net: fix double skb free in dcbnl

netlink_unicast() calls kfree_skb even in the error case.

dcbnl calls netlink_unicast() which when it fails free's the
skb and returns an error value.  dcbnl is free'ing the skb
again when this error occurs.  This patch removes the double
free.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix nlmsg len size for skb when error bit is set.

Currently, the nlmsg->len field is not set correctly in  netlink_ack()
for ack messages that include the nlmsg of the error frame.  This
corrects the length field passed to __nlmsg_put to use the correct
payload size.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix vlan_get_size to include vlan_flags size

Fix vlan_get_size to include vlan->flags.  Currently, the
size of the vlan flags is not included in the nlmsg size.

Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: cleanup unused prototype

The function e1000_enable_tx_pkt_filtering() was removed in
a previous cleanup patch.  this removes the no longer used
prototype.

Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: fix namespacecheck warnings

a couple of functions needed to be removed/declared static

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: drop unused functionality for eeprom write/read

eerd and eewr don't exist on pre PCIe devices

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: updated whitespace and comments

A large whitespace change to e1000_hw.[ch] in order to update it to kernel coding
style (by running lindent).  Updated function header comments into kdoc style.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: drop redunant line of code, cleanup

adapter was being assigned twice, also clarified variable name and unwrapped
line.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: remove races when changing mtu

this patch fixes a bug that occurs when routing packets and simultaneously
changing the mtu.  the rx_buffer_len variable is used during the rx cleanup
and if that changes on the fly without stopping traffic bad things happen

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: two workarounds were incomplete, fix them

1) 82544 does not need last_tx_tso workaround, it interferes with the 82544
workaround too
2) 82544 hang workaround was using the address of the page struct instead of
the physical address as its "workaround decider" not sure how that ever worked

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: fix tx waking queue after queue stopped during shutdown

This fix closes a race where the adapter can be shutting down while
hard_start_xmit is being called and interrupts are being handled.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: test link state conclusively

e1000 was using one particular way to detect link, but with the advent
of some of the newer hardware designs using SERDES connections, tests
for link must completely cover all cases.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: stop timers at appropriate times

there were some hotplug cases that made timers still run after the driver
had been removed, make sure to stop all the timers and not allow racy
reschedules.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: use netif_tx_disable

we can use netif_tx_disable now because LLTX has been removed.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: remove unused functions

after removal of pcie, need to remove some unnecessary functions

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

e1000: drop dead pcie code from e1000

this patch is the first in a series of clean up patches for e1000 to drop
unused code, and update the driver to kernel spec, and then, to update the
driver to have all available bug fixes.

Call it the e1000 weight loss plan.

Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Signed-off-by: Don Skidmore <donald.c.skidmore@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

at91_can: Forgotten git 'add' of at91_can.c

Signed-off-by: David S. Miller <davem@davemloft.net>

TI Davinci EMAC: Fix in vector definition for EMAC_VERSION_2

In the emac_poll function when looking for interrupt status masks
correct definition must be chosen based on EMAC_VERSION(the bit
mask has changed from version 1 to version 2).

Signed-off-by: Sriram <srk@ti.com>
Acked-by: Chaithrika U S <chaithrika@ti.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ax25: Fix ax25_cb refcounting in ax25_ctl_ioctl

Use ax25_cb_put after ax25_find_cb in ax25_ctl_ioctl.

Reported-by: Bernard Pidoux F6BVP <f6bvp@free.fr>
Signed-off-by: Jarek Poplawski <jarkao2@gmail.com>
Reviewed-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of ssh:///linux/kernel/git/rusty/linux-2.6-for-davem

genetlink: fix netns vs. netlink table locking (2)

Similar to commit d136f1bd366fdb7e747ca7e0218171e7a00a98a5,
there's a bug when unregistering a generic netlink family,
which is caught by the might_sleep() added in that commit:

    BUG: sleeping function called from invalid context at net/netlink/af_netlink.c:183
    in_atomic(): 1, irqs_disabled(): 0, pid: 1510, name: rmmod
    2 locks held by rmmod/1510:
     #0:  (genl_mutex){+.+.+.}, at: [<ffffffff8138283b>] genl_unregister_family+0x2b/0x130
     #1:  (rcu_read_lock){.+.+..}, at: [<ffffffff8138270c>] __genl_unregister_mc_group+0x1c/0x120
    Pid: 1510, comm: rmmod Not tainted 2.6.31-wl #444
    Call Trace:
     [<ffffffff81044ff9>] __might_sleep+0x119/0x150
     [<ffffffff81380501>] netlink_table_grab+0x21/0x100
     [<ffffffff813813a3>] netlink_clear_multicast_users+0x23/0x60
     [<ffffffff81382761>] __genl_unregister_mc_group+0x71/0x120
     [<ffffffff81382866>] genl_unregister_family+0x56/0x130
     [<ffffffffa0007d85>] nl80211_exit+0x15/0x20 [cfg80211]
     [<ffffffffa000005a>] cfg80211_exit+0x1a/0x40 [cfg80211]

Fix in the same way by grabbing the netlink table lock
before doing rcu_read_lock().

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

3c59x: Get rid of "Trying to free already-free IRQ"

Following trace pops up if we try to suspend with 3c59x ethernet NIC
brought down:

  root@b1:~# ifconfig eth16 down
  root@b1:~# echo mem > /sys/power/state
  ...
  3c59x 0000:00:10.0: suspend
  3c59x 0000:00:10.0: PME# disabled
  Trying to free already-free IRQ 48
  ------------[ cut here ]------------
  Badness at c00554e4 [verbose debug info unavailable]
  NIP: c00554e4 LR: c00554e4 CTR: c019a098
  REGS: c7975c60 TRAP: 0700   Not tainted  (2.6.31-rc4)
  MSR: 00021032 <ME,CE,IR,DR>  CR: 28242422  XER: 20000000
  TASK = c79cb0c0[1746] 'bash' THREAD: c7974000
  ...
  NIP [c00554e4] __free_irq+0x108/0x1b0
  LR [c00554e4] __free_irq+0x108/0x1b0
  Call Trace:
  [c7975d10] [c00554e4] __free_irq+0x108/0x1b0 (unreliable)
  [c7975d30] [c005559c] free_irq+0x10/0x24
  [c7975d40] [c01e21ec] vortex_suspend+0x70/0xc4
  [c7975d60] [c017e584] pci_legacy_suspend+0x58/0x100

This is because the driver manages interrupts without checking for
netif_running().

Though, there are few other issues with suspend/resume in this driver.
The intention of calling free_irq() in suspend() was to avoid any
possible spurious interrupts (see commit 5b039e681b8c5f30aac9cc04385
"3c59x PM fixes"). But,

- On resume, the driver was requesting IRQ just after pci_set_master(),
  but before vortex_up() (which actually resets 3c59x chips).

