splice: fix i_mutex locking in generic_splice_write()

Rearrange locking of i_mutex on destination so it's only held while
buffers are copied with the pipe_to_file() actor, and not while
waiting for more data on the pipe.

Signed-off-by: Miklos Szeredi <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

splice: remove i_mutex locking in splice_from_pipe()

splice_from_pipe() is only called from two places:

  - generic_splice_sendpage()
  - splice_write_null()

Neither of these require i_mutex to be taken on the destination inode.

Signed-off-by: Miklos Szeredi <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

splice: split up __splice_from_pipe()

Split up __splice_from_pipe() into four helper functions:

  splice_from_pipe_begin()
  splice_from_pipe_next()
  splice_from_pipe_feed()
  splice_from_pipe_end()

splice_from_pipe_next() will wait (if necessary) for more buffers to
be added to the pipe.  splice_from_pipe_feed() will feed the buffers
to the supplied actor and return when there's no more data available
(or if all of the requested data has been copied).

This is necessary so that implementations can do locking around the
non-waiting splice_from_pipe_feed().

This patch should not cause any change in behavior.

Signed-off-by: Miklos Szeredi <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

block: fix SG_IO to return a proper error value

blk_rq_unmap_user() returns -EFAULT if a program passes an invalid
address to kernel. SG_IO path needs to pass the returned value to user
space instead of ignoring it.

Signed-off-by: FUJITA Tomonori <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

cfq-iosched: don't delay queue kick for a merged request

"Zhang, Yanmin" <[email protected]> reports that commit
b029195dda0129b427c6e579a3bb3ae752da3a93 introduced a regression
of about 50% with sequential threaded read workloads. The test
case is:

tiotest -k0 -k1 -k3 -f 80 -t 32

which starts 32 threads each reading a 80MB file. Twiddle the kick
queue logic so that we do start IO immediately, if it appears to be
a fully merged request. We can't really detect that, so just check
if the request is bigger than a page or not. The assumption is that
since single bio issues will first queue a single request with just
one page attached and then later do merges on that, if we already
have more than a page worth of data in the request, then the request
is most likely good to go.

Verified that this doesn't cause a regression with the test case that
commit b029195dda0129b427c6e579a3bb3ae752da3a93 was fixing. It does not,
we still see maximum sized requests for the queue-then-merge cases.

Signed-off-by: Jens Axboe <[email protected]>

buffer: switch do_emergency_thaw() away from pdflush_operation()

This is (again) a preparatory patch similar to commit
a2a9537ac0b37a5da6fbe7e1e9cb06c524d2a9c4. It open codes a simple
async way of executing do_thaw_all() out of context, so we can get
rid of pdflush.

Signed-off-by: Jens Axboe <[email protected]>

block: update biodoc.txt on plugging

We do per-device plugging, get rid of any references to tq_disk as that
has been dead since 2.6.5 or so.

Signed-off-by: Jens Axboe <[email protected]>

as-iosched: get rid of private REQ_SYNC/REQ_ASYNC defines

We can just use the block layer BLK_RW_SYNC/ASYNC defines now.

Signed-off-by: Jens Axboe <[email protected]>

cfq-iosched: get rid of private SYNC/ASYNC defines

We can just use the block layer BLK_RW_SYNC/ASYNC defines now.

Signed-off-by: Jens Axboe <[email protected]>

cfq-iosched: use rw_is_sync() to see if rw flags are sync or not

Signed-off-by: Jens Axboe <[email protected]>

Document and move the various READ/WRITE types

It's a somewhat twisty maze of hints and behavioural modifiers, try
and clear it up a bit with some documentation.

Signed-off-by: Jens Axboe <[email protected]>

block: fix bad spelling of quiesce

Credit goes to Andrew Morton for spotting this one.

Signed-off-by: Jens Axboe <[email protected]>

block: move bio list helpers into bio.h

It's used by DM and MD and generally useful, so move the bio list
helpers into bio.h.

Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Alasdair G Kergon <[email protected]>
Signed-off-by: Jens Axboe <[email protected]>

Linux 2.6.30-rc2

Merge branch 'drm-intel-next' of git://git./linux/kernel/git/anholt/drm-intel

* 'drm-intel-next' of git://git.kernel.org/pub/scm/linux/kernel/git/anholt/drm-intel:
  drm/i915: fix scheduling while holding the new active list spinlock
  drm/i915: Allow tiling of objects with bit 17 swizzling by the CPU.
  drm/i915: Correctly set the write flag for get_user_pages in pread.
  drm/i915: Fix use of uninitialized var in 40a5f0de
  drm/i915: indicate framebuffer restore key in SysRq help message
  drm/i915: sync hdmi detection by hdmi identifier with 2D
  drm/i915: Fix a mismerge of the IGD patch (new .find_pll hooks missed)
  drm/i915: Implement batch and ring buffer dumping

x86 microcode: revert some work_on_cpu

Revert part of af5c820a3169e81af869c113e18ec7588836cd50 ("x86: cpumask:
use work_on_cpu in arch/x86/kernel/microcode_core.c")

That change is causing only one Intel CPU's microcode to be updated e.g.
microcode: CPU3 updated from revision 0x9 to 0x17, date = 2005-04-22
where before it announced that also for CPU0 and CPU1 and CPU2.

