pandora-kernel.git
12 years agodrm/i915: Fix inconsistent backlight level during disabled
Takashi Iwai [Wed, 16 Nov 2011 09:58:03 +0000 (10:58 +0100)]
drm/i915: Fix inconsistent backlight level during disabled

When the brightness property is inquired while the backlight is disabled,
the driver returns a wrong value (zero) because it probes the value after
the backlight was turned off.  This caused a black screen even after the
backlight is enabled again.  It should return the internal backlight_level
instead, so that it won't be influenced by the backlight-enable state.

BugLink: https://bugs.freedesktop.org/show_bug.cgi?id=41926
BugLink: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/872652
Tested-by: Kamal Mostafa <kamal@canonical.com>
Cc: Alex Davis <alex14641@yahoo.com>
Cc: <stable@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Keith Packard <keithp@keithp.com>
12 years agodrm, i915: Fix memory leak in i915_gem_busy_ioctl().
Rakib Mullick [Tue, 15 Nov 2011 18:49:28 +0000 (00:49 +0600)]
drm, i915: Fix memory leak in i915_gem_busy_ioctl().

A call to i915_add_request() has been made in function i915_gem_busy_ioctl(). i915_add_request can fail,
so in it's exit path previously allocated memory needs to be freed.

Signed-off-by: Rakib Mullick <rakib.mullick@gmail.com>
Reviewed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
12 years agoPM / shmobile: Fix build of sh7372_pm_init() for CONFIG_PM unset
Rafael J. Wysocki [Thu, 17 Nov 2011 20:39:55 +0000 (21:39 +0100)]
PM / shmobile: Fix build of sh7372_pm_init() for CONFIG_PM unset

Fix build regression introduced by commit 056879d2f244001b2888cdc8cf
(ARM: mach-shmobile: sh7372 A3SP no_suspend_console fix) by moving
the intialization of the A3SP domain to a separate function and
providing an empty definition of it for CONFIG_PM unset.

Reported-and-tested-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
12 years agoPM Sleep: Do not extend wakeup paths to devices with ignore_children set
Rafael J. Wysocki [Thu, 17 Nov 2011 20:39:33 +0000 (21:39 +0100)]
PM Sleep: Do not extend wakeup paths to devices with ignore_children set

Commit 4ca46ff3e0d8c234cb40ebb6457653b59584426c (PM / Sleep: Mark
devices involved in wakeup signaling during suspend) introduced
the power.wakeup_path field in struct dev_pm_info to mark devices
whose children are enabled to wake up the system from sleep states,
so that power domains containing the parents that provide their
children with wakeup power and/or relay their wakeup signals are not
turned off.  Unfortunately, that introduced a PM regression on SH7372
whose power consumption in the system "memory sleep" state increased
as a result of it, because it prevented the power domain containing
the I2C controller from being turned off when some children of that
controller were enabled to wake up the system, although the
controller was not necessary for them to signal wakeup.

To fix this issue use the observation that devices whose
power.ignore_children flag is set for runtime PM should be treated
analogously during system suspend.  Namely, they shouldn't be
included in wakeup paths going through their children.  Since the
SH7372 I2C controller's power.ignore_children flag is set, doing so
will restore the previous behavior of that SOC.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@suse.de>
12 years agoTTY: ldisc, wait for ldisc infinitely in hangup
Jiri Slaby [Wed, 16 Nov 2011 15:27:09 +0000 (16:27 +0100)]
TTY: ldisc, wait for ldisc infinitely in hangup

For /dev/console case, we do not kill all ldisc users. It's due to
redirected_tty_write test in __tty_hangup. In that case there still
might be a process waiting e.g. in n_tty_read for input.

We wait for such processes to disappear. The problem is that we use a
timeout. After this timeout, we continue closing the ldisc and start
freeing tty resources. It obviously leads to crashes when the other
process is woken.

So to fix this, we wait infinitely before reiniting the ldisc. (The
tiocsetd remains untouched -- times out after 5s.)

This is nicely reproducible with this run from shell:
  exec 0<>/dev/console 1<>/dev/console 2<>/dev/console
and stopping a getty like:
  systemctl stop serial-getty@ttyS0.service

The crash proper may be produced only under load or with constified
timing the same as for 92f6fa09b.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Dave Young <hidave.darkstar@gmail.com>
Cc: Dave Jones <davej@redhat.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: Dmitriy Matrosov <sgf.dma@gmail.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
12 years agoTTY: ldisc, move wait idle to caller
Jiri Slaby [Wed, 16 Nov 2011 15:27:08 +0000 (16:27 +0100)]
TTY: ldisc, move wait idle to caller

It is the only place where reinit is called from. And we really need
to wait for the old ldisc to go once. Actually this is the place where
the waiting originally was (before removed and re-added later).

This will make the fix in the following patch easier to implement.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Dave Young <hidave.darkstar@gmail.com>
Cc: Dave Jones <davej@redhat.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: Dmitriy Matrosov <sgf.dma@gmail.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
12 years agoTTY: ldisc, allow waiting for ldisc arbitrarily long
Jiri Slaby [Wed, 16 Nov 2011 15:27:07 +0000 (16:27 +0100)]
TTY: ldisc, allow waiting for ldisc arbitrarily long

To fix a nasty bug in ldisc hup vs. reinit we need to wait infinitely
long for ldisc to be gone. So here we add a parameter to
tty_ldisc_wait_idle to allow that.

This is only a preparation for the real fix which is done in the
following patches.

Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Dave Young <hidave.darkstar@gmail.com>
Cc: Dave Jones <davej@redhat.com>
Cc: Ben Hutchings <ben@decadent.org.uk>
Cc: Dmitriy Matrosov <sgf.dma@gmail.com>
Cc: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
12 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/linville/wirel...
John W. Linville [Thu, 17 Nov 2011 19:15:47 +0000 (14:15 -0500)]
Merge branch 'master' of git://git./linux/kernel/git/linville/wireless into for-davem

12 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6
Linus Torvalds [Thu, 17 Nov 2011 17:28:54 +0000 (15:28 -0200)]
Merge git://git./linux/kernel/git/jejb/scsi-rc-fixes-2.6

* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6:
  [SCSI] hpsa: Disable ASPM
  [SCSI] aacraid: controller hangs if kernel uses non-default ASPM policy
  [SCSI] mpt2sas: add missing allocation.
  [SCSI] Silencing 'killing requests for dead queue'
  [SCSI] fix WARNING: at drivers/scsi/scsi_lib.c:1704

12 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound
Linus Torvalds [Thu, 17 Nov 2011 17:13:37 +0000 (15:13 -0200)]
Merge branch 'for-linus' of git://git./linux/kernel/git/tiwai/sound

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound:
  ALSA: hda - Fix the connection selection of ADCs on Cirrus codecs
  ALSA: hda - Update URLs in document
  ALSA: hda - move eld->spk_alloc fixup to hdmi_update_eld()
  ALSA: hda - delayed ELD repoll
  ALSA: hda - fix ELD memory leak
  ALSA: hda/realtek: remove redundant semicolon
  ALSA: hda - pwr_nids cleanup for IDT codecs

12 years agoARM: wire up process_vm_writev and process_vm_readv syscalls
Russell King [Thu, 17 Nov 2011 16:58:00 +0000 (16:58 +0000)]
ARM: wire up process_vm_writev and process_vm_readv syscalls

These two syscalls were introduced during the last merge window.
Add the entries into the ARM call tables for them.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
12 years agoRevert "KVM: PPC: Add support for explicit HIOR setting"
Alexander Graf [Thu, 17 Nov 2011 14:26:35 +0000 (15:26 +0100)]
Revert "KVM: PPC: Add support for explicit HIOR setting"

This reverts commit a15bd354f083f20f257db450488db52ac27df439.