- Issuing free_irq() on a shared IRQ doesn't guarantee that a buggy
  HW won't trigger spurious interrupts in another driver that
  requested the same interrupt. So, if we want to protect from
  unexpected interrupts, then on suspend we should issue disable_irq(),
  not free_irq().

Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

tunnel: eliminate recursion field

It seems recursion field from "struct ip_tunnel" is not anymore needed.
recursion prevention is done at the upper level (in dev_queue_xmit()),
since we use HARD_TX_LOCK protection for tunnels.

This avoids a cache line ping pong on "struct ip_tunnel" : This structure
should be now mostly read on xmit and receive paths.

Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ems_pci: fix size of CAN controllers BAR mapping for CPC-PCI v2

The driver mapped only 128 bytes of the CAN controller address space when a
CPC-PCI v2 was detected (incl. CPC-104P). This patch will fix it by always
mapping the whole address space (4096 bytes on all boards) of the
corresponding PCI BAR.

Signed-off-by: Sebastian Haas <haas@ems-wuensche.com>
Signed-off-by: Wolfgang Grandegger <wg@grandegger.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net: fix htmldocs sunrpc, clnt.c

DOCPROC Documentation/DocBook/networking.xml
  Warning(net/sunrpc/clnt.c:647): No description found for parameter 'req'
  Warning(net/sunrpc/clnt.c:647): No description found for parameter 'tk_ops'
  Warning(net/sunrpc/clnt.c:647): Excess function parameter 'ops' description in 'rpc_run_bc_task'

Signed-off-by: Jaswinder Singh Rajput <jaswinderrajput@gmail.com>
Cc: Ricardo Labiaga <Ricardo.Labiaga@netapp.com>
Cc: Benny Halevy <bhalevy@panasas.com>
Cc: Andy Adamson <andros@netapp.com>
Cc: Trond Myklebust <Trond.Myklebust@netapp.com>
Cc: Randy Dunlap <randy.dunlap@oracle.com>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: David S. Miller <davem@davemloft.net>

Phonet: error on broadcast sending (unimplemented)

If we ever implement this, then we can stop returning an error.

Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

Phonet: fix race for port number in concurrent bind()

Allocating a port number to a socket and hashing that socket shall be
an atomic operation with regards to other port allocation. Otherwise,
we could allocate a port that is already being allocated to another
socket.

Signed-off-by: Rémi Denis-Courmont <remi.denis-courmont@nokia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pktgen: better scheduler friendliness

Previous update did not resched in inner loop causing watchdogs.
Rewrite inner loop to:
  * account for delays better with less clock calls
  * more accurate timing of delay:
    - only delay if packet was successfully sent
    - if delay is 100ns and it takes 10ns to build packet then
      account for that
  * use wait_event_interruptible_timeout rather than open coding it.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

pktgen: T_TERMINATE flag is unused

Get rid of unused flag bit.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

ipv4: check optlen for IP_MULTICAST_IF option

Due to man page of setsockopt, if optlen is not valid, kernel should return
-EINVAL. But a simple testcase as following, errno is 0, which means setsockopt
is successful.
addr.s_addr = inet_addr("192.1.2.3");
setsockopt(s, IPPROTO_IP, IP_MULTICAST_IF, &addr, 1);
printf("errno is %d\n", errno);

Xiaotian Feng(dfeng@redhat.com) caught the bug. We fix it firstly checking
the availability of optlen and then dealing with the logic like other options.

Reported-by: Xiaotian Feng <dfeng@redhat.com>
Signed-off-by: Shan Wei <shanwei@cn.fujitsu.com>
Acked-by: Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>

Merge branch 'master' of /home/davem/src/GIT/linux-2.6/

Conflicts:
drivers/staging/Kconfig
drivers/staging/Makefile
drivers/staging/cpc-usb/TODO
drivers/staging/cpc-usb/cpc-usb_drv.c
drivers/staging/cpc-usb/cpc.h
drivers/staging/cpc-usb/cpc_int.h
drivers/staging/cpc-usb/cpcusb.h

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: (57 commits)
  drm/i915: Handle ERESTARTSYS during page fault
  drm/i915: Warn before mmaping a purgeable buffer.
  drm/i915: Track purged state.
  drm/i915: Remove eviction debug spam
  drm/i915: Immediately discard any backing storage for uneeded objects
  drm/i915: Do not mis-classify clean objects as purgeable
  drm/i915: Whitespace correction for madv
  drm/i915: BUG_ON page refleak during unbind
  drm/i915: Search harder for a reusable object
  drm/i915: Clean up evict from list.
  drm/i915: Add tracepoints
  drm/i915: framebuffer compression for GM45+
  drm/i915: split display functions by chip type
  drm/i915: Skip the sanity checks if the current relocation is valid
  drm/i915: Check that the relocation points to within the target
  drm/i915: correct FBC update when pipe base update occurs
  drm/i915: blacklist Acer AspireOne lid status
  ACPI: make ACPI button funcs no-ops if not built in
  drm/i915: prevent FIFO calculation overflows on 32 bits with high dotclocks
  drm/i915: intel_display.c handle latency variable efficiently
  ...

Fix up trivial conflicts in drivers/gpu/drm/i915/{i915_dma.c|i915_drv.h}

Merge branch 'linux-next' of git://git./linux/kernel/git/jbarnes/pci-2.6

* 'linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6: (21 commits)
  x86/PCI: make 32 bit NUMA node array int, not unsigned char
  x86/PCI: default pcibus cpumask to all cpus if it lacks affinity
  MAINTAINTERS: remove hotplug driver entries
  PCI: pciehp: remove slot capabilities definitions
  PCI: pciehp: remove error message definitions
  PCI: pciehp: remove number field
  PCI: pciehp: remove hpc_ops
  PCI: pciehp: remove pci_dev field
  PCI: pciehp: remove crit_sect mutex
  PCI: pciehp: remove slot_bus field
  PCI: pciehp: remove first_slot field
  PCI: pciehp: remove slot_device_offset field
  PCI: pciehp: remove hp_slot field
  PCI: pciehp: remove device field
  PCI: pciehp: remove bus field
  PCI: pciehp: remove slot_num_inc field
  PCI: pciehp: remove num_slots field
  PCI: pciehp: remove slot_list field
  PCI: fix VGA arbiter header file
  PCI: Disable AER with pci=nomsi
  ...