We cannot use work_on_cpu() in the CONFIG_MICROCODE_OLD_INTERFACE code,
because Intel's request_microcode_user() involves a copy_from_user() from
/sbin/microcode_ctl, which therefore needs to be on that CPU at the time.

Signed-off-by: Hugh Dickins <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drm/i915: fix scheduling while holding the new active list spinlock

regression caused by commit 5e118f4139feafe97e913df67b1f7c1e5083e535:
i915_gem_object_move_to_inactive() should be called in task context,
as it calls fput();

Signed-off-by: Shaohua Li<[email protected]>
[anholt: Add more detail to the comment about the lock break that's added]
Signed-off-by: Eric Anholt <[email protected]>

Merge branch 'core-fixes-for-linus' of git://git./linux/kernel/git/tip/linux-2.6-tip

* 'core-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
  lockdep: warn about lockdep disabling after kernel taint, fix

Merge branch 'for-linus' of git://git./linux/kernel/git/mszeredi/fuse

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
  fuse: fix "direct_io" private mmap
  fuse: fix argument type in fuse_get_user_pages()

Merge branch 'for-linus' of git://git./linux/kernel/git/ryusuke/nilfs2

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2:
  nilfs2: fix possible mismatch of sufile counters on recovery
  nilfs2: segment usage file cleanups
  nilfs2: fix wrong accounting and duplicate brelse in nilfs_sufile_set_error
  nilfs2: simplify handling of active state of segments fix
  nilfs2: remove module version
  nilfs2: fix lockdep recursive locking warning on meta data files
  nilfs2: fix lockdep recursive locking warning on bmap
  nilfs2: return f_fsid for statfs2

Merge branch 'fixes-for-linus' of git://git.monstr.eu/linux-2.6-microblaze

* 'fixes-for-linus' of git://git.monstr.eu/linux-2.6-microblaze:
  microblaze: Add missing FILE tag to MAINTAINERS
  microblaze: remove duplicated #include's
  microblaze: struct device - replace bus_id with dev_name()
  microblaze: Simplify copy_thread()
  microblaze: Add TIMESTAMPING constants to socket.h
  microblaze: Add missing empty ftrace.h file
  microblaze: Fix problem with removing zero length files

Merge git://git./linux/kernel/git/lethal/sh-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-2.6:
  sh: Add in PCI bus for DMA API debugging.
  sh: Pre-allocate a reasonable number of DMA debug entries.
  sh: sh7786: modify usb setup timeout judgment bug.
  MAINTAINERS: Update sh architecture file patterns.
  sh: ap325: use edge control for ov772x camera
  sh: Plug in support for ARCH=sh64 using sh SRCARCH.
  sh: urquell: Fix up address mapping in board comments.
  sh: Add support for DMA API debugging.
  sh: Provide cpumask_of_pcibus() to fix NUMA build.
  sh: urquell: Add board comment
  sh: wire up sys_preadv/sys_pwritev() syscalls.
  sh: sh7785lcr: fix PCI address map for 32-bit mode
  sh: intc: Added resume from hibernation support to the intc

Fix lpfc_parse_bg_err()'s use of do_div()

Fix lpfc_parse_bg_err()'s use of do_div().  It should be passing a 64-bit
variable as the first parameter.  However, since it's only using a 32-bit
variable, it doesn't need to use do_div() at all, but can instead use the
division operator.

This deals with the following warnings:

    CC      drivers/scsi/lpfc/lpfc_scsi.o
  drivers/scsi/lpfc/lpfc_scsi.c: In function 'lpfc_parse_bg_err':
  drivers/scsi/lpfc/lpfc_scsi.c:1397: warning: comparison of distinct pointer types lacks a cast
  drivers/scsi/lpfc/lpfc_scsi.c:1397: warning: right shift count >= width of type
  drivers/scsi/lpfc/lpfc_scsi.c:1397: warning: passing argument 1 of '__div64_32' from incompatible pointer type

Signed-off-by: David Howells <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

tty: Update some of the USB kernel doc

Updates some usb_serial_port members documentation.

Signed-off-by: Leandro Dorileo <[email protected]>
Signed-off-by: Alan Cox <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

parport_pc: Fix build failure drivers/parport/parport_pc.c for powerpc

In commit 51dcdfec6a274afc1c6fce180d582add9ff512c0 ("parport: Use the
PCI IRQ if offered") parport_pc_probe_port() gained an irqflags arg.
This isn't being supplied on powerpc.  This patch make powerpc fallback
to the old behaviour, that is using "0" for irqflags.