It exceeded the padding on the SREGS struct, rendering the ABI
backwards-incompatible.

Conflicts:

arch/powerpc/kvm/powerpc.c
include/linux/kvm.h

Signed-off-by: Avi Kivity <avi@redhat.com>
12 years agoKVM: VMX: Check for automatic switch msr table overflow
Gleb Natapov [Wed, 5 Oct 2011 12:01:24 +0000 (14:01 +0200)]
KVM: VMX: Check for automatic switch msr table overflow

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
12 years agoKVM: VMX: Add support for guest/host-only profiling
Gleb Natapov [Wed, 5 Oct 2011 12:01:23 +0000 (14:01 +0200)]
KVM: VMX: Add support for guest/host-only profiling

Support guest/host-only profiling by switch perf msrs on
a guest entry if needed.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
12 years agoKVM: VMX: add support for switching of PERF_GLOBAL_CTRL
Gleb Natapov [Wed, 5 Oct 2011 12:01:22 +0000 (14:01 +0200)]
KVM: VMX: add support for switching of PERF_GLOBAL_CTRL

Some cpus have special support for switching PERF_GLOBAL_CTRL msr.
Add logic to detect if such support exists and works properly and extend
msr switching code to use it if available. Also extend number of generic
msr switching entries to 8.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
12 years agoKVM: s390: announce SYNC_MMU
Christian Borntraeger [Thu, 17 Nov 2011 10:00:44 +0000 (11:00 +0100)]
KVM: s390: announce SYNC_MMU

KVM on s390 always had a sync mmu. Any mapping change in userspace
mapping was always reflected immediately in the guest mapping.
- In older code the guest mapping was just an offset
- In newer code the last level page table is shared

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
12 years agoKVM: s390: Fix tprot locking
Christian Borntraeger [Thu, 17 Nov 2011 10:00:43 +0000 (11:00 +0100)]
KVM: s390: Fix tprot locking

There is a potential host deadlock in the tprot intercept handling.
We must not hold the mmap semaphore while resolving the guest
address. If userspace is remapping, then the memory detection in
the guest is broken anyway so we can safely separate the
address translation from walking the vmas.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
12 years agoKVM: s390: handle SIGP sense running intercepts
Cornelia Huck [Thu, 17 Nov 2011 10:00:42 +0000 (11:00 +0100)]
KVM: s390: handle SIGP sense running intercepts

SIGP sense running may cause an intercept on higher level
virtualization, so handle it by checking the CPUSTAT_RUNNING flag.

Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
12 years agoKVM: s390: Fix RUNNING flag misinterpretation
Cornelia Huck [Thu, 17 Nov 2011 10:00:41 +0000 (11:00 +0100)]
KVM: s390: Fix RUNNING flag misinterpretation

CPUSTAT_RUNNING was implemented signifying that a vcpu is not stopped.
This is not, however, what the architecture says: RUNNING should be
set when the host is acting on the behalf of the guest operating
system.

CPUSTAT_RUNNING has been changed to be set in kvm_arch_vcpu_load()
and to be unset in kvm_arch_vcpu_put().

For signifying stopped state of a vcpu, a host-controlled bit has
been used and is set/unset basically on the reverse as the old
CPUSTAT_RUNNING bit (including pushing it down into stop handling
proper in handle_stop()).

Cc: stable@kernel.org
Signed-off-by: Cornelia Huck <cornelia.huck@de.ibm.com>
Signed-off-by: Carsten Otte <cotte@de.ibm.com>
Signed-off-by: Avi Kivity <avi@redhat.com>
12 years agoMerge branch 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Thu, 17 Nov 2011 13:46:26 +0000 (11:46 -0200)]
Merge branch 'irq-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq: Fix irqfixup, irqpoll regression

12 years agofix typo/thinko in get_random_bytes()
Luck, Tony [Wed, 16 Nov 2011 18:50:56 +0000 (10:50 -0800)]
fix typo/thinko in get_random_bytes()

If there is an architecture-specific random number generator we use it
to acquire randomness one "long" at a time.  We should put these random
words into consecutive words in the result buffer - not just overwrite
the first word again and again.

Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: H. Peter Anvin <hpa@zytor.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agowriteback: remove vm_dirties and task->dirties
Wu Fengguang [Wed, 7 Sep 2011 16:41:32 +0000 (10:41 -0600)]
writeback: remove vm_dirties and task->dirties

They are not used any more.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
12 years agowriteback: hard throttle 1000+ dd on a slow USB stick
Wu Fengguang [Mon, 14 Nov 2011 01:47:32 +0000 (19:47 -0600)]
writeback: hard throttle 1000+ dd on a slow USB stick

The sleep based balance_dirty_pages() can pause at most MAX_PAUSE=200ms
on every 1 4KB-page, which means it cannot throttle a task under
4KB/200ms=20KB/s. So when there are more than 512 dd writing to a
10MB/s USB stick, its bdi dirty pages could grow out of control.

Even if we can increase MAX_PAUSE, the minimal (task_ratelimit = 1)
means a limit of 4KB/s.

They can eventually be safeguarded by the global limit check
(nr_dirty < dirty_thresh). However if someone is also writing to an
HDD at the same time, it'll get poor HDD write performance.

We at least want to maintain good write performance for other devices
when one device is attacked by some "massive parallel" workload, or
suffers from slow write bandwidth, or somehow get stalled due to some
error condition (eg. NFS server not responding).

For a stalled device, we need to completely block its dirtiers, too,
before its bdi dirty pages grow all the way up to the global limit and
leave no space for the other functional devices.

So change the loop exit condition to

/*
 * Always enforce global dirty limit; also enforce bdi dirty limit
 * if the normal max_pause sleeps cannot keep things under control.
 */
if (nr_dirty < dirty_thresh &&
    (bdi_dirty < bdi_thresh || bdi->dirty_ratelimit > 1))
break;

which can be further simplified to

if (task_ratelimit)
break;

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
12 years agoplatform/x86: fix dell-laptop function prototypes
Randy Dunlap [Thu, 17 Nov 2011 02:20:51 +0000 (18:20 -0800)]
platform/x86: fix dell-laptop function prototypes

Fix build warnings:
  drivers/platform/x86/dell-laptop.c:592:13: warning: function declaration isn't a prototype
  drivers/platform/x86/dell-laptop.c:599:13: warning: function declaration isn't a prototype

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Cc: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agomedia/staging: fix allyesconfig build error
Randy Dunlap [Thu, 17 Nov 2011 02:17:03 +0000 (18:17 -0800)]
media/staging: fix allyesconfig build error

Fix x86 allyesconfig builds.  Builds fail due to a non-static variable
named 'debug' in drivers/staging/media/as102:

  arch/x86/built-in.o:arch/x86/kernel/entry_32.S:1296: first defined here
  ld: Warning: size of symbol `debug' changed from 90 in arch/x86/built-in.o to 4 in drivers/built-in.o

Thou shalt have no non-static identifiers that are named 'debug'.

Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Cc: Pierrick Hascoet <pierrick.hascoet@abilis.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
12 years agohwmon: (exynos4_tmu) Fix Kconfig dependency
Donggeun Kim [Thu, 17 Nov 2011 10:33:05 +0000 (05:33 -0500)]
hwmon: (exynos4_tmu) Fix Kconfig dependency

Rename dependency of EXYNOS4_TMU in Kconfig to the existing one.

Signed-off-by: Donggeun Kim <dg77.kim@samsung.com>
Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>
12 years agostmmac: fix pm functions avoiding sleep on spinlock
Francesco Virlinzi [Wed, 16 Nov 2011 21:58:02 +0000 (21:58 +0000)]
stmmac: fix pm functions avoiding sleep on spinlock

This patch fixes the pm functions to avoid the system
sleeps while a spinlock is taken.

Signed-off-by: Francesco Virlinzi <francesco.virlinzi@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agostmmac: remove spin_lock in stmmac_ioctl.
Srinivas Kandagatla [Wed, 16 Nov 2011 21:58:01 +0000 (21:58 +0000)]
stmmac: remove spin_lock in stmmac_ioctl.

This patch removes un-needed spin_lock in stmmac_ioctl while reading and
writing mdio registers. While holding spin_lock the code must be
atomic, which is not true in this case as both mdiobus_read and writes
have mutex locks.

Without this patch reading mdio registers via mii-tool results in below
BUG:
mii-tool -vvv eth0"
Using SIOCGMIIPHY=0x8947
BUG: sleeping function called from invalid context at kernel/mutex.c:287
in_atomic(): 1, irqs_disabled(): 0, pid: 614, name: mii-tool
2 locks held by mii-tool/614:
 #0:  (rtnl_mutex){......}, at: [<c01fd80c>] dev_ioctl+0x550/0x674
 #1:  (&priv->lock){......}, at: [<c01b34ec>] stmmac_ioctl+0x4c/0x78
[<c002ea14>] (unwind_backtrace+0x0/0xcc) from [<c0272c38>]
(mutex_lock_nested+0x24/0x35c)
[<c0272c38>] (mutex_lock_nested+0x24/0x35c) from [<c01b237c>]
(mdiobus_read+0x44/0x70)
[<c01b237c>] (mdiobus_read+0x44/0x70) from [<c01b0c64>]
(phy_mii_ioctl+0x4c/0x138)
[<c01b0c64>] (phy_mii_ioctl+0x4c/0x138) from [<c01b34fc>]
(stmmac_ioctl+0x5c/0x78)
[<c01b34fc>] (stmmac_ioctl+0x5c/0x78) from [<c01fcec8>]
(dev_ifsioc+0x2a4/0x2c8)
[<c01fcec8>] (dev_ifsioc+0x2a4/0x2c8) from [<c01fd81c>]
(dev_ioctl+0x560/0x674)
[<c01fd81c>] (dev_ioctl+0x560/0x674) from [<c00c36e0>]
(vfs_ioctl+0x2c/0x8c)
[<c00c36e0>] (vfs_ioctl+0x2c/0x8c) from [<c00c4130>]
(do_vfs_ioctl+0x530/0x578)
[<c00c4130>] (do_vfs_ioctl+0x530/0x578) from [<c00c41ac>]
(sys_ioctl+0x34/0x54)
[<c00c41ac>] (sys_ioctl+0x34/0x54) from [<c0028aa0>]
(ret_fast_syscall+0x0/0x2c)

Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@st.com>
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agostmmac: parameters auto-tuning through HW cap reg
Giuseppe CAVALLARO [Wed, 16 Nov 2011 21:58:00 +0000 (21:58 +0000)]
stmmac: parameters auto-tuning through HW cap reg

New GMAC devices (newer than the databook 3.50a) have the
HW capability register that provides which features are actually
supported by the hardware.

On old devices many information have to be passed through the
platform, for example: enhanced descriptor structure,
TX COE etc. These are mandatory to properly configure the driver.
This remains still valid because the driver has to support old
Synopsys devices but now it's also able to override them using the
values from the HW capability register if supported.

Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agostmmac: fix advertising 1000Base capabilties for non GMII iface
Srinivas Kandagatla [Wed, 16 Nov 2011 21:57:59 +0000 (21:57 +0000)]
stmmac: fix advertising 1000Base capabilties for non GMII iface

This patch fixes the way to stop the 1000Base advertising
capabilties for non GMII interfaces.

Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@st.com>
Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agostmmac: use mdelay on timeout of sw reset
Francesco Virlinzi [Wed, 16 Nov 2011 21:57:58 +0000 (21:57 +0000)]
stmmac: use mdelay on timeout of sw reset

This patch uses an mdelay to manage the timeout on
sw reset to be independant of cpu_clk.

Signed-off-by: Francesco Virlinzi <francesco.virlinzi@st.com>
Reviewed-by: Giuseppe Cavallaro <peppe.cavallaro@st.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agopowerpc/signal32: Fix sigset_t conversion when copying to user
Will Deacon [Tue, 8 Nov 2011 04:51:19 +0000 (04:51 +0000)]
powerpc/signal32: Fix sigset_t conversion when copying to user

On PPC64, put_sigset_t converts a sigset_t to a compat_sigset_t
before copying it to userspace. There is a typo in the case that
we have 4 words to copy, meaning that we corrupt the compat_sigset_t.

It appears that _NSIG_WORDS can't be greater than 2 at the moment
so this code is probably always optimised away anyway.

Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Fix atomic_xxx_return barrier semantics
Benjamin Herrenschmidt [Tue, 15 Nov 2011 17:11:27 +0000 (17:11 +0000)]
powerpc: Fix atomic_xxx_return barrier semantics

The Documentation/memory-barriers.txt document requires that atomic
operations that return a value act as a memory barrier both before
and after the actual atomic operation.

Our current implementation doesn't guarantee this. More specifically,
while a load following the isync can not be issued before stwcx. has
completed, that completion doesn't architecturally means that the
result of stwcx. is visible to other processors (or any previous stores
for that matter) (typically, the other processors L1 caches can still
hold the old value).

This has caused an actual crash in RCU torture testing on Power 7

This fixes it by changing those atomic ops to use new macros instead
of RELEASE/ACQUIRE barriers, called ATOMIC_ENTRY and ATMOIC_EXIT barriers,
which are then defined respectively to lwsync and sync.

I haven't had a chance to measure the performance impact (or rather
what I measured with kernel compiles is in the noise, I yet have to
find a more precise benchmark)

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
12 years agopowerpc: Remove buggy 9-year-old test for binutils < 2.12.1
Kyle Moffett [Mon, 7 Nov 2011 13:30:43 +0000 (13:30 +0000)]
powerpc: Remove buggy 9-year-old test for binutils < 2.12.1

Recent binutils refuses to assemble AltiVec opcodes when in e500/SPE
mode, as some of those opcodes alias the "SPE" instructions.  This
triggers an ancient binutils version check even when building a kernel
with CONFIG_ALTIVEC disabled.

In theory, the check could be conditionalized on CONFIG_ALTIVEC, but in
practice it has long outlived its utility.  It is virtually impossible
to find binutils older than 2.12.1 (released 2002) in the wild anymore.
Even ancient RedHat Enterprise Linux 4 has binutils-2.14.

To fix the kernel build when done natively on e500 systems with this new
binutils, the test is simply removed.

