8 years agoPM / QoS: add pm_qos_update_request_timeout() API
MyungJoo Ham [Wed, 28 Mar 2012 21:31:24 +0000]
PM / QoS: add pm_qos_update_request_timeout() API

The new API, pm_qos_update_request_timeout() is to provide a timeout
with pm_qos_update_request.

For example, pm_qos_update_request_timeout(req, 100, 1000), means that
QoS request on req with value 100 will be active for 1000 microseconds.
After 1000 microseconds, the QoS request thru req is reset. If there
were another pm_qos_update_request(req, x) during the 1000 us, this
new request with value x will override as this is another request on the
same req handle. A new request on the same req handle will always
override the previous request whether it is the conventional request or
it is the new timeout request.

Signed-off-by: MyungJoo Ham <myungjoo.ham@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Acked-by: Mark Gross <markgross@thegnar.org>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>

8 years agofirmware_class: Move request_firmware_nowait() to workqueues
Stephen Boyd [Wed, 28 Mar 2012 21:31:00 +0000]
firmware_class: Move request_firmware_nowait() to workqueues

Oddly enough a work_struct was already part of the firmware_work
structure but nobody was using it. Instead of creating a new
kthread for each request_firmware_nowait() call just schedule the
work on the system workqueue. This should avoid some overhead
in forking new threads when they're not strictly necessary.

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>

8 years agofirmware_class: Reorganize fw_create_instance()
Stephen Boyd [Wed, 28 Mar 2012 21:30:43 +0000]
firmware_class: Reorganize fw_create_instance()

Recent patches to split up the three phases of request_firmware()
lead to a casting away of const in fw_create_instance(). We can
avoid this cast by splitting up fw_create_instance() a bit.

Make _request_firmware_setup() return a struct fw_priv and use
that struct instead of passing struct firmware to
_request_firmware(). Move the uevent and device file creation
bits to the loading phase and rename the function to
_request_firmware_load() to better reflect its purpose.

Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>

8 years agoPM / Sleep: Mitigate race between the freezer and request_firmware()
Rafael J. Wysocki [Wed, 28 Mar 2012 21:30:28 +0000]
PM / Sleep: Mitigate race between the freezer and request_firmware()

There is a race condition between the freezer and request_firmware()
such that if request_firmware() is run on one CPU and
freeze_processes() is run on another CPU and usermodehelper_disable()
called by it succeeds to grab umhelper_sem for writing before
usermodehelper_read_trylock() called from request_firmware()
acquires it for reading, the request_firmware() will fail and
trigger a WARN_ON() complaining that it was called at a wrong time.
However, in fact, it wasn't called at a wrong time and
freeze_processes() simply happened to be executed simultaneously.

To avoid this race, at least in some cases, modify
usermodehelper_read_trylock() so that it doesn't fail if the
freezing of tasks has just started and hasn't been completed yet.
Instead, during the freezing of tasks, it will try to freeze the
task that has called it so that it can wait until user space is
thawed without triggering the scary warning.

For this purpose, change usermodehelper_disabled so that it can
take three different values, UMH_ENABLED (0), UMH_FREEZING and
UMH_DISABLED.  The first one means that usermode helpers are
enabled, the last one means "hard disable" (i.e. the system is not
ready for usermode helpers to be used) and the second one
is reserved for the freezer.  Namely, when freeze_processes() is
started, it sets usermodehelper_disabled to UMH_FREEZING which
tells usermodehelper_read_trylock() that it shouldn't fail just
yet and should call try_to_freeze() if woken up and cannot
return immediately.  This way all freezable tasks that happen
to call request_firmware() right before freeze_processes() is
started and lose the race for umhelper_sem with it will be
frozen and will sleep until thaw_processes() unsets
usermodehelper_disabled.  [For the non-freezable callers of
request_firmware() the race for umhelper_sem against
freeze_processes() is unfortunately unavoidable.]