Fixes build failure:

  In file included from drivers/parport/parport_pc.c:68:
  arch/powerpc/include/asm/parport.h: In function 'parport_pc_find_nonpci_ports':
  arch/powerpc/include/asm/parport.h:32: error: too few arguments to function 'parport_pc_probe_port'
  arch/powerpc/include/asm/parport.h:32: error: too few arguments to function 'parport_pc_probe_port'
  arch/powerpc/include/asm/parport.h:32: error: too few arguments to function 'parport_pc_probe_port'
  make[3]: *** [drivers/parport/parport_pc.o] Error 1

Signed-off-by: Tony Breeds <[email protected]>
Signed-off-by: Alan Cox <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

parport: Fix various uses of parport_pc

These got overlooked first time around.

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

tty: Fix leak in ti-usb

If the ti-usb adapter returns an zero data length frame (which happens)
then we leak a kref.  Found by Christoph Mair <[email protected]>
who proposed a patch.  The patch here is different as Christoph's patch
didn't work for the case where tty = NULL and data arrived but Christoph
did all the hard work chasing it down.

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

cdc-acm: Fix long standing abuse of tty->low_latency

ACM sets the low latency flag but calls the flip buffer routines from
IRQ context which isn't permitted (and as of 2.6.29 causes a warning
hence this one was caught)

Fortunatelt ACM doesn't need to set this flag in the first place as it
only set it to work around problems in ancient (pre tty flip rewrite)
kernels.

Reported-by: Chuck Ebbert <[email protected]>
Signed-off-by: Alan Cox <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

microblaze: Add missing FILE tag to MAINTAINERS

Signed-off-by: Michal Simek <[email protected]>

lockdep: warn about lockdep disabling after kernel taint, fix

Impact: build fix for Sparc and s390

Stephen Rothwell reported that the Sparc build broke:

 In file included from kernel/panic.c:12:
 include/linux/debug_locks.h: In function '__debug_locks_off':
 include/linux/debug_locks.h:15: error: implicit declaration of function 'xchg'

due to:

 9eeba61: lockdep: warn about lockdep disabling after kernel taint

There is some inconsistency between architectures about where exactly
xchg() is defined.

The traditional place is in system.h but the more logical point for it
is in atomic.h - where most architectures (especially new ones) have
it defined. These architecture also still offer it via system.h.

Some, such as Sparc or s390 only have it in asm/system.h and not available
via asm/atomic.h at all.

Use the widest set of headers in debug_locks.h and also include asm/system.h.

Reported-by: Stephen Rothwell <[email protected]>
Cc: Frederic Weisbecker <[email protected]>
Cc: "David S. Miller" <[email protected]>
Cc: Linus Torvalds <[email protected]>
LKML-Reference: <20090414144317.026498df[email protected]>
Signed-off-by: Ingo Molnar <[email protected]>

microblaze: remove duplicated #include's

Remove duplicated #include's in
  arch/microblaze/include/asm/io.h
  arch/microblaze/kernel/prom.c
  arch/microblaze/kernel/ptrace.c
  arch/microblaze/kernel/signal.c
  arch/microblaze/kernel/sys_microblaze.c

Signed-off-by: Huang Weiyi <[email protected]>
Signed-off-by: Michal Simek <[email protected]>

microblaze: struct device - replace bus_id with dev_name()

Signed-off-by: Michal Simek <[email protected]>

microblaze: Simplify copy_thread()

Removing nr parameter from copy_tread function
commit 6f2c55b843836d26528c56a0968689accaedbc67

Signed-off-by: Michal Simek <[email protected]>

microblaze: Add TIMESTAMPING constants to socket.h

This changes was introduce with
commit: cb9eff097831007afb30d64373f29d99825d0068

net: new user space API for time stamping of incoming and outgoing packets

Signed-off-by: Michal Simek <[email protected]>

microblaze: Add missing empty ftrace.h file

Signed-off-by: Michal Simek <[email protected]>

microblaze: Fix problem with removing zero length files

Adding one new line was recommended solution.
Test with make distclean

Tested-by: Jeff Garzik <[email protected]>
Signed-off-by: Michal Simek <[email protected]>

sh: Add in PCI bus for DMA API debugging.

This adds in the pci_bus_type for DMA API debug.

Signed-off-by: Paul Mundt <[email protected]>

sh: Pre-allocate a reasonable number of DMA debug entries.

This prevents the DMA API debugging from running out of entries right
away on boot. Defines 4096 entries by default, which while a bit on the
heavy side, ought to leave enough breathing room for some time.

Signed-off-by: Paul Mundt <[email protected]>

Merge branch 'for-linus' of git://git./linux/kernel/git/jmorris/security-testing-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/security-testing-2.6:
  tomoyo: version bump to 2.2.0.
  tomoyo: add Documentation/tomoyo.txt

Fix quilt merge error in acpi-cpufreq.c

We ended up incorrectly using '&cur' instead of '&readin' in the
work_on_cpu() -> smp_call_function_single() transformation in commit
01599fca6758d2cd133e78f87426fc851c9ea725 ("cpufreq: use
smp_call_function_[single|many]() in acpi-cpufreq.c").