Signed-off-by: Kyle Moffett <Kyle.D.Moffett@boeing.com>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/book3e-64: Fix debug support for userspace
Kumar Gala [Thu, 6 Oct 2011 02:53:40 +0000 (02:53 +0000)]
powerpc/book3e-64: Fix debug support for userspace

With the introduction of CONFIG_PPC_ADV_DEBUG_REGS user space debug is
broken on Book-E 64-bit parts that support delayed debug events.  When
switch_booke_debug_regs() sets DBCR0 we'll start getting debug events as
MSR_DE is also set and we aren't able to handle debug events from kernel
space.

We can remove the hack that always enables MSR_DE and loads up DBCR0 and
just utilize switch_booke_debug_regs() to get user space debug working
again.

We still need to handle critical/debug exception stacks & proper
save/restore of state for those exception levles to support debug events
from kernel space like we have on 32-bit.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Remove extraneous CONFIG_PPC_ADV_DEBUG_REGS define
Kumar Gala [Thu, 6 Oct 2011 02:53:39 +0000 (02:53 +0000)]
powerpc: Remove extraneous CONFIG_PPC_ADV_DEBUG_REGS define

All of DebugException is already protected by CONFIG_PPC_ADV_DEBUG_REGS
there is no need to have another such ifdef inside the function.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Revert show_regs() define for readability
Kumar Gala [Thu, 6 Oct 2011 02:53:38 +0000 (02:53 +0000)]
powerpc: Revert show_regs() define for readability

We had an existing ifdef for 4xx & BOOKE processors that got changed to
CONFIG_PPC_ADV_DEBUG_REGS.  The define has nothing to do with
CONFIG_PPC_ADV_DEBUG_REGS.  The define really should be:

 #if defined(CONFIG_4xx) || defined(CONFIG_BOOKE)

and not

 #ifdef CONFIG_PPC_ADV_DEBUG_REGS

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agodrm/i915: Use DPCD value for max DP lanes.
Keith Packard [Wed, 2 Nov 2011 20:03:47 +0000 (13:03 -0700)]
drm/i915: Use DPCD value for max DP lanes.

The BIOS VBT value for an eDP panel has been shown to be incorrect on
one machine, and we haven't found any machines where the DPCD value
was wrong, so we'll use the DPCD value everywhere.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
12 years agodrm/i915: Initiate DP link training only on the lanes we'll be using
Keith Packard [Wed, 2 Nov 2011 17:17:59 +0000 (10:17 -0700)]
drm/i915: Initiate DP link training only on the lanes we'll be using

Limit the link training setting command to the lanes needed for the
current mode. It seems vaguely possible that a monitor will try to
train the other lanes and fail in some way, so this seems like the
safer plan.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
12 years agodrm/i915: Remove trailing white space
Keith Packard [Wed, 2 Nov 2011 03:01:35 +0000 (20:01 -0700)]
drm/i915: Remove trailing white space

Found a couple of bare tabs in intel_dp.c

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
12 years agodrm/i915: Try harder during dp pattern 1 link training
Keith Packard [Wed, 2 Nov 2011 03:00:06 +0000 (20:00 -0700)]
drm/i915: Try harder during dp pattern 1 link training

Instead of going through the sequence just once, run through the whole
set up to 5 times to see if something can work. This isn't part of the
DP spec, but the BIOS seems to do it, and given that link training
failure is so bad, it seems reasonable to follow suit.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
12 years agodrm/i915: Make DP prepare/commit consistent with DP dpms
Keith Packard [Wed, 2 Nov 2011 03:25:21 +0000 (20:25 -0700)]
drm/i915: Make DP prepare/commit consistent with DP dpms

Make sure the sequence of operations in all three functions makes
sense:

 1) The backlight must be off unless the screen is running
 2) The link must be running to turn the eDP panel on/off
 3) The CPU eDP PLL must be running until everything is off

Signed-off-by: Keith Packard <keithp@keithp.com>
12 years agodrm/i915: Let panel power sequencing hardware do its job
Keith Packard [Wed, 2 Nov 2011 02:57:50 +0000 (19:57 -0700)]
drm/i915: Let panel power sequencing hardware do its job

The panel power sequencing hardware tracks the stages of panel power
sequencing and signals when the panel is completely on or off. Instead
of blindly assuming the panel timings will work, poll the panel power
status register until it shows the correct values.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
12 years agodrm/i915: Treat PCH eDP like DP in most places
Keith Packard [Wed, 2 Nov 2011 02:54:11 +0000 (19:54 -0700)]
drm/i915: Treat PCH eDP like DP in most places

PCH eDP has many of the same needs as regular PCH DP connections,
including the DP_CTl bit settings, the TRANS_DP_CTL register.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
12 years agodrm/i915: Remove link_status field from intel_dp structure
Keith Packard [Wed, 2 Nov 2011 02:45:03 +0000 (19:45 -0700)]
drm/i915: Remove link_status field from intel_dp structure

No persistent data was ever stored here, so link_status is instead
allocated on the stack as needed.

Signed-off-by: Keith Packard <keithp@keithp.com>
Reviewed-by: Jesse Barnes <jbarnes@virtuousgeek.org>
12 years agodrm/i915: Move common PCH_PP_CONTROL setup to ironlake_get_pp_control
Keith Packard [Wed, 2 Nov 2011 02:34:06 +0000 (19:34 -0700)]
drm/i915: Move common PCH_PP_CONTROL setup to ironlake_get_pp_control

Every usage of PCH_PP_CONTROL sets the PANEL_UNLOCK_REGS value to
ensure that writes will be respected, move this to a common function
to make the driver cleaner.

No functional changes.

Signed-off-by: Keith Packard <keithp@keithp.com>
12 years agonew helper: mount_subtree()
Al Viro [Thu, 17 Nov 2011 02:43:59 +0000 (21:43 -0500)]
new helper: mount_subtree()

takes vfsmount and relative path, does lookup within that vfsmount
(possibly triggering automounts) and returns the result as root
of subtree suitable for return by ->mount() (i.e. a reference to
dentry and an active reference to its superblock grabbed, superblock
locked exclusive).

btrfs and nfs switched to it instead of open-coding the sucker.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
12 years agosky2: version 1.30
stephen hemminger [Wed, 16 Nov 2011 13:43:00 +0000 (13:43 +0000)]
sky2: version 1.30

Update version number.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosky2: used fixed RSS key
stephen hemminger [Wed, 16 Nov 2011 13:42:59 +0000 (13:42 +0000)]
sky2: used fixed RSS key

Rather than generating a different RSS key on each boot, just use
a predetermined value that will map same flow to same value on
every device.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosky2: reduce default Tx ring size
stephen hemminger [Wed, 16 Nov 2011 13:42:58 +0000 (13:42 +0000)]
sky2: reduce default Tx ring size

The default Tx ring size for the sky2 driver is quite large and could
cause excess buffer bloat for many users. The minimum ring size
possible and still allow handling  the worst case packet on 64bit platforms
is 38 which gets rounded up to a power of 2. But most packets only require
a couple of ring elements.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosky2: rename up/down functions
stephen hemminger [Wed, 16 Nov 2011 13:42:57 +0000 (13:42 +0000)]
sky2: rename up/down functions

The code is clearer if the up/down functions are renamed to
open/close like other drivers.  Purely syntax change.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosky2: pci posting issues
stephen hemminger [Wed, 16 Nov 2011 13:42:56 +0000 (13:42 +0000)]
sky2: pci posting issues

A couple of the reset and setup paths have possible PCI posting issues.
When setting registers, a read is necessary to force the writes to complete.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agosky2: fix hang on shutdown (and other irq issues)
stephen hemminger [Wed, 16 Nov 2011 13:42:55 +0000 (13:42 +0000)]
sky2: fix hang on shutdown (and other irq issues)

There are several problems with recent change to how IRQ's are setup.
   * synchronize_irq in sky2_shutdown would hang because there
     was no IRQ setup.
   * when device was set to down, some IRQ bits left enabled so a
     hardware error would produce IRQ with no handler
   * quick link on Optima chip set was enabled without handler
   * suspend/resume would leave IRQ on with no handler if device
     was down

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agor6040: fix check against MCRO_HASHEN bit in r6040_multicast_list
Florian Fainelli [Wed, 16 Nov 2011 06:00:08 +0000 (06:00 +0000)]
r6040: fix check against MCRO_HASHEN bit in r6040_multicast_list

We are checking whether the MCR0_HASHEN bit is set using a logical and
instead of bitwise and, fix that.