Reported-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable@vger.kernel.org

8 years agoPM / Sleep: Move disabling of usermode helpers to the freezer
Rafael J. Wysocki [Wed, 28 Mar 2012 21:30:21 +0000]
PM / Sleep: Move disabling of usermode helpers to the freezer

The core suspend/hibernation code calls usermodehelper_disable() to
avoid race conditions between the freezer and the starting of
usermode helpers and each code path has to do that on its own.
However, it is always called right before freeze_processes()
and usermodehelper_enable() is always called right after
thaw_processes().  For this reason, to avoid code duplication and
to make the connection between usermodehelper_disable() and the
freezer more visible, make freeze_processes() call it and remove the
direct usermodehelper_disable() and usermodehelper_enable() calls
from all suspend/hibernation code paths.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable@vger.kernel.org

8 years agoPM / Hibernate: Disable usermode helpers right before freezing tasks
Rafael J. Wysocki [Wed, 28 Mar 2012 21:30:14 +0000]
PM / Hibernate: Disable usermode helpers right before freezing tasks

There is no reason to call usermodehelper_disable() before creating
memory bitmaps in hibernate() and software_resume(), so call it right
before freeze_processes(), in accordance with the other suspend and
hibernation code.  Consequently, call usermodehelper_enable() right
after the thawing of tasks rather than after freeing the memory

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable@vger.kernel.org

8 years agofirmware_class: Do not warn that system is not ready from async loads
Rafael J. Wysocki [Wed, 28 Mar 2012 21:30:02 +0000]
firmware_class: Do not warn that system is not ready from async loads

If firmware is requested asynchronously, by calling
request_firmware_nowait(), there is no reason to fail the request
(and warn the user) when the system is (presumably temporarily)
unready to handle it (because user space is not available yet or
frozen).  For this reason, introduce an alternative routine for
read-locking umhelper_sem, usermodehelper_read_lock_wait(), that
will wait for usermodehelper_disabled to be unset (possibly with
a timeout) and make request_firmware_work_func() use it instead of

Accordingly, modify request_firmware() so that it uses
usermodehelper_read_trylock() to acquire umhelper_sem and remove
the code related to that lock from _request_firmware().

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable@vger.kernel.org

8 years agofirmware_class: Split _request_firmware() into three functions, v2
Rafael J. Wysocki [Wed, 28 Mar 2012 21:29:55 +0000]
firmware_class: Split _request_firmware() into three functions, v2

Split _request_firmware() into three functions,
_request_firmware_prepare() doing preparatory work that need not be
done under umhelper_sem, _request_firmware_cleanup() doing the
post-error cleanup and _request_firmware() carrying out the remaining

This change is requisite for moving the acquisition of umhelper_sem
from _request_firmware() to the callers, which is going to be done

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Stephen Boyd <sboyd@codeaurora.org>
Cc: stable@vger.kernel.org

8 years agofirmware_class: Rework usermodehelper check
Rafael J. Wysocki [Wed, 28 Mar 2012 21:29:45 +0000]
firmware_class: Rework usermodehelper check

Instead of two functions, read_lock_usermodehelper() and
usermodehelper_is_disabled(), used in combination, introduce
usermodehelper_read_trylock() that will only return with umhelper_sem
held if usermodehelper_disabled is unset (and will return -EAGAIN
otherwise) and make _request_firmware() use it.

Rename read_unlock_usermodehelper() to
usermodehelper_read_unlock() to follow the naming convention of the
new function.

Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: stable@vger.kernel.org

8 years agoPM / Runtime: don't forget to wake up waitqueue on failure
Alan Stern [Mon, 26 Mar 2012 20:46:52 +0000]
PM / Runtime: don't forget to wake up waitqueue on failure

This patch (as1535) fixes a bug in the runtime PM core.  When a
runtime suspend attempt completes, whether successfully or not, the
device's power.wait_queue is supposed to be signalled.  But this
doesn't happen in the failure pathway of rpm_suspend() when another
autosuspend attempt is rescheduled.  As a result, a task can get stuck
indefinitely on the wait queue (I have seen this happen in testing).