Fixed up trivial conflicts in MAINTAINERS

Merge branch 'cputime' of git://git390.marist.edu/linux-2.6

* 'cputime' of git://git390.marist.edu/pub/scm/linux-2.6:
  [PATCH] Fix idle time field in /proc/uptime

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

* 'for-linus' of git://git.monstr.eu/linux-2.6-microblaze: (24 commits)
  microblaze: Disable heartbeat/enable emaclite in defconfigs
  microblaze: Support simpleImage.dts make target
  microblaze: Fix _start symbol to physical address
  microblaze: Use LOAD_OFFSET macro to get correct LMA for all sections
  microblaze: Create the LOAD_OFFSET macro used to compute VMA vs LMA offsets
  microblaze: Copy ppc asm-compat.h for clean handling of constants in asm and C
  microblaze: Actually show KiB rather than pages in "Freeing initrd memory:"
  microblaze: Support ptrace syscall tracing.
  microblaze: Updated CPU version and FPGA family codes in PVR
  microblaze: Generate correct signal and siginfo for integer div-by-zero
  microblaze: Don't be noisy when userspace causes hardware exceptions
  microblaze: Remove ipc.h file which points to non-existing asm-generic file
  microblaze: Clear sticky FSR register after generating exception signals
  microblaze: Ensure CPU usermode is set on new userspace processes
  microblaze: Use correct kbuild variable KBUILD_CFLAGS
  microblaze: Save and restore msr in hw exception
  microblaze: Add architectural support for USB EHCI host controllers
  microblaze: Implement include/asm/syscall.h.
  microblaze: Improve checking mechanism for MSR instruction
  microblaze: Add checking mechanism for MSR instruction
  ...

Merge git://git./linux/kernel/git/rusty/linux-2.6-for-linus

* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
  module: don't call percpu_modfree on NULL pointer.
  module: fix memory leak when load fails after srcversion/version allocated
  module: preferred way to use MODULE_AUTHOR
  param: allow whitespace as kernel parameter separator
  module: reduce string table for loaded modules (v2)
  module: reduce symbol table for loaded modules (v2)

Merge branch 'for-linus' of git://git./linux/kernel/git/mason/btrfs-unstable

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/btrfs-unstable: (42 commits)
  Btrfs: hash the btree inode during  fill_super
  Btrfs: relocate file extents in clusters
  Btrfs: don't rename file into dummy directory
  Btrfs: check size of inode backref before adding hardlink
  Btrfs: fix releasepage to avoid unlocking extents we haven't locked
  Btrfs: Fix test_range_bit for whole file extents
  Btrfs: fix errors handling cached state in set/clear_extent_bit
  Btrfs: fix early enospc during balancing
  Btrfs: deal with NULL space info
  Btrfs: account for space used by the super mirrors
  Btrfs: fix extent entry threshold calculation
  Btrfs: remove dead code
  Btrfs: fix bitmap size tracking
  Btrfs: don't keep retrying a block group if we fail to allocate a cluster
  Btrfs: make balance code choose more wisely when relocating
  Btrfs: fix arithmetic error in clone ioctl
  Btrfs: add snapshot/subvolume destroy ioctl
  Btrfs: change how subvolumes are organized
  Btrfs: do not reuse objectid of deleted snapshot/subvol
  Btrfs: speed up snapshot dropping
  ...

Merge branch 'for-linus' of git://git./linux/kernel/git/viro/vfs-2.6

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
  truncate: use new helpers
  truncate: new helpers
  fs: fix overflow in sys_mount() for in-kernel calls
  fs: Make unload_nls() NULL pointer safe
  freeze_bdev: grab active reference to frozen superblocks
  freeze_bdev: kill bd_mount_sem
  exofs: remove BKL from super operations
  fs/romfs: correct error-handling code
  vfs: seq_file: add helpers for data filling
  vfs: remove redundant position check in do_sendfile
  vfs: change sb->s_maxbytes to a loff_t
  vfs: explicitly cast s_maxbytes in fiemap_check_ranges
  libfs: return error code on failed attr set
  seq_file: return a negative error code when seq_path_root() fails.
  vfs: optimize touch_time() too
  vfs: optimization for touch_atime()
  vfs: split generic_forget_inode() so that hugetlbfs does not have to copy it
  fs/inode.c: add dev-id and inode number for debugging in init_special_inode()
  libfs: make simple_read_from_buffer conventional

Merge git://git./linux/kernel/git/viro/audit-current

* git://git.kernel.org/pub/scm/linux/kernel/git/viro/audit-current:
  lsm: Use a compressed IPv6 string format in audit events
  Audit: send signal info if selinux is disabled
  Audit: rearrange audit_context to save 16 bytes per struct
  Audit: reorganize struct audit_watch to save 8 bytes

module: don't call percpu_modfree on NULL pointer.

The general one handles NULL, the static obsolescent
(CONFIG_HAVE_LEGACY_PER_CPU_AREA) one in module.c doesn't; Eric's
commit 720eba31 assumed it did, and various frobbings since then kept
that assumption.

All other callers in module.c all protect it with an if; this effectively
does the same as free_init is only goto if we fail percpu_modalloc().

Reported-by: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Eric Dumazet <dada1@cosmosbay.com>
Cc: Masami Hiramatsu <mhiramat@redhat.com>
Cc: Américo Wang <xiyou.wangcong@gmail.com>
Tested-by: Kamalesh Babulal <kamalesh@linux.vnet.ibm.com>

module: fix memory leak when load fails after srcversion/version allocated

Normally the twisty paths of sysfs will free the attributes, but not if
we fail before we hook it into sysfs (which is the last thing we do in
load_module).

(This sysfs code is a turd, no doubt there are other issues lurking too).

Reported-by: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Tested-by: Tetsuo Handa <penguin-kernel@i-love.sakura.ne.jp>

module: preferred way to use MODULE_AUTHOR

For the longest time now we've been using multiple MODULE_AUTHOR()
statements when a module has more than one author, but the comment here
disagrees.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: Luciano Coelho <luciano.coelho@nokia.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

param: allow whitespace as kernel parameter separator

Some boot mechanisms require that kernel parameters are stored in a
separate file which is loaded to memory without further processing
(e.g. the "Load from FTP" method on s390). When such a file contains
newline characters, the kernel parameter preceding the newline might
not be correctly parsed (due to the newline being stuck to the end of
the actual parameter value) which can lead to boot failures.

This patch improves kernel command line usability in such a situation
by allowing generic whitespace characters as separators between kernel
parameters.

Signed-off-by: Peter Oberparleiter <oberpar@linux.vnet.ibm.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

module: reduce string table for loaded modules (v2)

Also remove all parts of the string table (referenced by the symbol
table) that are not needed for kallsyms use (i.e. which were only
referenced by symbols discarded by the previous patch, or not
referenced at all for whatever reason).

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

module: reduce symbol table for loaded modules (v2)

Discard all symbols not interesting for kallsyms use: absolute,
section, and in the common case (!KALLSYMS_ALL) data ones.