Signed-off-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agorandom: Fix handing of arch_get_random_long in get_random_bytes()
Luck, Tony [Wed, 16 Nov 2011 18:50:56 +0000 (10:50 -0800)]
random: Fix handing of arch_get_random_long in get_random_bytes()

If there is an architecture-specific random number generator we use
it to acquire randomness one "long" at a time. We should put these
random words into consecutive words in the result buffer - not just
overwrite the first word again and again.

Signed-off-by: Tony Luck <tony.luck@intel.com>
Acked-by: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/4ec4061010261a4cb0@agluck-desktop.sc.intel.com
Signed-off-by: H. Peter Anvin <hpa@linux.intel.com>
12 years agoMAINTAINERS: change email address for shemminger
Stephen Hemminger [Wed, 16 Nov 2011 22:58:46 +0000 (17:58 -0500)]
MAINTAINERS: change email address for shemminger

My old email account at linux-foundation is no longer usable after
the LF breakin.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agopch_gbe: Move #include of module.h
Joe Perches [Tue, 15 Nov 2011 13:56:14 +0000 (13:56 +0000)]
pch_gbe: Move #include of module.h

The first #include must be pch_gbe.h as it
does a #define of pr_fmt.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agobnx2x: cache-in compressed fw image
Dmitry Kravkov [Tue, 15 Nov 2011 12:07:33 +0000 (12:07 +0000)]
bnx2x: cache-in compressed fw image

Re-request fw from fs may fail for different reasons, once the fw was
loaded we won't release it until driver is removed.

This also resolves the boot problem when initial fw is located on initrd,
but rootfs is still unavailable, in this case device reset will fail due
to absence of fw files.

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonet/cadence: enable by default NET_ATMEL
Jean-Christophe PLAGNIOL-VILLARD [Tue, 15 Nov 2011 09:36:30 +0000 (09:36 +0000)]
net/cadence: enable by default NET_ATMEL

so the defconfig of the atmel continue to have the support of the network
as before

Signed-off-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
Cc: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agobridge: correct IPv6 checksum after pull
stephen hemminger [Tue, 15 Nov 2011 08:09:14 +0000 (08:09 +0000)]
bridge: correct IPv6 checksum after pull

Bridge multicast snooping of ICMPv6 would incorrectly report a checksum problem
when used with Ethernet devices like sky2 that use CHECKSUM_COMPLETE.
When bytes are removed from skb, the computed checksum needs to be adjusted.

Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Tested-by: Martin Volf <martin.volf.42@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoqeth: Reduce CPU consumption through less SIGA-r calls
Einar Lueck [Tue, 15 Nov 2011 02:31:16 +0000 (02:31 +0000)]
qeth: Reduce CPU consumption through less SIGA-r calls

Patch avoids SIGA-r calls in case of SIGA-r required. It only calls
SIGA-r if a threshold of free buffer is reached. CPU consumption is
reduced as a consequence.

Signed-off-by: Einar Lueck <elelueck@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoqeth: l3 fix rcu splat in xmit
Frank Blaschka [Tue, 15 Nov 2011 02:31:15 +0000 (02:31 +0000)]
qeth: l3 fix rcu splat in xmit

when use dst_get_neighbour to get neighbour, we need
rcu_read_lock to protect, since dst_get_neighbour uses
rcu_dereference.

Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agonetiucv: reinsert dev_alloc_name for device naming
Ursula Braun [Tue, 15 Nov 2011 02:31:14 +0000 (02:31 +0000)]
netiucv: reinsert dev_alloc_name for device naming

Invocation of dev_alloc_name() is re-inserted, because the created
net_device name is used to create the device name for the iucv bus.
This device is created before the register_netdev call.

Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoqeth: remove WARN_ON leftover
Christian Borntraeger [Tue, 15 Nov 2011 02:31:13 +0000 (02:31 +0000)]
qeth: remove WARN_ON leftover

The patch "qeth: exploit asynchronous delivery of storage blocks"
added a WARN_ON in qeth_schedule_recovery. A device recovery should
not cause a kernel warning. This is obviously a debugging  left-over
that we forgot to remove.

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoqeth: return with -EPERM if sniffing is not enabled
Ursula Braun [Tue, 15 Nov 2011 02:31:12 +0000 (02:31 +0000)]
qeth: return with -EPERM if sniffing is not enabled

Without appropriate configuration at the SE, a HiperSockets device
cannot be used for sniffing. Setting the sniffer attribute is rejected
with -EPERM.

Signed-off-by: Ursula Braun <ursula.braun@de.ibm.com>
Signed-off-by: Frank Blaschka <frank.blaschka@de.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agotcp: clear xmit timers in tcp_v4_syn_recv_sock()
Eric Dumazet [Mon, 14 Nov 2011 10:56:56 +0000 (10:56 +0000)]
tcp: clear xmit timers in tcp_v4_syn_recv_sock()

Simon Kirby reported divides by zero errors in __tcp_select_window()

This happens when inet_csk_route_child_sock() returns a NULL pointer :

We free new socket while we eventually armed keepalive timer in
tcp_create_openreq_child()

Fix this by a call to tcp_clear_xmit_timers()

[ This is a followup to commit 918eb39962dff (net: add missing
bh_unlock_sock() calls) ]

Reported-by: Simon Kirby <sim@hostway.ca>
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com>
Tested-by: Simon Kirby <sim@hostway.ca>
Signed-off-by: David S. Miller <davem@davemloft.net>
12 years agoswitch create_mnt_ns() to saner calling conventions, fix double mntput() in nfs
Al Viro [Wed, 16 Nov 2011 21:12:14 +0000 (16:12 -0500)]
switch create_mnt_ns() to saner calling conventions, fix double mntput() in nfs

Life is much saner if create_mnt_ns(mnt) drops mnt in case of error...
Switch it to such calling conventions, switch callers, fix double mntput() in
fs/nfs/super.c one.

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
12 years agobtrfs: fix double mntput() in mount_subvol()
Al Viro [Wed, 16 Nov 2011 21:06:09 +0000 (16:06 -0500)]
btrfs: fix double mntput() in mount_subvol()

Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
12 years agowatchdog: fix initialisation printout in s3c2410_wdt
Dmitry Artamonow [Wed, 16 Nov 2011 08:46:13 +0000 (12:46 +0400)]
watchdog: fix initialisation printout in s3c2410_wdt

Looks like a typo creeped in, and driver prints
s3c2410-wdt s3c2410-wdt: watchdog active, reset abled, irq abled

instead of
s3c2410-wdt s3c2410-wdt: watchdog active, reset enabled, irq enabled

Also it may completely disinform about irq status, as it prints
"irq enabled" when S3C2410_WTCON_INTEN is in fact 0.