The patch fixes the problem by moving the wake_up_all() call up near
the start of the failure code.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
CC: <stable@vger.kernel.org>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>

8 years agolsm_audit: don't specify the audit pre/post callbacks in 'struct common_audit_data'
Linus Torvalds [Mon, 2 Apr 2012 22:48:12 +0000]
lsm_audit: don't specify the audit pre/post callbacks in 'struct common_audit_data'

It just bloats the audit data structure for no good reason, since the
only time those fields are filled are just before calling the
common_lsm_audit() function, which is also the only user of those

So just make them be the arguments to common_lsm_audit(), rather than
bloating that structure that is passed around everywhere, and is
initialized in hot paths.

Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoSELinux: do not allocate stack space for AVC data unless needed
Eric Paris [Tue, 3 Apr 2012 16:38:00 +0000]
SELinux: do not allocate stack space for AVC data unless needed

Instead of declaring the entire selinux_audit_data on the stack when we
start an operation on declare it on the stack if we are going to use it.
We know it's usefulness at the end of the security decision and can declare
it there.

Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoSELinux: remove avd from slow_avc_audit()
Eric Paris [Mon, 2 Apr 2012 17:15:55 +0000]
SELinux: remove avd from slow_avc_audit()

We don't use the argument, so remove it.

Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoSELinux: remove avd from selinux_audit_data
Eric Paris [Mon, 2 Apr 2012 17:15:50 +0000]
SELinux: remove avd from selinux_audit_data

We do not use it.  Remove it.

Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoLSM: shrink the common_audit_data data union
Eric Paris [Mon, 2 Apr 2012 17:15:44 +0000]
LSM: shrink the common_audit_data data union

After shrinking the common_audit_data stack usage for private LSM data I'm
not going to shrink the data union.  To do this I'm going to move anything
larger than 2 void * ptrs to it's own structure and require it to be declared
separately on the calling stack.  Thus hot paths which don't need more than
a couple pointer don't have to declare space to hold large unneeded
structures.  I could get this down to one void * by dealing with the key
struct and the struct path.  We'll see if that is helpful after taking care of

Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoLSM: shrink sizeof LSM specific portion of common_audit_data
Eric Paris [Tue, 3 Apr 2012 16:37:02 +0000]
LSM: shrink sizeof LSM specific portion of common_audit_data

Linus found that the gigantic size of the common audit data caused a big
perf hit on something as simple as running stat() in a loop.  This patch
requires LSMs to declare the LSM specific portion separately rather than
doing it in a union.  Thus each LSM can be responsible for shrinking their
portion and don't have to pay a penalty just because other LSMs have a
bigger space requirement.

Signed-off-by: Eric Paris <eparis@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoregulator: Remove non-existent parameter from fixed-helper.c kernel doc
Russ Dill [Thu, 22 Mar 2012 05:19:45 +0000]
regulator: Remove non-existent parameter from fixed-helper.c kernel doc

Signed-off-by: Russ Dill <russ.dill@ti.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Fix setting new voltage in s5m8767_set_voltage
Axel Lin [Mon, 19 Mar 2012 04:22:10 +0000]
regulator: Fix setting new voltage in s5m8767_set_voltage

Current code does not really update the register with new value, fix it.
I rename the variable i to sel for better readability.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked-by: Sangbeom Kim <sbkim73@samsung.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: fix sysfs name collision between dummy and fixed dummy regulator
Shawn Guo [Sat, 31 Mar 2012 08:33:31 +0000]
regulator: fix sysfs name collision between dummy and fixed dummy regulator

When regulator_register_fixed() is being used to register fixed dummy
regulator, the following line will be seen in the boot log.  And the
sysfs entry for fixed dummy regulator is not shown.

  dummy: Failed to create debugfs directory

The patch renames the fixed dummy supply to "fixed-dummy" to avoid
the name collision with dummy regulator.

Signed-off-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Fix deadlock on removal of regulators with supplies
Mark Brown [Wed, 28 Mar 2012 20:17:55 +0000]
regulator: Fix deadlock on removal of regulators with supplies

If a regulator with a supply is being unregistered we will call
regulator_put() to release the supply with the regulator_list_mutex held
but this deadlocks as regulator_put() takes the same lock. Fix this by
releasing the supply before we take the mutex in regulator_unregister().