Andrew explains:
 "OK, the acpi tree went and had conflicting changes merged into it after
  I'd written the patch and it appears that I incorrectly reverted part
  of 18b2646fe3babeb40b34a0c1751e0bf5adfdc64c while fixing the resulting
  rejects.

  Switching it to `readin' looks correct."

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

tomoyo: version bump to 2.2.0.

Signed-off-by: Kentaro Takeda <[email protected]>
Signed-off-by: Tetsuo Handa <[email protected]>
Signed-off-by: Toshiharu Harada <[email protected]>
Signed-off-by: James Morris <[email protected]>

tomoyo: add Documentation/tomoyo.txt

Signed-off-by: Kentaro Takeda <[email protected]>
Signed-off-by: Tetsuo Handa <[email protected]>
Signed-off-by: Toshiharu Harada <[email protected]>
Signed-off-by: James Morris <[email protected]>

Merge branch 'for-rc1/xen/core' of git://git./linux/kernel/git/jeremy/xen

* 'for-rc1/xen/core' of git://git.kernel.org/pub/scm/linux/kernel/git/jeremy/xen:
  xen: add FIX_TEXT_POKE to fixmap
  xen: honour VCPU availability on boot
  xen: clean up gate trap/interrupt constants
  xen: set _PAGE_NX in __supported_pte_mask before pagetable construction
  xen: resume interrupts before system devices.
  xen/mmu: weaken flush_tlb_other test
  xen/mmu: some early pagetable cleanups
  Xen: Add virt_to_pfn helper function
  x86-64: remove PGE from must-have feature list
  xen: mask XSAVE from cpuid
  NULL noise: arch/x86/xen/smp.c
  xen: remove xen_load_gdt debug
  xen: make xen_load_gdt simpler
  xen: clean up xen_load_gdt
  xen: split construction of p2m mfn tables from registration
  xen: separate p2m allocation from setting
  xen: disable preempt for leave_lazy_mmu

sh: sh7786: modify usb setup timeout judgment bug.

This corrects a race with the PHY RST bit not being set properly if the
PLL status changes right before timeout. This resulted in it potentially
failing even if the device came up in time.

Special thanks to Mr. Juha Leppanen and Iwamatsu-san for reporting this
out and reviewing it.

Reported-by: Juha Leppanen <[email protected]>
Reviewed-by: Nobuhiro Iwamatsu <[email protected]>
Tested-by: Paul Mundt <[email protected]>
Signed-off-by: Kuninori Morimoto <[email protected]>
Signed-off-by: Paul Mundt <[email protected]>

edac: use to_delayed_work()

The edac-core driver includes code which assumes that the work_struct
which is included in every delayed_work is the first member of that
structure.  This is currently the case but might change in the future, so
use to_delayed_work() instead, which doesn't make such an assumption.

linux-2.6.30-rc1 has the to_delayed_work() function that will allow this
patch to work

Signed-off-by: Jean Delvare <[email protected]>
Signed-off-by: Doug Thompson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

edac: fix local pci_write_bits32

Fix the edac local pci_write_bits32 to properly note the 'escape' mask if
all ones in a 32-bit word.

Currently no consumer of this function uses that mask, so there is no
danger to existing code.

Signed-off-by: Jeff Haran <[email protected]>
Signed-off-by: Doug Thompson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sgi-xpc: clean up numerous globals

Introduce xpc_arch_ops and eliminate numerous individual global definitions.

Signed-off-by: Robin Holt <[email protected]>
Cc: Dean Nelson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sgi-xpc: implement opencomplete messaging

sgi-xpc has a window of failure where an open message can be sent and a
subsequent data message can get lost.  We have added a new message
(opencomplete) which closes that window.

Signed-off-by: Robin Holt <[email protected]>
Signed-off-by: Dean Nelson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sgi-xpc: prevent false heartbeat failures

The heartbeat timeout functionality in sgi-xpc is currently not trained to
the connection time.  If a connection is made and the code is in the last
polling window prior to doing a timeout, the next polling window will see
the heartbeat as unchanged and initiate a no-heartbeat disconnect.

Signed-off-by: Robin Holt <[email protected]>
Signed-off-by: Dean Nelson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sgi-xpc: update SGI XP/XPC/XPNET maintainer

Dean has moved on to other work.  His responsibilities for XP/XPC/XPNET
have been handed to me.

Signed-off-by: Robin Holt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

doc: use correct debugfs mountpoint

Use the default mountpoint of debugfs in the pktcdvd ABI.

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

memcg: fix documentation

The description about various statistics from memory.stat is not accurate
and confusing at times.

Correct this along with a few other minor cleanups.

Signed-off-by: Bharata B Rao <[email protected]>
Acked-by: Balbir Singh <[email protected]>
Acked-by: KAMEZAWA Hiroyuki <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Replace all DMA_nBIT_MASK macro with DMA_BIT_MASK(n)

This is the second go through of the old DMA_nBIT_MASK macro,and there're not
so many of them left,so I put them into one patch.I hope this is the last round.
After this the definition of the old DMA_nBIT_MASK macro could be removed.