Fix it.

Signed-off-by: Dmitry Artamonow <mad_soft@inbox.ru>
Tested-by: Thomas Abraham <thomas.abraham@linaro.org>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
12 years agowatchdog: Don't overwrite error value in wm831x_wdt_set_timeout()
Mark Brown [Wed, 16 Nov 2011 11:52:12 +0000 (11:52 +0000)]
watchdog: Don't overwrite error value in wm831x_wdt_set_timeout()

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Wim Van Sebroeck <wim@iguana.be>
12 years agoencrypted-keys: module build fixes
Mimi Zohar [Sun, 16 Oct 2011 23:17:48 +0000 (19:17 -0400)]
encrypted-keys: module build fixes

Encrypted keys are encrypted/decrypted using either a trusted or
user-defined key type, which is referred to as the 'master' key.
The master key may be of type trusted iff the trusted key is
builtin or both the trusted key and encrypted keys are built as
modules.  This patch resolves the build dependency problem.

- Use "masterkey-$(CONFIG_TRUSTED_KEYS)-$(CONFIG_ENCRYPTED_KEYS)" construct
to encapsulate the above logic. (Suggested by Dimtry Kasatkin.)
- Fixing the encrypted-keys Makefile, results in a module name change
from encrypted.ko to encrypted-keys.ko.
- Add module dependency for request_trusted_key() definition

Signed-off-by: Mimi Zohar <zohar@us.ibm.com>
12 years agoencrypted-keys: fix error return code
Mimi Zohar [Mon, 24 Oct 2011 12:17:42 +0000 (08:17 -0400)]
encrypted-keys: fix error return code

Fix request_master_key() error return code.

Signed-off-by: Mimi Zohar <zohar@us.ibm.com>
12 years agoxen-gntalloc: signedness bug in add_grefs()
Dan Carpenter [Fri, 4 Nov 2011 18:24:36 +0000 (21:24 +0300)]
xen-gntalloc: signedness bug in add_grefs()

gref->gref_id is unsigned so the error handling didn't work.
gnttab_grant_foreign_access() returns an int type, so we can add a
cast here, and it doesn't cause any problems.
gnttab_grant_foreign_access() can return a variety of errors
including -ENOSPC, -ENOSYS and -ENOMEM.

CC: stable@kernel.org
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoxen-gntalloc: integer overflow in gntalloc_ioctl_alloc()
Dan Carpenter [Fri, 4 Nov 2011 18:24:08 +0000 (21:24 +0300)]
xen-gntalloc: integer overflow in gntalloc_ioctl_alloc()

On 32 bit systems a high value of op.count could lead to an integer
overflow in the kzalloc() and gref_ids would be smaller than
expected.  If the you triggered another integer overflow in
"if (gref_size + op.count > limit)" then you'd probably get memory
corruption inside add_grefs().

CC: stable@kernel.org
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoxen-gntdev: integer overflow in gntdev_alloc_map()
Dan Carpenter [Fri, 4 Nov 2011 18:23:32 +0000 (21:23 +0300)]
xen-gntdev: integer overflow in gntdev_alloc_map()

The multiplications here can overflow resulting in smaller buffer
sizes than expected.  "count" comes from a copy_from_user().

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoxen:pvhvm: enable PVHVM VCPU placement when using more than 32 CPUs.
Zhenzhong Duan [Fri, 28 Oct 2011 05:28:59 +0000 (22:28 -0700)]
xen:pvhvm: enable PVHVM VCPU placement when using more than 32 CPUs.

PVHVM running with more than 32 vcpus and pv_irq/pv_time enabled
need VCPU placement to work, or else it will softlockup.

CC: stable@kernel.org
Acked-by: Stefano Stabellini <stefano.stabellini@eu.citrix.com>
Signed-off-by: Zhenzhong Duan <zhenzhong.duan@oracle.com>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoxen/balloon: Avoid OOM when requesting highmem
Daniel De Graaf [Wed, 19 Oct 2011 21:59:37 +0000 (17:59 -0400)]
xen/balloon: Avoid OOM when requesting highmem

If highmem pages are requested from the balloon on a system without
highmem, the implementation of alloc_xenballooned_pages will allocate
all available memory trying to find highmem pages to return. Allow
low memory to be returned when highmem pages are requested to avoid
this loop.

Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoxen: Remove hanging references to CONFIG_XEN_PLATFORM_PCI
Daniel De Graaf [Wed, 19 Oct 2011 22:05:27 +0000 (18:05 -0400)]
xen: Remove hanging references to CONFIG_XEN_PLATFORM_PCI

In 5fbdc10395cd500d6ff844825a918c4e6f38de37 the XEN_PLATFORM_PCI config
option was removed, but references in header files remained. Clean up
those references.

Signed-off-by: Daniel De Graaf <dgdegra@tycho.nsa.gov>
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
12 years agoxen: map foreign pages for shared rings by updating the PTEs directly
David Vrabel [Thu, 29 Sep 2011 15:53:32 +0000 (16:53 +0100)]
xen: map foreign pages for shared rings by updating the PTEs directly

When mapping a foreign page with xenbus_map_ring_valloc() with the
GNTTABOP_map_grant_ref hypercall, set the GNTMAP_contains_pte flag and
pass a pointer to the PTE (in init_mm).

After the page is mapped, the usual fault mechanism can be used to
update additional MMs.  This allows the vmalloc_sync_all() to be
removed from alloc_vm_area().

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Acked-by: Andrew Morton <akpm@linux-foundation.org>
[v1: Squashed fix by Michal for no-mmu case]
Signed-off-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Michal Simek <monstr@monstr.eu>
12 years agoALSA: hda - Fix the connection selection of ADCs on Cirrus codecs
Takashi Iwai [Wed, 16 Nov 2011 17:05:11 +0000 (18:05 +0100)]
ALSA: hda - Fix the connection selection of ADCs on Cirrus codecs

spec->cur_adc isn't set until cs_capture_pcm_prepare() is called although
the driver tries to select the connection at init time and at auto-mic
switch.  This results in the access to the widget NID 0, which is
obviously invalid, also a wrong capture source.

This patch fixes the issue by issuing the connect-select verb conditionally
at appropriate places.

Reported-and-tested-by: Dylan Reid <dgreid@chromium.org>
Cc: <stable@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
12 years agomm: Make task in balance_dirty_pages() killable
Jan Kara [Wed, 16 Nov 2011 11:34:48 +0000 (19:34 +0800)]
mm: Make task in balance_dirty_pages() killable

There is no reason why task in balance_dirty_pages() shouldn't be killable
and it helps in recovering from some error conditions (like when filesystem
goes in error state and cannot accept writeback anymore but we still want to
kill processes using it to be able to unmount it).

There will be follow up patches to further abort the generic_perform_write()
and other filesystem write loops, to avoid large write + SIGKILL combination
exceeding the dirty limit and possibly strange OOM.

Reported-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Tested-by: Kazuya Mio <k-mio@sx.jp.nec.com>
Reviewed-by: Neil Brown <neilb@suse.de>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
12 years agoALSA: hda - Update URLs in document
Takashi Iwai [Wed, 16 Nov 2011 09:52:01 +0000 (10:52 +0100)]
ALSA: hda - Update URLs in document

Some stuff was moved from kernel.org to other places.