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Fix comments in include/linux/regulator/machine.h
Axel Lin [Thu, 29 Mar 2012 04:21:17 +0000]
regulator: Fix comments in include/linux/regulator/machine.h

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Only update [LDOx|DCx]_HIB_MODE bits in wm8350_[ldo|dcdc]_set_suspend_disable
Axel Lin [Thu, 29 Mar 2012 02:47:36 +0000]
regulator: Only update [LDOx|DCx]_HIB_MODE bits in wm8350_[ldo|dcdc]_set_suspend_disable

What we want is to disable output by setting [LDOx|DCx]_HIB_MODE bits.
Current code also clears other bits in LDOx/DCDCx Low Power register.

R202 (CAh) LDO1 Low Power
BIT[13:12] LDO1 Hibernate behaviour:
00 = Select voltage image settings
01 = disable output
10 = reserved
11 = reserved

R182 (B6h) DCDC1 Low Power
BIT[14:12] DC-DC1 Hibernate behaviour:
000 = Use current settings (no change)
001 = Select voltage image settings
010 = Force standby mode
011 = Force standby mode and voltage image settings.
100 = Force LDO mode
101 = Force LDO mode and voltage image settings.
110 = Reserved.
111 = Disable output

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Fix setting low power mode for wm831x aldo
Axel Lin [Thu, 29 Mar 2012 07:02:55 +0000]
regulator: Fix setting low power mode for wm831x aldo

Set BIT[8] of LDO7 ON Control mode for low power mode.

R16507 (407Bh)  LDO7 ON Control
BIT[8] LDO7_ON_MODE: LDO7 ON Operating Mode
        0 = Normal mode
        1 = Low Power mode

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Return microamps in wm8350_isink_get_current
Axel Lin [Tue, 27 Mar 2012 07:21:45 +0000]
regulator: Return microamps in wm8350_isink_get_current

The values in isink_cur array are microamps.
The regulator core expects get_current_limit callback to return microamps.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: wm8350: Fix the logic to choose best current limit setting
Axel Lin [Tue, 27 Mar 2012 07:20:08 +0000]
regulator: wm8350: Fix the logic to choose best current limit setting

Current implementation in get_isink_val actually choose the biggest current
limit setting falls within the specified range.
What we want is to choose the smallest current limit setting falls within the
specified range. Fix it.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: wm831x-isink: Fix the logic to choose best current limit setting
Axel Lin [Tue, 27 Mar 2012 07:18:53 +0000]
regulator: wm831x-isink: Fix the logic to choose best current limit setting

Current code in wm831x_isink_set_current actually set the current limit setting
smaller than specified range.

Fix the logic in wm831x_isink_set_current to choose the smallest current limit
setting falls within the specified range.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: wm831x-dcdc: Fix the logic to choose best current limit setting
Axel Lin [Tue, 27 Mar 2012 07:17:26 +0000]
regulator: wm831x-dcdc: Fix the logic to choose best current limit setting

Current code in wm831x_buckv_set_current_limit actually set the current limit
setting greater than specified range.

Fix the logic in wm831x_buckv_set_current_limit to choose the
smallest current limit setting falls within the specified range.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: anatop: patching to device-tree property "reg".
Ying-Chun Liu (PaulLiu) [Tue, 27 Mar 2012 07:54:01 +0000]
regulator: anatop: patching to device-tree property "reg".

Change "reg" to "anatop-reg-offset" due to there is a warning of handling no
size field in reg.

This patch also adds the missing device-tree binding documentation.

Signed-off-by: Ying-Chun Liu (PaulLiu) <paul.liu@linaro.org>
Acked-by: Shawn Guo <shawn.guo@linaro.org>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Do proper shift to set correct bit for DC[2|5]_HIB_MODE setting
Axel Lin [Thu, 22 Mar 2012 22:27:10 +0000]
regulator: Do proper shift to set correct bit for DC[2|5]_HIB_MODE setting