Signed-off-by: Yang Hongyang <[email protected]>
Cc: Russell King <[email protected]>
Cc: Tony Lindgren <[email protected]>
Cc: "David S. Miller" <[email protected]>
Cc: James Bottomley <[email protected]>
Cc: Greg KH <[email protected]>
Cc: Takashi Iwai <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

ext2: fix data corruption for racing writes

If two writers allocating blocks to file race with each other (e.g.
because writepages races with ordinary write or two writepages race with
each other), ext2_getblock() can be called on the same inode in parallel.
Before we are going to allocate new blocks, we have to recheck the block
chain we have obtained so far without holding truncate_mutex.  Otherwise
we could overwrite the indirect block pointer set by the other writer
leading to data loss.

The below test program by Ying is able to reproduce the data loss with ext2
on in BRD in a few minutes if the machine is under memory pressure:

long kMemSize  = 50 << 20;
int kPageSize = 4096;

int main(int argc, char **argv) {
int status;
int count = 0;
int i;
char *fname = "/mnt/test.mmap";
char *mem;
unlink(fname);
int fd = open(fname, O_CREAT | O_EXCL | O_RDWR, 0600);
status = ftruncate(fd, kMemSize);
mem = mmap(0, kMemSize, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0);
// Fill the memory with 1s.
memset(mem, 1, kMemSize);
sleep(2);
for (i = 0; i < kMemSize; i++) {
int byte_good = mem[i] != 0;
if (!byte_good && ((i % kPageSize) == 0)) {
//printf("%d ", i / kPageSize);
count++;
}
}
munmap(mem, kMemSize);
close(fd);
unlink(fname);

if (count > 0) {
printf("Running %d bad page\n", count);
return 1;
}
return 0;
}

Cc: Ying Han <[email protected]>
Cc: Nick Piggin <[email protected]>
Signed-off-by: Jan Kara <[email protected]>
Cc: Mingming Cao <[email protected]>
Cc: <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

shmem: respect MAX_LFS_FILESIZE

SHMEM_MAX_BYTES was derived from the maximum size of its triple-indirect
swap vector, forgetting to take the MAX_LFS_FILESIZE limit into account.
Never mind 256kB pages, even 8kB pages on 32-bit kernels allowed files to
grow slightly bigger than that supposed maximum.

Fix this by using the min of both (at build time not run time).  And it
happens that this calculation is good as far as 8MB pages on 32-bit or
16MB pages on 64-bit: though SHMSWP_MAX_INDEX gets truncated before that,
it's truncated to such large numbers that we don't need to care.

[[email protected]: it needs pagemap.h]
[[email protected]: fix sparc64 min() warnings]
Signed-off-by: Hugh Dickins <[email protected]>
Cc: Yuri Tikhonov <[email protected]>
Cc: Paul Mackerras <[email protected]>
Cc: Randy Dunlap <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

shmem: fix division by zero

Fix a division by zero which we have in shmem_truncate_range() and
shmem_unuse_inode() when using big PAGE_SIZE values (e.g.  256kB on
ppc44x).

With 256kB PAGE_SIZE, the ENTRIES_PER_PAGEPAGE constant becomes too large
(0x1.0000.0000) on a 32-bit kernel, so this patch just changes its type
from 'unsigned long' to 'unsigned long long'.

Hugh: reverted its unsigned long longs in shmem_truncate_range() and
shmem_getpage(): the pagecache index cannot be more than an unsigned long,
so the divisions by zero occurred in unreached code.  It's a pity we need
any ULL arithmetic here, but I found no pretty way to avoid it.

Signed-off-by: Yuri Tikhonov <[email protected]>
Signed-off-by: Hugh Dickins <[email protected]>
Cc: Paul Mackerras <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

intelfb: support i854

Support the Intel 854 Chipset in fbdev.

We test and use the patch on a Thomson IP1101 IPTV-Box.  On the VGA-Port
we get a normal signal.

Here is the link to the Mambux-Project: http://www.mambux.de

Cc: Keith Packard <[email protected]>
Cc: Dave Airlie <[email protected]>
Cc: Krzysztof Helt <[email protected]>
Signed-off-by: Stefan Husemann <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm: add documentation describing what tsk->active_mm means vs tsk->mm

I'm sure everyone knows this, but I didn't, so I googled it, and found a
nice explanation from Linus.  Might be worth sticking in Documentation.

Signed-off-by: Michael Ellerman <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

memcg: remove warning when CONFIG_DEBUG_VM=n

mm/memcontrol.c:318: warning: `mem_cgroup_is_obsolete' defined but not used

[[email protected]: simplify as suggested by Balbir]
Signed-off-by: KAMEZAWA Hiroyuki <[email protected]>
Reviewed-by: Balbir Singh <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

kernel/sys.c: clean up sys_shutdown exit path

Impact: cleanup, fix

Clean up sys_shutdown() exit path.  Factor out common code.  Return
correct error code instead of always 0 on failure.

