]
Merge branch 'for-rmk' of git://

13 years agosound: rawmidi: fix MIDI device O_APPEND error handling
Clemens Ladisch [Wed, 21 Oct 2009 07:11:43 +0000 (09:11 +0200)]
sound: rawmidi: fix MIDI device O_APPEND error handling

Commit 9a1b64caac82aa02cb74587ffc798e6f42c6170a in 2.6.30 broke the
error handling code in rawmidi_open_priv().

If only the output substream of a RawMIDI device has been opened and
if this device is then opened with O_RDWR | O_APPEND and if the
initialization of the input substream fails (either because of low
memory or because the device driver's open callback fails), then the
runtime structure of the already open output substream will be freed
and all following writes through the first handle will cause
snd_rawmidi_write() to use the NULL runtime pointer.

Signed-off-by: Clemens Ladisch <>
Cc: <>
Signed-off-by: Takashi Iwai <>