Signed-off-by: Jan Beulich <jbeulich@novell.com>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

Merge branch 'for-linus' of git://neil.brown.name/md

* 'for-linus' of git://neil.brown.name/md: (97 commits)
  md: raid-1/10: fix RW bits manipulation
  md: remove unnecessary memset from multipath.
  md: report device as congested when suspended
  md: Improve name of threads created by md_register_thread
  md: remove sparse warnings about lock context.
  md: remove sparse waring "symbol xxx shadows an earlier one"
  async_tx/raid6: add missing dma_unmap calls to the async fail case
  ioat3: fix uninitialized var warnings
  drivers/dma/ioat/dma_v2.c: fix warnings
  raid6test: fix stack overflow
  ioat2: clarify ring size limits
  md/raid6: cleanup ops_run_compute6_2
  md/raid6: eliminate BUG_ON with side effect
  dca: module load should not be an error message
  ioat: driver version 4.0
  dca: registering requesters in multiple dca domains
  async_tx: remove HIGHMEM64G restriction
  dmaengine: sh: Add Support SuperH DMA Engine driver
  dmaengine: Move all map_sg/unmap_sg for slave channel to its client
  fsldma: Add DMA_SLAVE support
  ...

Merge branch 'for_linus' of git://git./linux/kernel/git/mchehab/linux-2.6

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mchehab/linux-2.6:
  V4L/DVB (13039): dib0700: not building CONFIG_DVB_TUNER_DIB0070 breaks compilation
  V4L/DVB (13038): dvbdev: Remove an anoying/uneeded warning
  V4L/DVB (13037): go7007: Revert compatibility code added at the wrong place
  media: video: Fix build in saa7164

Merge branch 'hwpoison' of git://git./linux/kernel/git/ak/linux-mce-2.6

* 'hwpoison' of git://git.kernel.org/pub/scm/linux/kernel/git/ak/linux-mce-2.6: (21 commits)
  HWPOISON: Enable error_remove_page on btrfs
  HWPOISON: Add simple debugfs interface to inject hwpoison on arbitary PFNs
  HWPOISON: Add madvise() based injector for hardware poisoned pages v4
  HWPOISON: Enable error_remove_page for NFS
  HWPOISON: Enable .remove_error_page for migration aware file systems
  HWPOISON: The high level memory error handler in the VM v7
  HWPOISON: Add PR_MCE_KILL prctl to control early kill behaviour per process
  HWPOISON: shmem: call set_page_dirty() with locked page
  HWPOISON: Define a new error_remove_page address space op for async truncation
  HWPOISON: Add invalidate_inode_page
  HWPOISON: Refactor truncate to allow direct truncating of page v2
  HWPOISON: check and isolate corrupted free pages v2
  HWPOISON: Handle hardware poisoned pages in try_to_unmap
  HWPOISON: Use bitmask/action code for try_to_unmap behaviour
  HWPOISON: x86: Add VM_FAULT_HWPOISON handling to x86 page fault handler v2
  HWPOISON: Add poison check to page fault handling
  HWPOISON: Add basic support for poisoned pages in fault handler v3
  HWPOISON: Add new SIGBUS error codes for hardware poison signals
  HWPOISON: Add support for poison swap entries v2
  HWPOISON: Export some rmap vma locking to outside world
  ...

drivers/usb/serial/sierra.c: fix CONFIG_PM=n build

drivers/usb/serial/sierra.c: In function 'sierra_suspend':
drivers/usb/serial/sierra.c:936: error: 'struct usb_device' has no member named 'auto_pm'

Repairs

commit e6929a9020acbeb04d9a3ad9a88234c15be808fd
Author: Oliver Neukum <oliver@neukum.org>
Date:   Fri Sep 4 23:19:53 2009 +0200

    USB: support for autosuspend in sierra while online

Cc: Greg KH <greg@kroah.com>
Cc: Oliver Neukum <oliver@neukum.org>
Cc: Elina Pasheva <epasheva@sierrawireless.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

alpha: AGP update (fixes compile failure)

This brings Alpha AGP platforms in sync with the change to struct
agp_memory (unsigned long *memory => struct page **pages).

Only compile tested (I don't have titan/marvel hardware), but this change
looks pretty straightforward, so hopefully it's ok.

Signed-off-by: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Dave Airlie <airlied@linux.ie>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

task_struct cleanup: move binfmt field to mm_struct

Because the binfmt is not different between threads in the same process,
it can be moved from task_struct to mm_struct.  And binfmt moudle is
handled per mm_struct instead of task_struct.

Signed-off-by: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Cc: Rusty Russell <rusty@rustcorp.com.au>
Acked-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

include/linux/unaligned/{l,b}e_byteshift.h: fix usage for compressed kernels

When unaligned accesses are required for uncompressing a kernel (such as
for LZO decompression on ARM in a patch that follows), including
<linux/kernel.h> causes issues as it brings in a lot of things that are
not available in the decompression environment.

linux/kernel.h brings at least:
extern int console_printk[];
extern const char hex_asc[];
which causes errors at link-time as they are not available when
compiling the pre-boot environement. There are also a few others:

  arch/arm/boot/compressed/misc.o: In function `valid_user_regs':
   arch/arm/include/asm/ptrace.h:158: undefined reference to `elf_hwcap'
  arch/arm/boot/compressed/misc.o: In function `console_silent':
   include/linux/kernel.h:292: undefined reference to `console_printk'
  arch/arm/boot/compressed/misc.o: In function `console_verbose':
   include/linux/kernel.h:297: undefined reference to `console_printk'
  arch/arm/boot/compressed/misc.o: In function `pack_hex_byte':
   include/linux/kernel.h:360: undefined reference to `hex_asc'
  arch/arm/boot/compressed/misc.o: In function `hweight_long':
   include/linux/bitops.h:45: undefined reference to `hweight32'
  arch/arm/boot/compressed/misc.o: In function `__cmpxchg_local_generic':
   include/asm-generic/cmpxchg-local.h:21: undefined reference to `wrong_size_cmpxchg'
   include/asm-generic/cmpxchg-local.h:42: undefined reference to `wrong_size_cmpxchg'
  arch/arm/boot/compressed/misc.o: In function `__xchg':
   arch/arm/include/asm/system.h:309: undefined reference to `__bad_xchg'

However, those files apparently use nothing from <linux/kernel.h>, all
they need is the declaration of types such as u32 or u64, so
<linux/types.h> should be enough

Signed-off-by: Albin Tonnerre <albin.tonnerre@free-electrons.com>
Cc: Sam Ravnborg <sam@ravnborg.org>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Phillip Lougher <phillip@lougher.demon.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

lzma/gzip: fix potential oops when input data is truncated

If the lzma/gzip decompressors are called with insufficient input data
(len > 0 & fill = NULL), they will attempt to call the fill function to
obtain more data, leading to a kernel oops.