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

at25: make input buffers of at25_*write() const

| drivers/misc/eeprom/at25.c:358: warning: assignment from incompatible pointer type

Signed-off-by: Geert Uytterhoeven <[email protected]>
Cc: Kevin Hilman <[email protected]>
Cc: David Brownell <[email protected]>
Cc: Jean Delvare <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

at24: make input buffers of at24_*write() const

| drivers/misc/eeprom/at24.c:508: warning: assignment from incompatible pointer type

Signed-off-by: Geert Uytterhoeven <[email protected]>
Cc: Kevin Hilman <[email protected]>
Cc: David Brownell <[email protected]>
Cc: Jean Delvare <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

jbd: update locking coments

Update information about locking in JBD revoke code.

Reported-by: Lin Tan <[email protected]>.
Signed-off-by: Jan Kara <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm: document get_user_pages_fast()

While better than get_user_pages(), the usage of gupf(), especially the
return values and the fact that it can potentially only partially pin the
range, warranted some documentation.

Signed-off-by: Andy Grover <[email protected]>
Cc: Ingo Molnar <[email protected]>
Cc: Nick Piggin <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

initramfs: fix initramfs to work with hardlinked init

Change cb6ff208076b5f434db1b8c983429269d719cef5 ("NOMMU: Support XIP on
initramfs") seems to have broken booting from initramfs with /sbin/init
being a hardlink.

It seems like the logic required for XIP on nommu, i.e.  ftruncate to
reported cpio header file size (body_len) is broken for hardlinks, which
have a reported size of 0, and the truncate thus nukes the contents of the
file (in my case busybox), making boot impossible and ending with runaway
loop modprobe binfmt-0000 - and of course 0000 is not a valid binary
format.

My fix is to only call ftruncate if size is non-zero which fixes things
for me, but I'm not certain whether this will break XIP for those files on
nommu systems, although I would guess not.

Signed-off-by: Randy Robertson <[email protected]>
Acked-by: David Howells <[email protected]>
Acked-by: Paul Mundt <[email protected]>
Acked-by: H. Peter Anvin <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

ptrace: fix exit_ptrace() vs ptrace_traceme() race

Pointed out by Roland.  The bug was recently introduced by me in
"forget_original_parent: split out the un-ptrace part", commit
39c626ae47c469abdfd30c6e42eff884931380d6.

Since that patch we have a window after exit_ptrace() drops tasklist and
before forget_original_parent() takes it again.  In this window the child
can do ptrace(PTRACE_TRACEME) and nobody can untrace this child after
that.

Change ptrace_traceme() to not attach to the exiting ->real_parent.  We
don't report the error in this case, we pretend we attach right before
->real_parent calls exit_ptrace() which should untrace us anyway.

Signed-off-by: Oleg Nesterov <[email protected]>
Acked-by: Roland McGrath <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

MAINTAINERS: bluesmoke-devel list is moderated for non-subscribers

Signed-off-by: Jean Delvare <[email protected]>
Acked-by: Doug Thompson <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm: reformat the Unevictable-LRU documentation

Do a bit of reformatting on the Unevictable-LRU documentation.

Signed-off-by: David Howells <[email protected]>
Acked-by: Lee Schermerhorn <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm: point the UNEVICTABLE_LRU config option at the documentation

Point the UNEVICTABLE_LRU config option at the documentation describing
the option.

Signed-off-by: David Howells <[email protected]>
Cc: Lee Schermerhorn <[email protected]>
Cc: Rik van Riel <[email protected]>
Cc: KOSAKI Motohiro <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sysrq: remove duplicated #include

Remove duplicated #include in drivers/char/sysrq.c.

Signed-off-by: Huang Weiyi <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

include/video/cyblafb.h: remove it, it's unused

commit ddb53d48da5b0e691f35e703ac29118747f86c99 ("fbdev: remove cyblafb
driver") removed drivers/video/cyblafb.c, but not its .h file

Signed-off-by: Joe Perches <[email protected]>
Cc: Krzysztof Helt <[email protected]>
Cc: "Jani Monoses" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

res_counter: update documentation

After the introduction of resource counters hierarchies
(28dbc4b6a01fb579a9441c7b81e3d3413dc452df) the prototypes of
res_counter_init() and res_counter_charge() have been changed.

Keep the documentation consistent with the actual function prototypes.

Signed-off-by: Andrea Righi <[email protected]>
Cc: Paul Menage <[email protected]>
Cc: Pavel Emelyanov <[email protected]>
Cc: Balbir Singh <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

spi: spi_write_then_read() bugfixes

The "simplify spi_write_then_read()" patch included two regressions from
the 2.6.27 behaviors:

 - The data it wrote out during the (full duplex) read side
   of the transfer was not zeroed.

 - It fails completely on half duplex hardware, such as
   Microwire and most "3-wire" SPI variants.

So, revert that patch.  A revised version should be submitted at some
point, which can get the speedup on standard hardware (full duplex)
without breaking on less-capable half-duplex stuff.