DC[2|5]_HIB_MODE is BIT 12 of DCDC[2|5] Control register.
WM8350_DC2_HIB_MODE_ACTIVE/WM8350_DC2_HIB_MODE_DISABLE are defined as 1/0.
Thus we need to left shift WM8350_DC2_HIB_MODE_SHIFT bits.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Fix restoring pmic.dcdcx_hib_mode settings in wm8350_dcdc_set_suspend_enable
Axel Lin [Thu, 22 Mar 2012 22:25:05 +0000]
regulator: Fix restoring pmic.dcdcx_hib_mode settings in wm8350_dcdc_set_suspend_enable

What we want is to restore wm8350->pmic.dcdcx_hib_mode settings to
WM8350_DCDCx_LOW_POWER registers. Current code also clears all other
bits of WM8350_DCDCx_LOW_POWER registers which is wrong.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Fix unbalanced lock/unlock in mc13892_regulator_probe error path
Axel Lin [Thu, 15 Mar 2012 04:49:09 +0000]
regulator: Fix unbalanced lock/unlock in mc13892_regulator_probe error path

We do not hold a lock while registering regulator, thus should not call unlock
if regulator_register fails.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: Fix set and get current limit for wm831x_buckv
Axel Lin [Thu, 15 Mar 2012 07:53:05 +0000]
regulator: Fix set and get current limit for wm831x_buckv

WM831X_DC1_HC_THR_MASK is 0x0070.
We need to do proper shift for setting and getting current limit.

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoregulator: tps6586x: Fix list minimal voltage setting for LDO0
Axel Lin [Sat, 24 Mar 2012 01:37:30 +0000]
regulator: tps6586x: Fix list minimal voltage setting for LDO0

According to the datasheet, LDO0 has minimal voltage 1.2V rather than 1.25V.

Table 3-39. VLDO0[2:0] Settings
VLDOx[2:0] VOUT (V) VLDOx[2:0] VOUT (V)
000        1.20     100        2.70
001        1.50     101        2.85
010        1.80     110        3.10
011        2.50     111        3.30

Signed-off-by: Axel Lin <axel.lin@gmail.com>
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>

8 years agoperf/x86/p4: Add format attributes
Peter Zijlstra [Tue, 27 Mar 2012 14:50:42 +0000]
perf/x86/p4: Add format attributes

Steven reported his P4 not booting properly, the missing format
attributes cause a NULL ptr deref. Cure this by adding the
missing format specification.

I took the format description out of the comment near
p4_config_pack*() and hope that comment is still relatively

Reported-by: Steven Rostedt <rostedt@goodmis.org>
Reported-by: Bruno Prémont <bonbons@linux-vserver.org>
Tested-by: Steven Rostedt <rostedt@goodmis.org>
Signed-off-by: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Cyrill Gorcunov <gorcunov@openvz.org>
Cc: Lin Ming <ming.m.lin@intel.com>
Cc: Stephane Eranian <eranian@google.com>
Link: http://lkml.kernel.org/r/1332859842.16159.227.camel@twins
Signed-off-by: Ingo Molnar <mingo@kernel.org>

8 years agotracing, sched, vfs: Fix 'old_pid' usage in trace_sched_process_exec()
Oleg Nesterov [Fri, 30 Mar 2012 16:26:36 +0000]
tracing, sched, vfs: Fix 'old_pid' usage in trace_sched_process_exec()

1. TRACE_EVENT(sched_process_exec) forgets to actually use the
   old pid argument, it sets ->old_pid = p->pid.

2. search_binary_handler() uses the wrong pid number. tracepoint
   needs the global pid_t from the root namespace, while old_pid
   is the virtual pid number as it seen by the tracer/parent.

With this patch we have two pid_t's in search_binary_handler(),
not really nice. Perhaps we should switch to "struct pid*", but
in this case it would be better to cleanup the current code
first and move the "depth == 0" code outside.

Signed-off-by: Oleg Nesterov <oleg@redhat.com>
Cc: David Smith <dsmith@redhat.com>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Denys Vlasenko <dvlasenk@redhat.com>
Link: http://lkml.kernel.org/r/20120330162636.GA4857@redhat.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>

8 years agox86, kvm: Call restore_sched_clock_state() only after %gs is initialized
Marcelo Tosatti [Sun, 1 Apr 2012 16:53:36 +0000]
x86, kvm: Call restore_sched_clock_state() only after %gs is initialized

s2ram broke due to this KVM commit:

  b74f05d61b73 x86: kvmclock: abstract save/restore sched_clock_state

restore_sched_clock_state() methods use percpu data, therefore
they must run after %gs is initialized, but before mtrr_bp_restore()
(due to lockstat using sched_clock).