Signed-off-by: Phillip Lougher <phillip@lougher.demon.co.uk>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/vlynq/vlynq.c: fix resource size off by 1 error

In this case, the calls to request_mem_region, ioremap, and
release_mem_region all have a consistent length argument, len, but since
in other files (res->end - res->start) + 1, equivalent to
resource_size(res), is used for a resource-typed structure res, one could
consider whether the same should be done here.

The problem was found using the following semantic patch:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@
struct resource *res;
@@

- (res->end - res->start) + 1
+ resource_size(res)

@@
struct resource *res;
@@

- res->end - res->start
+ BAD(resource_size(res))
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/romfs: correct error-handling code

romfs_iget returns an ERR_PTR value in an error case instead of NULL.

A simplified version of the semantic match that finds this problem is as
follows: (http://coccinelle.lip6.fr/)

// <smpl>
@match exists@
expression x, E;
statement S1, S2;
@@

x = romfs_iget(...)
... when != x = E
(
*  if (x == NULL || ...) S1 else S2
|
*  if (x == NULL && ...) S1 else S2
)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gru: allocation may fail in quicktest1()

The allocation may fail.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Acked-by: Jack Steiner <steiner@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

gru: use proc_create()

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Jack Steiner <steiner@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

aio: ifdef fields in mm_struct

->ioctx_lock and ->ioctx_list are used only under CONFIG_AIO.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Zach Brown <zach.brown@oracle.com>
Cc: Benjamin LaHaise <bcrl@kvack.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memstick: move dev_dbg

id_reg.if_mode might be unitialized when (*mrq)->error is nonzero.  move
dev_dbg() inside the if so that we are sure we can use id_reg values.

Signed-off-by: Jiri Slaby <jirislaby@gmail.com>
Cc: Alex Dubov <oakad@yahoo.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

adfs: remove redundant test on unsigned

unsigned block cannot be less than 0.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

edac: core: remove completion-wait for complete with rcu_barrier

Module edac_core.ko uses call_rcu() callbacks in edac_device.c, edac_mc.c
and edac_pci.c.

They all use a wait_for_completion() scheme, but this scheme it not 100%
safe on multiple CPUs.  See the _rcu_barrier() implementation which
explains why extra precausion is needed.

The patch adds a comment about rcu_barrier() and as a precausion calls
rcu_barrier().  A maintainer needs to look at removing the
wait_for_completion code.

[dougthompson@xmission.com: remove the wait_for_completion code]
Signed-off-by Jesper Dangaard Brouer <hawk@comx.dk>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

edac: i3200 memory controller driver

A driver for the Intel 3200 and 3210 memory controllers.  It has only had
light testing so far, and currently makes no attempt to decode error
addresses at anything finer than csrow granularity.

Signed-off-by: Jason Uhlenkott <juhlenko@akamai.com>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

edac: fix resource size calculation

Use the function resource_size, which reduces the chance of introducing
off-by-one errors in calculating the resource size.

The semantic patch that makes this change is as follows:
(http://www.emn.fr/x-info/coccinelle/)

// <smpl>
@@
struct resource *res;
@@

- (res->end - res->start) + 1
+ resource_size(res)
// </smpl>

Signed-off-by: Julia Lawall <julia@diku.dk>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

edac: mpc85xx add mpc83xx support

Add support for the Freescale MPC83xx memory controller to the existing
driver for the Freescale MPC85xx memory controller.  The only difference
between the two processors are in the CS_BNDS register parsing code, which
has been changed so it will work on both processors.

The L2 cache controller does not exist on the MPC83xx, but the OF
subsystem will not use the driver if the device is not present in the OF
device tree.

I had to change the nr_pages calculation to make the math work out.  I
checked it on my board and did the math by hand for a 64GB 85xx using 64K
pages.  In both cases, nr_pages * PAGE_SIZE comes out to the correct
value.

Signed-off-by: Ira W. Snyder <iws@ovro.caltech.edu>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Cc: Kumar Gala <galak@gate.crashing.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

edac: mpc85xx add P2020DS support

Based on Kumar's new compatible types patch, add P2020 into MPC85xx EDAC
compatible lists so that EDAC can recognize P2020 meomry controller and L2
cache controller and export the relevant fields to sysfs.

EDAC MPC85xx DDR3 support is needed if DDR3 memory stick is installed on a
P2020DS board so that EDAC core can recognize DDR3 memory type.

Signed-off-by: Yang Shi <yang.shi@windriver.com>
Acked-by: Dave Jiang <djiang@mvista.com>
Signed-off-by: Doug Thompson <dougthompson@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

linux/futex.h: place kernel types behind __KERNEL__

The forward decls for some kernel types are only needed by the code behind
__KERNEL__, so don't bleed these types to userspace.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

pidns: deny CLONE_PARENT|CLONE_NEWPID combination

CLONE_PARENT was used to implement an older threading model.  For
consistency with the CLONE_THREAD check in copy_pid_ns(), disable
CLONE_PARENT with CLONE_NEWPID, at least until the required semantics of
pid namespaces are clear.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Acked-by: Roland McGrath <roland@redhat.com>
Acked-by: Serge Hallyn <serue@us.ibm.com>
Cc: Oren Laadan <orenl@cs.columbia.edu>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fork(): disable CLONE_PARENT for init

When global or container-init processes use CLONE_PARENT, they create a
multi-rooted process tree.  Besides siblings of global init remain as
zombies on exit since they are not reaped by their parent (swapper).  So
prevent global and container-inits from creating siblings.

Signed-off-by: Sukadev Bhattiprolu <sukadev@us.ibm.com>
Acked-by: Eric W. Biederman <ebiederm@xmission.com>
Acked-by: Roland McGrath <roland@redhat.com>
Cc: Oren Laadan <orenl@cs.columbia.edu>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Serge Hallyn <serue@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

sysctl: remove "struct file *" argument of ->proc_handler

It's unused.

It isn't needed -- read or write flag is already passed and sysctl
shouldn't care about the rest.

It _was_ used in two places at arch/frv for some reason.

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Cc: David Howells <dhowells@redhat.com>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Ralf Baechle <ralf@linux-mips.org>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: James Morris <jmorris@namei.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

MAINTAINERS: add Matt Mackall and Herbert Xu to HARDWARE RANDOM NUMBER GENERATOR

Signed-off-by: Joe Perches <joe@perches.com>
Cc: David Daney <ddaney@caviumnetworks.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Matt Mackall <mpm@selenic.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

bfin-otp: add writing support

The on-chip OTP may be written at runtime, so enable support for it in the
driver.  However, since writing should really be done only on development
systems, don't bend over backwards to make sure the simple software lock
is per-fd -- per-device is OK.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Signed-off-by: Bryan Wu <cooloney@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/char/uv_mmtimer.c: add memory mapped RTC driver for UV

This driver memory maps the UV Hub RTC.