Signed-off-by: David Brownell <[email protected]>
Cc: <[email protected]> [2.6.28.x, 2.6.29.x]
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

asm-generic/siginfo.h: update NSIGTRAP definition

Impact: (nearly) trivial

The patch

commit da654b74bda14c45a7d98c731bf3c1a43b6b74e2
Author: Srinivasa Ds <[email protected]>
Date:   Tue Sep 23 15:23:52 2008 +0530

    signals: demultiplexing SIGTRAP signal

forgot to update the NSIGTRAP define in asm-generic/siginfo.h to the new
number of sigtrap subcodes.  Nothing in the tree seems to use it, but
presumably something in user space might.  So update it.

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

cirrusfb: do not allow unsupported pixel depth

Do not allow modes with unsupported pixel depth.  Otherwise, one can hang
a computer by setting incorrect value with fbset command.

Signed-off-by: Krzysztof Helt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

filemap: fix kernel-doc warnings

Fix filemap.c kernel-doc warnings:

Warning(mm/filemap.c:575): No description found for parameter 'page'
Warning(mm/filemap.c:575): No description found for parameter 'waiter'

Signed-off-by: Randy Dunlap <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

include/linux/fiemap.h: include types.h now that it's exported

Include <linux/types.h> in fiemap.h.  Sam Ravnborg pointed out that this
was missing in this newly-exported header which uses the __u32 and __u64
types.

Signed-off-by: Eric Sandeen <[email protected]>
Cc: Sam Ravnborg <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

drivers/video/aty/radeon_base.c: fix typo in comment

Signed-off-by: Paul Menzel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sisfb: fix color component length for pseudocolor modes

sisfb incorrectly sets the length of the color fields to 6 bits
for PSEUDOCOLOR modes, even though 8 bits are always used per pixel.
Fix this by setting the length to 8.

Signed-off-by: Michal Januszewski <[email protected]>
Cc: Thomas Winischhofer <[email protected]>
Acked-by: Krzysztof Helt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

sa1100fb: fix color component length for pseudocolor modes

sa1100fb incorrectly sets the length of the color fields to 8 bits for
PSEUDOCOLOR modes for which only 4 bits are used per pixel.  Fix this by
setting the length to 4 bits for these modes.

Signed-off-by: Michal Januszewski <[email protected]>
Cc: Krzysztof Helt <[email protected]>
Cc: Russell King <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

s3fb: fix color component length for pseudocolor modes

s3fb incorrectly sets the length of the color fields to 6 bits for
PSEUDOCOLOR modes, even though 8 or 4 bits are used per pixel.  Fix this
by setting the length to 8 or 4, respectively.

Signed-off-by: Michal Januszewski <[email protected]>
Acked-by: Ondrej Zajicek <[email protected]>
Cc: Krzysztof Helt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

namespaces: move get_mq() inside #ifdef CONFIG_SYSCTL

| ipc/mq_sysctl.c:26: warning: 'get_mq' defined but not used

Signed-off-by: Geert Uytterhoeven <[email protected]>
Acked-by: Serge Hallyn <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

aio: remove INIT_KIOCTX

Unused after 20dcae32439384b6863c626bb3b2a09bed65b33e aka
"[PATCH] aio: remove kioctx from mm_struct".

Signed-off-by: Alexey Dobriyan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

hfs: fix memory leak when unmounting

When an HFS filesystem is unmounted, it leaks a 2-page bitmap.  Also,
under extreme memory pressure, it's possible that hfs_releasepage() may
use a tree pointer that has not been initialized, and if so, the release
request should just be rejected.

[[email protected]: free_pages(0) is legal, remove obvious comment]
Signed-off-by: Dave Anderson <[email protected]>
Tested-by: Eugene Teo <[email protected]>
Cc: Roman Zippel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

hwmon: sht15 humidity sensor driver

Data sheet at:
http://www.sensirion.ch/en/pdf/product_information/Datasheet-humidity-sensor-SHT1x.pdf

These sensors communicate over a 2 wire bus running a device specific
protocol.  The complexity of the driver is mainly due to handling the
substantial delays between requesting a reading and the device pulling the
data line low to indicate that the data is available.  This is handled by
an interrupt that is disabled under all other conditions.

I wasn't terribly clear on the best way to handle this, so comments on
that aspect would be particularly welcome!

Interpretation of the temperature depends on knowing the supply voltage.
If configured in a board config as a regulator consumer this is obtained
from the regulator subsystem.  If not it should be provided in the
platform data.

I've placed this driver in the hwmon subsystem as it is definitely a
device that may be used for hardware monitoring and with it's relatively
slow response times (up to 120 millisecs to get a reading) a caching
strategy certainly seems to make sense!

Signed-off-by: Jonathan Cameron <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

efifb: exit if framebuffer address is invalid

efifb will attempt to ioremap a framebuffer even if its starting address
is 0, failing and causing an ugly backtrace in the process.  Exit before
probing if this is the case.

Signed-off-by: Matthew Garrett <[email protected]>
Acked-by: Peter Jones <[email protected]>
Cc: Krzysztof Helt <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

uvesafb: fix color component length for pseudocolor modes

uvesafb incorrectly sets the length of the color fields to 6 bits for
PSEUDOCOLOR modes, even though 8 bits are always used per pixel.  Fix this
by setting the length to 8.