Move it to the correct place.

Reported-and-tested-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com>
Cc: Avi Kivity <avi@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>

8 years agox86: Use -mno-avx when available
Andi Kleen [Wed, 28 Mar 2012 18:51:17 +0000]
x86: Use -mno-avx when available

On gccs that support AVX it's a good idea to disable that too, similar to
how SSE2, SSE1 etc. are already disabled. This prevents the compiler
from generating AVX ever implicitely.

No failure observed, just from review.

[ hpa: Marking this for urgent and stable, simply because the patch
  will either have absolutely no effect *or* it will avoid potentially
  very hard to debug failures. ]

Signed-off-by: Andi Kleen <ak@linux.intel.com>
Link: http://lkml.kernel.org/r/1332960678-11879-1-git-send-email-andi@firstfloor.org
Signed-off-by: H. Peter Anvin <hpa@zytor.com>
Cc: <stable@vger.kernel.org>

8 years agox86: Remove the ancient and deprecated disable_hlt() and enable_hlt() facility
Len Brown [Thu, 29 Mar 2012 21:49:17 +0000]
x86: Remove the ancient and deprecated disable_hlt() and enable_hlt() facility

The X86_32-only disable_hlt/enable_hlt mechanism was used by the
32-bit floppy driver. Its effect was to replace the use of the
HLT instruction inside default_idle() with cpu_relax() - essentially
it turned off the use of HLT.

This workaround was commented in the code as:

 "disable hlt during certain critical i/o operations"

 "This halt magic was a workaround for ancient floppy DMA
  wreckage. It should be safe to remove."

H. Peter Anvin additionally adds:

 "To the best of my knowledge, no-hlt only existed because of
  flaky power distributions on 386/486 systems which were sold to
  run DOS.  Since DOS did no power management of any kind,
  including HLT, the power draw was fairly uniform; when exposed
  to the much hhigher noise levels you got when Linux used HLT
  caused some of these systems to fail.

  They were by far in the minority even back then."

Alan Cox further says:

 "Also for the Cyrix 5510 which tended to go castors up if a HLT
  occurred during a DMA cycle and on a few other boxes HLT during
  DMA tended to go astray.

  Do we care ? I doubt it. The 5510 was pretty obscure, the 5520
  fixed it, the 5530 is probably the oldest still in any kind of

So, let's finally drop this.

Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Josh Boyer <jwboyer@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Acked-by: "H. Peter Anvin" <hpa@zytor.com>
Acked-by: Alan Cox <alan@lxorguk.ukuu.org.uk>
Cc: Stephen Hemminger <shemminger@vyatta.com
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: <stable@kernel.org>
Link: http://lkml.kernel.org/n/tip-3rhk9bzf0x9rljkv488tloib@git.kernel.org
[ If anyone cares then alternative instruction patching could be
  used to replace HLT with a one-byte NOP instruction. Much simpler. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>

8 years agox86: Preserve lazy irq disable semantics in fixup_irqs()
Liu, Chuansheng [Mon, 26 Mar 2012 07:11:50 +0000]
x86: Preserve lazy irq disable semantics in fixup_irqs()

The default irq_disable() sematics are to mark the interrupt disabled,
but keep it unmasked. If the interrupt is delivered while marked
disabled, the low level interrupt handler masks it and marks it
pending. This is important for detecting wakeup interrupts during
suspend and for edge type interrupts to avoid losing interrupts.

fixup_irqs() moves the interrupts away from an offlined cpu. For
certain interrupt types it needs to mask the interrupt line before
changing the affinity. After affinity has changed the interrupt line
is unmasked again, but only if it is not marked disabled.