Signed-off-by: Dimitri Sivanich <sivanich@sgi.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

drivers/char/rio/rioctrl.c: off by one error in rioctrl.c

If DownLoad.ProductCode == MAX_PRODUCT, that would be a problem when we do
RIOBootTable[DownLoad.ProductCode] a couple lines down.

Found by smatch (http://repo.or.cz/w/smatch.git).

Signed-off-by: Dan Carpenter <error27@gmail.com>
Cc: Jiri Slaby <jirislaby@gmail.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

hpet: hpet driver periodic timer setup bug fixes

The periodic interrupt from drivers/char/hpet.c does not work correctly,
both when using the periodic capability of the hardware and while
emulating the periodic interrupt (when hardware does not support periodic
mode).

With timers capable of periodic interrupts, the comparator field is first
set with the period value followed by set of hidden accumulator, which has
the side effect of overwriting the comparator value.  This results in
wrong periodicity for the interrupts.  For, periodic interrupts to work,
following steps are necessary, in that order.

* Set config with Tn_VAL_SET_CNF bit

* Write to hidden accumulator, the value written is the time when the
  first interrupt should be generated

* Write compartor with period interval for subsequent interrupts
  (http://www.intel.com/hardwaredesign/hpetspec_1.pdf )

When emulating periodic timer with timers not capable of periodic
interrupt, driver is adding the period to counter value instead of
comparator value, which causes slow drift when using this emulation.

Also, driver seems to add hpetp->hp_delta both while setting up periodic
interrupt and while emulating periodic interrupts with timers not capable
of doing periodic interrupts.  This hp_delta will result in slower than
expected interrupt rate and should not be used while setting the interval.

Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com>
Signed-off-by: Nils Carlson <nils.carlson@ericsson.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

mwave: fix read buffer overflow

Check whether index is within bounds before grabbing the element.

Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Cc: Kay Sievers <kay.sievers@vrfy.org>
Cc: Greg Kroah-Hartman <gregkh@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fs/char_dev.c: remove useless loop

There are two useless lines in fs/char_dev.c.

In register_chrdev there is a loop to change all '/' into '!' in the
kernel object name.
This code is useless as the same substitution is in kobject_set_name_vargs in
lib/kobject.c:
228         /* ewww... some of these buggers have '/' in the name ... */
229         while ((s = strchr(kobj->name, '/')))
230                 s[0] = '!';

kobject_set_name_vargs is called by kobject_set_name.
kobject_set_name is called just above the useless loop.

[hidave.darkstar@gmail.com: fix warning, remove the unused char *s]
Signed-off-by: Renzo Davoli <renzo@cs.unibo.it>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Dave Young <hidave.darkstar@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

/dev/zero: avoid repeated access_ok() checks

In read_zero, we check for access_ok() once for the count bytes.  It is
unnecessarily checked again in clear_user.  Use __clear_user, which does
not check for access_ok().

Signed-off-by: Nikanth Karthikesan <knikanth@suse.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

flat: use IS_ERR_VALUE() helper macro

There is a common macro now for testing mixed pointer/errno values, so use
that rather than handling the casts ourself.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Acked-by: David McCullough <david_mccullough@securecomputing.com>
Acked-by: Greg Ungerer <gerg@uclinux.org>
Cc: David Howells <dhowells@redhat.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fdpic: ignore the loader's PT_GNU_STACK when calculating the stack size

Ignore the loader's PT_GNU_STACK when calculating the stack size, and only
consider the executable's PT_GNU_STACK, assuming the executable has one.

Currently the behaviour is to take the largest stack size and use that,
but that means you can't reduce the stack size in the executable.  The
loader's stack size should probably only be used when executing the loader
directly.

WARNING: This patch is slightly dangerous - it may render a system
inoperable if the loader's stack size is larger than that of important
executables, and the system relies unknowingly on this increasing the size
of the stack.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
Acked-by: Paul Mundt <lethal@linux-sh.org>
Cc: Pavel Machek <pavel@ucw.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

elf: clean up fill_note_info()

Introduce a helper function elf_note_info_init() to help fill_note_info()
to do initializations, also fix the potential memory leaks.

[akpm@linux-foundation.org: remove NUM_NOTES]
Signed-off-by: WANG Cong <amwang@redhat.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: David Howells <dhowells@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

signals: inline __fatal_signal_pending

__fatal_signal_pending inlines to one instruction on x86, probably two
instructions on other machines.  It takes two longer x86 instructions just
to call it and test its return value, not to mention the function itself.

On my random x86_64 config, this saved 70 bytes of text (59 of those being
__fatal_signal_pending itself).

Signed-off-by: Roland McGrath <roland@redhat.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

fcntl: add F_[SG]ETOWN_EX

In order to direct the SIGIO signal to a particular thread of a
multi-threaded application we cannot, like suggested by the manpage, put a
TID into the regular fcntl(F_SETOWN) call.  It will still be send to the
whole process of which that thread is part.

Since people do want to properly direct SIGIO we introduce F_SETOWN_EX.

The need to direct SIGIO comes from self-monitoring profiling such as with
perf-counters.  Perf-counters uses SIGIO to notify that new sample data is
available.  If the signal is delivered to the same task that generated the
new sample it can augment that data by inspecting the task's user-space
state right after it returns from the kernel.  This is esp.  convenient
for interpreted or virtual machine driven environments.

Both F_SETOWN_EX and F_GETOWN_EX take a pointer to a struct f_owner_ex
as argument:

struct f_owner_ex {
int   type;
pid_t pid;
};

Where type is one of F_OWNER_TID, F_OWNER_PID or F_OWNER_GID.

Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Tested-by: stephane eranian <eranian@googlemail.com>
Cc: Michael Kerrisk <mtk.manpages@googlemail.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

signals: send_sigio: use do_send_sig_info() to avoid check_kill_permission()

group_send_sig_info()->check_kill_permission() assumes that current is the
sender and uses current_cred().

This is not true in send_sigio_to_task() case.  From the security pov the
sender is not current, but the task which did fcntl(F_SETOWN), that is why
we have sigio_perm() which uses the right creds to check.

Fortunately, send_sigio() always sends either SEND_SIG_PRIV or
SI_FROMKERNEL() signal, so check_kill_permission() does nothing.  But
still it would be tidier to avoid this bogus security check and save a
couple of cycles.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: stephane eranian <eranian@googlemail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

signals: introduce do_send_sig_info() helper

Introduce do_send_sig_info() and convert group_send_sig_info(),
send_sig_info(), do_send_specific() to use this helper.

Hopefully it will have more users soon, it allows to specify
specific/group behaviour via "bool group" argument.

Shaves 80 bytes from .text.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: stephane eranian <eranian@googlemail.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

exec: fix set_binfmt() vs sys_delete_module() race

sys_delete_module() can set MODULE_STATE_GOING after
search_binary_handler() does try_module_get().  In this case
set_binfmt()->try_module_get() fails but since none of the callers
check the returned error, the task will run with the wrong old
->binfmt.