Signed-off-by: Takashi Iwai <tiwai@suse.de>
12 years agoALSA: hda - move eld->spk_alloc fixup to hdmi_update_eld()
Wu Fengguang [Wed, 16 Nov 2011 08:29:48 +0000 (16:29 +0800)]
ALSA: hda - move eld->spk_alloc fixup to hdmi_update_eld()

It looks more natural and saves two lines of code.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
12 years agoALSA: hda - delayed ELD repoll
Wu Fengguang [Wed, 16 Nov 2011 08:29:47 +0000 (16:29 +0800)]
ALSA: hda - delayed ELD repoll

The Intel HDMI chips (ironlake at least) are found to have ~250ms delay
between the ELD_Valid=1 hotplug event is send and the ELD buffer becomes
actually readable. During the time the ELD buffer is mysteriously all 0.

Fix it by scheduling a delayed work to re-read ELD buffer after 300ms.

Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
12 years agoALSA: hda - fix ELD memory leak
Wu Fengguang [Wed, 16 Nov 2011 08:29:46 +0000 (16:29 +0800)]
ALSA: hda - fix ELD memory leak

memset(eld) clears eld->proc_entry which will leak the struct
snd_info_entry when unloading module.

Fix it by
- memset only the fields before eld->eld_buffer
- set eld->eld_valid to true _after_ all eld fields have been filled

Cc: <stable@kernel.org>
Cc: Pierre-louis Bossart <pierre-louis.bossart@intel.com>
Acked-by: Stephen Warren <swarren@nvidia.com>
Signed-off-by: Wu Fengguang <fengguang.wu@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
12 years agoMerge branch 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6
Linus Torvalds [Wed, 16 Nov 2011 09:33:50 +0000 (07:33 -0200)]
Merge branch 'for-linus' of git://git390.marist.edu/linux-2.6

* 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6:
  [S390] topology: fix topology on z10 machines
  [S390] crypto: avoid MSA3 and MSA4 instructions in ESA mode
  [S390] avoid STCKF if running in ESA mode
  [S390] zfcpdump: Do not initialize zfcpdump in kdump mode
  [S390] ap: Setup processing for messages in request queue.
  [S390] Kconfig: Select CONFIG_KEXEC for CONFIG_CRASH_DUMP
  [S390] incorrect note program header
  [S390] pfault: ignore leftover completion interrupts
  [S390] fix pgste update logic
  [S390] wire up process_vm syscalls

12 years agoblock: add missed trace_block_plug
Shaohua Li [Wed, 16 Nov 2011 08:21:50 +0000 (09:21 +0100)]
block: add missed trace_block_plug

After flush plug list, the list has no request, so we need to add a
trace_block_plug().

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agoparide: fix potential information leak in pg_read()
Dan Carpenter [Wed, 16 Nov 2011 08:21:50 +0000 (09:21 +0100)]
paride: fix potential information leak in pg_read()

Smatch has a new check for Rosenberg type information leaks where structs
are copied to the user with uninitialized stack data in them.  i In this
case, the pg_write_hdr struct has a hole in it.

struct pg_write_hdr {
        char                       magic;                /*     0     1 */
        char                       func;                 /*     1     1 */
        /* XXX 2 bytes hole, try to pack */
        int                        dlen;                 /*     4     4 */

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Tim Waugh <tim@cyberelk.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agobio: change some signed vars to unsigned
Dan Carpenter [Wed, 16 Nov 2011 08:21:50 +0000 (09:21 +0100)]
bio: change some signed vars to unsigned

This is just a cleanup patch to silence a static checker warning.

The problem is that we cap "nr_iovecs" so it can't be larger than
"UIO_MAXIOV" but we don't check for negative values.  It turns out this is
prevented at other layers, but logically it doesn't make sense to have
negative nr_iovecs so making it unsigned is nicer.

Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agoblock: avoid unnecessary plug list flush
Shaohua Li [Wed, 16 Nov 2011 08:21:50 +0000 (09:21 +0100)]
block: avoid unnecessary plug list flush

get_request_wait() could sleep and flush the plug list.  If the list is
already flushed, don't flush again.

Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Reviewed-by: Namhyung Kim <namhyung@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agocciss: auto engage SCSI mid layer at driver load time
Stephen M. Cameron [Wed, 16 Nov 2011 08:21:49 +0000 (09:21 +0100)]
cciss: auto engage SCSI mid layer at driver load time

A long time ago, probably in 2002, one of the distros, or maybe more than
one, loaded block drivers prior to loading the SCSI mid layer.  This meant
that the cciss driver, being a block driver, could not engage the SCSI mid
layer at init time without panicking, and relied on being poked by a
userland program after the system was up (and the SCSI mid layer was
therefore present) to engage the SCSI mid layer.

This is no longer the case, and cciss can safely rely on the SCSI mid
layer being present at init time and engage the SCSI mid layer straight
away.  This means that users will see their tape drives and medium
changers at driver load time without need for a script in /etc/rc.d that
does this:

for x in /proc/driver/cciss/cciss*
do
echo "engage scsi" > $x
done

However, if no tape drives or medium changers are detected, the SCSI mid
layer will not be engaged.  If a tape drive or medium change is later
hot-added to the system it will then be necessary to use the above script
or similar for the device(s) to be acceesible.

Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agoloop: cleanup set_status interface
Dmitry Monakhov [Wed, 16 Nov 2011 08:21:49 +0000 (09:21 +0100)]
loop: cleanup set_status interface

1) Anyone who has read access to loopdev has permission to call set_status
   and may change important parameters such as lo_offset, lo_sizelimit and
   so on, which contradicts to read access pattern and definitely equals
   to write access pattern.
2) Add lo_offset over i_size check to prevent blkdev_size overflow.
   ##Testcase_bagin
   #dd if=/dev/zero of=./file bs=1k count=1
   #losetup /dev/loop0 ./file
   /* userspace_application */
   struct loop_info64 loinf;
   fd = open("/dev/loop0", O_RDONLY);
   ioctl(fd, LOOP_GET_STATUS64, &loinf);
   /* Set offset to any value which is bigger than i_size, and sizelimit
    * to nonzero value*/
   loinf.lo_offset = 4096*1024;
   loinf.lo_sizelimit = 1024;
   ioctl(fd, LOOP_SET_STATUS64, &loinf);
   /* After this loop device will have size similar to 0x7fffffffffxxxx */
   #blockdev --getsz /dev/loop0
   ##OUTPUT: 36028797018955968
   ##Testcase_end

[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agoinclude/linux/bio.h: use a static inline function for bio_integrity_clone()
Stephen Rothwell [Wed, 16 Nov 2011 08:21:48 +0000 (09:21 +0100)]
include/linux/bio.h: use a static inline function for bio_integrity_clone()

When CONFIG_BLK_DEV_INTEGRITY is not set, we get these warnings:

drivers/md/dm.c: In function 'split_bvec':
drivers/md/dm.c:1061:3: warning: statement with no effect
drivers/md/dm.c: In function 'clone_bio':
drivers/md/dm.c:1088:3: warning: statement with no effect

Signed-off-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agoloop: prevent information leak after failed read
Dmitry Monakhov [Wed, 16 Nov 2011 08:21:48 +0000 (09:21 +0100)]
loop: prevent information leak after failed read

If read was not fully successful we have to fail whole bio to prevent
information leak of old pages

##Testcase_begin
dd if=/dev/zero of=./file bs=1M count=1
losetup /dev/loop0 ./file -o 4096
truncate -s 0 ./file
# OOps loop offset is now beyond i_size, so read will silently fail.
# So bio's pages would not be cleared, may which result in information leak.
hexdump -C /dev/loop0
##testcase_end

Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
12 years agopowerpc/ps3: Fix SMP lockdep boot warning
Geoff Levand [Tue, 8 Nov 2011 12:38:21 +0000 (12:38 +0000)]
powerpc/ps3: Fix SMP lockdep boot warning

Move the PS3 IPI message setup from ps3_smp_setup_cpu() to ps3_smp_probe().