The switch of the DAC width from the default 6 bits to 8 bits is retained
and tracked internally in the driver, but never exposed to userspace.

Signed-off-by: Michal Januszewski <[email protected]>
Acked-by: Krzysztof Helt <[email protected]>
Cc: <[email protected]>
Cc: Geert Uytterhoeven <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: fix color component field length documentation

The documentation about the meaning of the color component bitfield
lengths in pseudocolor modes is inconsistent.  Fix it, so that it
indicates the correct interpretation everywhere, i.e.  that 1 << length is
the number of palette entries.

Signed-off-by: Michal Januszewski <[email protected]>
Acked-by: Krzysztof Helt <[email protected]>
Cc: <[email protected]>
Acked-by: Geert Uytterhoeven <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

fbdev: fix info->lock deadlock in fbcon_event_notify()

fb_notifier_call_chain() is called with info->lock held, i.e.  in
do_fb_ioctl() => FBIOPUT_VSCREENINFO => fb_set_var() and the some
notifier callbacks, like fbcon_event_notify(), try to re-acquire
info->lock again.

Remove the lock/unlock_fb_info() in all the framebuffer notifier
callbacks' and be sure to always call fb_notifier_call_chain() with
info->lock held.

Reported-by: Pavel Roskin <[email protected]>
Reported-by: Eric Miao <[email protected]>
Signed-off-by: Andrea Righi <[email protected]>
Cc: Stefan Richter <[email protected]>
Cc: Krzysztof Helt <[email protected]>
Cc: Geert Uytterhoeven <[email protected]>
Cc: "Rafael J. Wysocki" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

init/initramfs: fix warning with CONFIG_BLK_DEV_RAM=n

init/initramfs.c:520: warning: 'clean_rootfs' defined but not used

Signed-off-by: Nikanth Karthikesan <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

hp_accel: remove unused #include <version.h>

Remove unused #include <version.h> in drivers/hwmon/hp_accel.c.

Signed-off-by: Huang Weiyi <[email protected]>
Acked-by: Pavel Machek <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Documentation/sysctl/net.txt: fix a typo

s/spicified/specified

Signed-off-by: Li Zefan <[email protected]>
Cc: "David S. Miller" <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

MAINTAINERS: fix Andreas's email address

Correct typo in email address.

Signed-off-by: Randy Dunlap <[email protected]>
Cc: Andreas Herrmann <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

MAINTAINERS: switch maintainer of the hso driver

Since Denis is no longer working for Option, I will maintain the driver from
now on.

Signed-off-by: Jan Dumon <[email protected]>
Cc: "David S. Miller" <[email protected]>
Cc: Greg KH <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

mm: move the scan_unevictable_pages sysctl to the vm table

vm knobs should go in the vm table.  Probably too late for
randomize_va_space though.

Signed-off-by: Peter Zijlstra <[email protected]>
Acked-by: Lee Schermerhorn <[email protected]>
Acked-by: Rik van Riel <[email protected]>
Signed-off-by: Andrew Morton <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>

Merge branch 'for-linus' of git://oss.sgi.com/xfs/xfs

* 'for-linus' of git://oss.sgi.com/xfs/xfs:
  xfs: remove xfs_flush_space
  xfs: flush delayed allcoation blocks on ENOSPC in create
  xfs: block callers of xfs_flush_inodes() correctly
  xfs: make inode flush at ENOSPC synchronous
  xfs: use xfs_sync_inodes() for device flushing
  xfs: inform the xfsaild of the push target before sleeping
  xfs: prevent unwritten extent conversion from blocking I/O completion
  xfs: fix double free of inode
  xfs: validate log feature fields correctly

MAINTAINERS: Update sh architecture file patterns.

Fill in missing entries for drivers/sh/ and Documentation/sh/.

Signed-off-by: Paul Mundt <[email protected]>

Merge branch 'master' of git://git./linux/kernel/git/torvalds/linux-2.6

Merge git://git./linux/kernel/git/sam/kbuild-fixes

* git://git.kernel.org/pub/scm/linux/kernel/git/sam/kbuild-fixes:
  docbook: make cleandocs
  kbuild: fix spurious initramfs rebuild
  Documentation: explain the difference between __bitwise and __bitwise__
  kbuild: make it possible for the linker to discard local symbols from vmlinux
  kbuild: remove pointless strdup() on arguments passed to new_module() in modpost
  kbuild: fix a few typos in top-level Makefile
  kbuild: introduce destination-y for exported headers
  kbuild: use git svn instead of git-svn in setlocalversion
  kconfig: fix update-po-config to accect backslash in input
  kbuild: fix option processing for -I in headerdep

Merge branch 'upstream-linus' of git://git./linux/kernel/git/jgarzik/libata-dev

* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
  ata: fix obviously wrong comment
  ahci: force CAP_NCQ for earlier NV MCPs
  [libata] sata_via: kill uninit'd var warning