The proper fix should change all ->load_binary() methods, but we can
rely on fact that the caller must hold a reference to binfmt->module
and use __module_get() which never fails.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: Hiroshi Shimamoto <h-shimamoto@ct.jp.nec.com>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

exec: allow do_coredump() to wait for user space pipe readers to complete

Allow core_pattern pipes to wait for user space to complete

One of the things that user space processes like to do is look at metadata
for a crashing process in their /proc/<pid> directory.  this is racy
however, since do_coredump in the kernel doesn't wait for the user space
process to complete before it reaps the crashing process.  This patch
corrects that.  Allowing the kernel to wait for the user space process to
complete before cleaning up the crashing process.  This is a bit tricky to
do for a few reasons:

1) The user space process isn't our child, so we can't sys_wait4 on it
2) We need to close the pipe before waiting for the user process to complete,
since the user process may rely on an EOF condition

I've discussed several solutions with Oleg Nesterov off-list about this,
and this is the one we've come up with.  We add ourselves as a pipe reader
(to prevent premature cleanup of the pipe_inode_info), and remove
ourselves as a writer (to provide an EOF condition to the writer in user
space), then we iterate until the user space process exits (which we
detect by pipe->readers == 1, hence the > 1 check in the loop).  When we
exit the loop, we restore the proper reader/writer values, then we return
and let filp_close in do_coredump clean up the pipe data properly.

Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Earl Chew <earl_chew@agilent.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

exec: let do_coredump() limit the number of concurrent dumps to pipes

Introduce core pipe limiting sysctl.

Since we can dump cores to pipe, rather than directly to the filesystem,
we create a condition in which a user can create a very high load on the
system simply by running bad applications.

If the pipe reader specified in core_pattern is poorly written, we can
have lots of ourstandig resources and processes in the system.

This sysctl introduces an ability to limit that resource consumption.
core_pipe_limit defines how many in-flight dumps may be run in parallel,
dumps beyond this value are skipped and a note is made in the kernel log.
A special value of 0 in core_pipe_limit denotes unlimited core dumps may
be handled (this is the default value).

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Earl Chew <earl_chew@agilent.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

exec: make do_coredump() more resilient to recursive crashes

Change how we detect recursive dumps.

Currently we have a mechanism by which we try to compare pathnames of the
crashing process to the core_pattern path.  This is broken for a dozen
reasons, and just doesn't work in any sort of robust way.

I'm replacing it with the use of a 0 RLIMIT_CORE value.  Since helper apps
set RLIMIT_CORE to zero, we don't write out core files for any process
with that particular limit set.  It the core_pattern is a pipe, any
non-zero limit is translated to RLIM_INFINITY.

This allows complete dumps to be captured, but prevents infinite recursion
in the event that the core_pattern process itself crashes.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Earl Chew <earl_chew@agilent.com>
Cc: Oleg Nesterov <oleg@tv-sign.ru>
Cc: Andi Kleen <andi@firstfloor.org>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

signals: tracehook_notify_jctl change

This changes tracehook_notify_jctl() so it's called with the siglock held,
and changes its argument and return value definition.  These clean-ups
make it a better fit for what new tracing hooks need to check.

Tracing needs the siglock here, held from the time TASK_STOPPED was set,
to avoid potential SIGCONT races if it wants to allow any blocking in its
tracing hooks.

This also folds the finish_stop() function into its caller
do_signal_stop().  The function is short, called only once and only
unconditionally.  It aids readability to fold it in.

[oleg@redhat.com: do not call tracehook_notify_jctl() in TASK_STOPPED state]
[oleg@redhat.com: introduce tracehook_finish_jctl() helper]
Signed-off-by: Roland McGrath <roland@redhat.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

wait_noreap_copyout(): check for ->wo_info != NULL

Current behaviour of sys_waitid() looks odd.  If user passes infop ==
NULL, sys_waitid() returns success.  When user additionally specifies flag
WNOWAIT, sys_waitid() returns -EFAULT on the same conditions.  When user
combines WNOWAIT with WCONTINUED, sys_waitid() again returns success.

This patch adds check for ->wo_info in wait_noreap_copyout().

User-visible change: starting from this commit, sys_waitid() always checks
infop != NULL and does not fail if it is NULL.

Signed-off-by: Vitaly Mayatskikh <v.mayatskih@gmail.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

do_wait: fix sys_waitid()-specific behaviour

do_wait() checks ->wo_info to figure out who is the caller.  If it's not
NULL the caller should be sys_waitid(), in that case do_wait() fixes up
the retval or zeros ->wo_info, depending on retval from underlying
function.

This is bug: user can pass ->wo_info == NULL and sys_waitid() will return
incorrect value.

man 2 waitid says:

waitid(): returns 0 on success

Test-case:

int main(void)
{
if (fork())
assert(waitid(P_ALL, 0, NULL, WEXITED) == 0);

return 0;
}

Result:

Assertion `waitid(P_ALL, 0, ((void *)0), 4) == 0' failed.

Move that code to sys_waitid().

User-visible change: sys_waitid() will return 0 on success, either
infop is set or not.

Note, there's another bug in wait_noreap_copyout() which affects
return value of sys_waitid(). It will be fixed in next patch.

Signed-off-by: Vitaly Mayatskikh <v.mayatskih@gmail.com>
Reviewed-by: Oleg Nesterov <oleg@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

wait_consider_task: kill "parent" argument

Kill the unused "parent" argument in wait_consider_task(), it was never used.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Ratan Nalumasu <rnalumasu@gmail.com>
Cc: Vitaly Mayatskikh <vmayatsk@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

do_wait-wakeup-optimization: simplify task_pid_type()

task_pid_type() is only used by eligible_pid() which has to check wo_type
!= PIDTYPE_MAX anyway.  Remove this check from task_pid_type() and factor
out ->pids[type] access, this shrinks .text a bit and simplifies the code.

The matches the behaviour of other similar helpers, say get_task_pid().
The caller must ensure that pid_type is valid, not the callee.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

do_wait-wakeup-optimization: fix child_wait_callback()->eligible_child() usage

child_wait_callback()->eligible_child() is not right, we can miss the
wakeup if the task was detached before __wake_up_parent() and the caller
of do_wait() didn't use __WALL.

Move ->wo_pid checks from eligible_child() to the new helper,
eligible_pid(), and change child_wait_callback() to use it instead of
eligible_child().

Note: actually I think it would be better to fix the __WCLONE check in
eligible_child(), it doesn't look exactly right.  But it is not clear what
is the supposed behaviour, and any change is user-visible.

Reported-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Tested-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

do_wait() wakeup optimization: child_wait_callback: check __WNOTHREAD case

Suggested by Roland.

do_wait(__WNOTHREAD) can only succeed if the caller is either ptracer, or
it is ->real_parent and the child is not traced. IOW, caller == p->parent
otherwise we should not wake up.