Fixes startup warnings like these:

  ------------[ cut here ]------------
  WARNING: at kernel/lockdep.c:2649
  Modules linked in:
  ...
  ---[ end trace 31fd0ba7d8756001 ]---

Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/ps3: Fix lost SMP IPIs
Geoff Levand [Tue, 8 Nov 2011 12:37:26 +0000 (12:37 +0000)]
powerpc/ps3: Fix lost SMP IPIs

Fixes the PS3 bootup hang introduced in 3.0-rc1 by:

  commit 317f394160e9beb97d19a84c39b7e5eb3d7815a
  sched: Move the second half of ttwu() to the remote cpu

Move the PS3's LV1 EOI call lv1_end_of_interrupt_ext() from ps3_chip_eoi()
to ps3_get_irq() for IPI messages.

If lv1_send_event_locally() is called between a previous call to
lv1_send_event_locally() and the coresponding call to
lv1_end_of_interrupt_ext() the second event will not be delivered to the
target cpu.

The PS3's SMP IPIs are implemented using lv1_send_event_locally(), so if two
IPI messages of the same type are sent to the same target in a relatively
short period of time the second IPI event can become lost when
lv1_end_of_interrupt_ext() is called from ps3_chip_eoi().

CC: stable@kernel.org
Signed-off-by: Geoff Levand <geoff@infradead.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Add hvcall.h include to book3s_hv.c
Michael Neuling [Wed, 9 Nov 2011 20:39:49 +0000 (20:39 +0000)]
powerpc: Add hvcall.h include to book3s_hv.c

If you build with KVM and UP it fails with the following due to a
missing include.

/arch/powerpc/kvm/book3s_hv.c: In function 'do_h_register_vpa':
arch/powerpc/kvm/book3s_hv.c:156:10: error: 'H_PARAMETER' undeclared (first use in this function)
arch/powerpc/kvm/book3s_hv.c:156:10: note: each undeclared identifier is reported only once for each function it appears in
arch/powerpc/kvm/book3s_hv.c:192:12: error: 'H_RESOURCE' undeclared (first use in this function)
arch/powerpc/kvm/book3s_hv.c:222:9: error: 'H_SUCCESS' undeclared (first use in this function)
arch/powerpc/kvm/book3s_hv.c: In function 'kvmppc_pseries_do_hcall':
arch/powerpc/kvm/book3s_hv.c:228:30: error: 'H_SUCCESS' undeclared (first use in this function)
arch/powerpc/kvm/book3s_hv.c:232:7: error: 'H_CEDE' undeclared (first use in this function)
arch/powerpc/kvm/book3s_hv.c:234:7: error: 'H_PROD' undeclared (first use in this function)
arch/powerpc/kvm/book3s_hv.c:238:10: error: 'H_PARAMETER' undeclared (first use in this function)
arch/powerpc/kvm/book3s_hv.c:250:7: error: 'H_CONFER' undeclared (first use in this function)
arch/powerpc/kvm/book3s_hv.c:252:7: error: 'H_REGISTER_VPA' undeclared (first use in this function)
make[2]: *** [arch/powerpc/kvm/book3s_hv.o] Error 1

Signed-off-by: Michael Neuling <mikey@neuling.org>
cc: stable@kernel.org (3.1 only)
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc/trace: Add a dummy stack frame for trace_hardirqs_off
Kevin Hao [Thu, 10 Nov 2011 16:04:17 +0000 (16:04 +0000)]
powerpc/trace: Add a dummy stack frame for trace_hardirqs_off

The trace_hardirqs_off will use CALLER_ADDR0 and CALLER_ADDR1.
If an exception occurs in user mode, there is only one stack frame
on the stack and accessing the CALLER_ADDR1 will causes the following
call trace. So we create a dummy stack frame to make
trace_hardirqs_off happy.

WARNING: at kernel/smp.c:459
Modules linked in:
NIP: c0093280 LR: c00930a0 CTR: c0010780
REGS: edb87ae0 TRAP: 0700   Not tainted  (3.1.0)
MSR: 00021002 <ME,CE>  CR: 28002888  XER: 00000000
TASK = edce2ac0[17658] 'mthread-lock-on' THREAD: edb86000 CPU: 5
GPR00: 00000001 edb87b90 edce2ac0 00000005 c0019594 edb87bd8 00000001 00000fe3
GPR08: 00041000 c084138c 4e20120d edb87b90 48002888 1001aa7c 00000000 00000000
GPR16: 48830000 10012a8c 00000000 10000af4 00000001 c0810000 00000000 00000000
GPR24: ee9aa920 c0816a18 00000000 00000005 c0019594 edb87bd8 ee20178c edb87b90
NIP [c0093280] smp_call_function_many+0x214/0x2b4
LR [c00930a0] smp_call_function_many+0x34/0x2b4
Call Trace:
[edb87b90] [c00930a0] smp_call_function_many+0x34/0x2b4 (unreliable)
[edb87bd0] [c00194ec] __flush_tlb_page+0xac/0x100
[edb87c00] [c001957c] flush_tlb_page+0x3c/0x54
[edb87c10] [c00180ac] ptep_set_access_flags+0x74/0x12c
[edb87c40] [c0128068] handle_pte_fault+0x2f0/0x9ac
[edb87cb0] [c0128c3c] handle_mm_fault+0x104/0x1dc
[edb87ce0] [c05f40f4] do_page_fault+0x2dc/0x630
[edb87e50] [c001078c] handle_page_fault+0xc/0x80

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: Copy down exception vectors after feature fixups
Anton Blanchard [Mon, 14 Nov 2011 12:54:47 +0000 (12:54 +0000)]
powerpc: Copy down exception vectors after feature fixups

kdump fails because we try to execute an HV only instruction. Feature
fixups are being applied after we copy the exception vectors down to 0
so they miss out on any updates.

We have always had this issue but it only became critical in v3.0
when we added CFAR support (breaks POWER5) and v3.1 when we added
POWERNV (breaks everyone).

Signed-off-by: Anton Blanchard <anton@samba.org>
Cc: <stable@kernel.org> [v3.0+]
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
12 years agopowerpc: panic if we can't instantiate RTAS
Anton Blanchard [Mon, 14 Nov 2011 12:55:47 +0000 (12:55 +0000)]
powerpc: panic if we can't instantiate RTAS

I had to debug a strange situation where all manner of things were
failing. SMT threads, storage and network were all completely broken.

The root cause was we couldn't find enough memory to instantiate RTAS -
this was a network install so the initrd was huge.

Instead of limping along and failing in mysterious ways we should just
panic up front if RTAS exists and we can't allocate space for it.

Signed-off-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>