This breaks the lazy irq disable semantics and causes problems in
suspend as the interrupt can be lost or wakeup functionality is

Check irqd_irq_masked() instead of irqd_irq_disabled() because
irqd_irq_masked() is only set, when the core code actually masked the
interrupt line. If it's not set, we unmask the interrupt and let the
lazy irq disable logic deal with an eventually incoming interrupt.

[ tglx: Massaged changelog and added a comment ]

Signed-off-by: liu chuansheng <chuansheng.liu@intel.com>
Cc: Yanmin Zhang <yanmin_zhang@linux.intel.com>
Link: http://lkml.kernel.org/r/27240C0AC20F114CBF8149A2696CBE4A05DFB3@SHSMSX101.ccr.corp.intel.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

8 years agohwmon: (ad7314) Adds missing spi_dev initialization
Graeme Smecher [Tue, 3 Apr 2012 23:42:21 +0000]
hwmon: (ad7314) Adds missing spi_dev initialization

This driver was recently moved from IIO (where it worked) to hwmon (where
it doesn't.) This breakage occured because the hwmon version neglected to
correctly initialize a reference to spi_dev in its drvdata. The result is a
segfault every time the temperature is queried.

Signed-off-by: Graeme Smecher <gsmecher@threespeedlogic.com>
Cc: stable@vger.kernel.org # 3.2+
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>

8 years agohwmon: (k10temp) Add support for AMD Trinity CPUs
Andre Przywara [Fri, 30 Mar 2012 20:48:20 +0000]
hwmon: (k10temp) Add support for AMD Trinity CPUs

The on-chip northbridge's temperature sensor of the upcoming
AMD Trinity CPUs works the same as for the previous CPUs.
Since it has a different PCI-ID, we just add the new one to the list
supported by k10temp.
This allows to use the k10temp driver on those CPUs.

Signed-off-by: Andre Przywara <andre.przywara@amd.com>
Cc: stable@vger.kernel.org # 3.0+
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>

8 years agohwmon: (w83627ehf) mark const init data with __initconst instead of __initdata
Uwe Kleine-König [Fri, 30 Mar 2012 20:04:55 +0000]
hwmon: (w83627ehf) mark const init data with __initconst instead of __initdata

As long as there is no other non-const variable marked __initdata in the
same compilation unit it doesn't hurt. If there were one however
compilation would fail with

error: $variablename causes a section type conflict

because a section containing const variables is marked read only and so
cannot contain non-const variables.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: Jean Delvare <khali@linux-fr.org>
Cc: Guenter Roeck <guenter.roeck@ericsson.com>
Cc: lm-sensors@lm-sensors.org
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>

8 years agohwmon: (acpi_power_meter) fix lockdep spew due to non-static lock class
Kyle McMartin [Wed, 28 Mar 2012 19:11:47 +0000]
hwmon: (acpi_power_meter) fix lockdep spew due to non-static lock class

Similar to a30dcb4f which fixed asus_atk0110.ko, I recently received a
bug report from someone hitting the same issue in acpi_power_meter.

[   13.963168] power_meter ACPI000D:00: Found ACPI power meter.
[   13.963900] BUG: key ffff8802161f3920 not in .data!
[   13.963904] ------------[ cut here ]------------
[   13.963915] WARNING: at kernel/lockdep.c:2986

So let's fix that up for them by statically declaring the

Signed-off-by: Kyle McMartin <kyle@redhat.com>
Cc: stable@vger.kernel.org # 3.0+
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>

8 years agohwmon: (adm1031) Fix compiler warning
Guenter Roeck [Sat, 24 Mar 2012 15:51:05 +0000]
hwmon: (adm1031) Fix compiler warning

Some configurations produce the following compile warning:

drivers/hwmon/adm1031.c: In function 'set_fan_auto_channel':
drivers/hwmon/adm1031.c:292: warning: 'reg' may be used uninitialized in this function

While this is a false positive, it can easily be fixed by overloading the return
value from get_fan_auto_nearest with both register value and error return code
(the register value is never negative). Coincidentially, that also reduces
module size by a few bytes.