Change child_wait_callback() to check this. Ratan reports the workload with
CPU load >99% caused by unnecessary wakeups, should be fixed by this patch.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Ratan Nalumasu <rnalumasu@gmail.com>
Cc: Vitaly Mayatskikh <vmayatsk@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

do_wait() wakeup optimization: change __wake_up_parent() to use filtered wakeup

Ratan Nalumasu reported that in a process with many threads doing
unnecessary wakeups.  Every waiting thread in the process wakes up to loop
through the children and see that the only ones it cares about are still
not ready.

Now that we have struct wait_opts we can change do_wait/__wake_up_parent
to use filtered wakeups.

We can make child_wait_callback() more clever later, right now it only
checks eligible_child().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Ratan Nalumasu <rnalumasu@gmail.com>
Cc: Vitaly Mayatskikh <vmayatsk@redhat.com>
Acked-by: James Morris <jmorris@namei.org>
Tested-by: Valdis Kletnieks <valdis.kletnieks@vt.edu>
Acked-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

do_wait() wakeup optimization: shift security_task_wait() from eligible_child() to wait_consider_task()

Preparation, no functional changes.

eligible_child() has a single caller, wait_consider_task(). We can move
security_task_wait() out from eligible_child(), this allows us to use it
for filtered wake_up().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Roland McGrath <roland@redhat.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Ratan Nalumasu <rnalumasu@gmail.com>
Cc: Vitaly Mayatskikh <vmayatsk@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

ptrace: __ptrace_detach: do __wake_up_parent() if we reap the tracee

The bug is old, it wasn't cause by recent changes.

Test case:

static void *tfunc(void *arg)
{
int pid = (long)arg;

assert(ptrace(PTRACE_ATTACH, pid, NULL, NULL) == 0);
kill(pid, SIGKILL);

sleep(1);
return NULL;
}

int main(void)
{
pthread_t th;
long pid = fork();

if (!pid)
pause();

signal(SIGCHLD, SIG_IGN);
assert(pthread_create(&th, NULL, tfunc, (void*)pid) == 0);

int r = waitpid(-1, NULL, __WNOTHREAD);
printf("waitpid: %d %m\n", r);

return 0;
}

Before the patch this program hangs, after this patch waitpid() correctly
fails with errno == -ECHILD.

The problem is, __ptrace_detach() reaps the EXIT_ZOMBIE tracee if its
->real_parent is our sub-thread and we ignore SIGCHLD.  But in this case
we should wake up other threads which can sleep in do_wait().

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: Roland McGrath <roland@redhat.com>
Cc: Vitaly Mayatskikh <vmayatsk@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: show swap usage in stat file

We now count MEM_CGROUP_STAT_SWAPOUT, so we can show swap usage.  It would
be useful for users to show swap usage in memory.stat file, because they
don't need calculate memsw.usage - res.usage to know swap usage.

Signed-off-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

memcg: improve resource counter scalability

Reduce the resource counter overhead (mostly spinlock) associated with the
root cgroup.  This is a part of the several patches to reduce mem cgroup
overhead.  I had posted other approaches earlier (including using percpu
counters).  Those patches will be a natural addition and will be added
iteratively on top of these.

The patch stops resource counter accounting for the root cgroup.  The data
for display is derived from the statisitcs we maintain via
mem_cgroup_charge_statistics (which is more scalable).  What happens today
is that, we do double accounting, once using res_counter_charge() and once
using memory_cgroup_charge_statistics().  For the root, since we don't
implement limits any more, we don't need to track every charge via
res_counter_charge() and check for limit being exceeded and reclaim.

The main mem->res usage_in_bytes can be derived by summing the cache and
rss usage data from memory statistics (MEM_CGROUP_STAT_RSS and
MEM_CGROUP_STAT_CACHE).  However, for memsw->res usage_in_bytes, we need
additional data about swapped out memory.  This patch adds a
MEM_CGROUP_STAT_SWAPOUT and uses that along with MEM_CGROUP_STAT_RSS and
MEM_CGROUP_STAT_CACHE to derive the memsw data.  This data is computed
recursively when hierarchy is enabled.

The tests results I see on a 24 way show that

1. The lock contention disappears from /proc/lock_stats
2. The results of the test are comparable to running with
   cgroup_disable=memory.

Here is a sample of my program runs

Without Patch

 Performance counter stats for '/home/balbir/parallel_pagefault':

 7192804.124144  task-clock-msecs         #     23.937 CPUs
         424691  context-switches         #      0.000 M/sec
            267  CPU-migrations           #      0.000 M/sec
       28498113  page-faults              #      0.004 M/sec
  5826093739340  cycles                   #    809.989 M/sec
   408883496292  instructions             #      0.070 IPC
     7057079452  cache-references         #      0.981 M/sec
     3036086243  cache-misses             #      0.422 M/sec

  300.485365680  seconds time elapsed

With cgroup_disable=memory

 Performance counter stats for '/home/balbir/parallel_pagefault':

 7182183.546587  task-clock-msecs         #     23.915 CPUs
         425458  context-switches         #      0.000 M/sec
            203  CPU-migrations           #      0.000 M/sec
       92545093  page-faults              #      0.013 M/sec
  6034363609986  cycles                   #    840.185 M/sec
   437204346785  instructions             #      0.072 IPC
     6636073192  cache-references         #      0.924 M/sec
     2358117732  cache-misses             #      0.328 M/sec

  300.320905827  seconds time elapsed

With this patch applied

 Performance counter stats for '/home/balbir/parallel_pagefault':

 7191619.223977  task-clock-msecs         #     23.955 CPUs
         422579  context-switches         #      0.000 M/sec
             88  CPU-migrations           #      0.000 M/sec
       91946060  page-faults              #      0.013 M/sec
  5957054385619  cycles                   #    828.333 M/sec
  1058117350365  instructions             #      0.178 IPC
     9161776218  cache-references         #      1.274 M/sec
     1920494280  cache-misses             #      0.267 M/sec

  300.218764862  seconds time elapsed

Data from Prarit (kernel compile with make -j64 on a 64
CPU/32G machine)

For a single run

Without patch

real 27m8.988s
user 87m24.916s
sys 382m6.037s

With patch

real    4m18.607s
user    84m58.943s
sys     50m52.682s

With config turned off

real    4m54.972s
user    90m13.456s
sys     50m19.711s

NOTE: The data looks counterintuitive due to the increased performance
with the patch, even over the config being turned off. We probably need
more runs, but so far all testing has shown that the patches definitely
help.

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Andi Kleen <andi@firstfloor.org>
Reviewed-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Reviewed-by: Daisuke Nishimura <nishimura@mxp.nes.nec.co.jp>
Cc: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Cc: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>