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Jean Delvare <khali@linux-fr.org>

8 years agohwmon: (f75375s) Fix warning message seen in some configurations
Guenter Roeck [Sat, 24 Mar 2012 15:38:21 +0000]
hwmon: (f75375s) Fix warning message seen in some configurations

In some configurations, BUG() does not result in an endless loop but returns to
the caller. This results in the following compiler warning:

drivers/hwmon/f75375s.c: In function 'duty_mode_enabled':
drivers/hwmon/f75375s.c:280: warning: control reaches end of non-void function
drivers/hwmon/f75375s.c: In function 'auto_mode_enabled':
drivers/hwmon/f75375s.c:295: warning: control reaches end of non-void function

Fix the warning by returning something sensible after BUG().

Cc: Nikolaus Schulz <schulz@macnetix.de>
Cc: Riku Voipio <riku.voipio@iki.fi>
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Jean Delvare <khali@linux-fr.org>
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>

8 years agohwmon: (max6639) Convert to dev_pm_ops
Mark Brown [Thu, 22 Mar 2012 20:23:58 +0000]
hwmon: (max6639) Convert to dev_pm_ops

The I2C specific PM operations have been deprecated and printing a
warning on boot for over a year now.

Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
[guenter.roeck@ericsson.com: Added missing #ifdef around pm functions]
Cc: stable@vger.kernel.org # 3.0+
Signed-off-by: Guenter Roeck <guenter.roeck@ericsson.com>

8 years agohwmon: (sht15) Fix Kconfig dependencies
Guenter Roeck [Wed, 21 Mar 2012 03:06:35 +0000]
hwmon: (sht15) Fix Kconfig dependencies

sht15 depends on GPIOLIB, not on GENERIC_GPIO.

This fixes the following build error, seen if GPIOLIB is not defined:

src/drivers/hwmon/sht15.c: error: implicit declaration of function 'gpio_direction_input': => 293:2
src/drivers/hwmon/sht15.c: error: implicit declaration of function 'gpio_direction_output': => 216:2
src/drivers/hwmon/sht15.c: error: implicit declaration of function 'gpio_free': => 1000:2
src/drivers/hwmon/sht15.c: error: implicit declaration of function 'gpio_get_value': => 296:2
src/drivers/hwmon/sht15.c: error: implicit declaration of function 'gpio_request': => 946:2
src/drivers/hwmon/sht15.c: error: implicit declaration of function 'gpio_set_value': => 218:2
src/drivers/hwmon/sht15.c: error: implicit declaration of function 'gpio_to_irq': => 514:2

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Jean Delvare <khali@linux-fr.org>

8 years agohwmon: (gpio-fan) Fix Kconfig dependencies
Guenter Roeck [Wed, 21 Mar 2012 03:06:34 +0000]
hwmon: (gpio-fan) Fix Kconfig dependencies

gpio-fan depends on GPIOLIB, not on GENERIC_GPIO.

This fixes the following build error, seen if GPIOLIB is not defined:

src/drivers/hwmon/gpio-fan.c: error: implicit declaration of function 'gpio_direction_output': => 372:3
src/drivers/hwmon/gpio-fan.c: error: implicit declaration of function 'gpio_free': => 130:2
src/drivers/hwmon/gpio-fan.c: error: implicit declaration of function 'gpio_get_value': => 79:2
src/drivers/hwmon/gpio-fan.c: error: implicit declaration of function 'gpio_request': => 98:2
src/drivers/hwmon/gpio-fan.c: error: implicit declaration of function 'gpio_set_value': => 156:3
src/drivers/hwmon/gpio-fan.c: error: implicit declaration of function 'gpio_to_irq': => 114:2

Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Acked-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Simon Guinot <sguinot@lacie.com>

8 years agoMCE, AMD: Drop too granulary family model checks
Borislav Petkov [Wed, 4 Apr 2012 12:21:02 +0000]
MCE, AMD: Drop too granulary family model checks

MCA details seldom change inbetween the models of a family so don't
be too conservative and enable decoding on everything starting from
K8 onwards. Minor adjustments can come in later but most importantly,
we have some decoding infrastructure in place for upcoming models by

Signed-off-by: Borislav Petkov <borislav.petkov@amd.com>

8 years agomd/raid1,raid10: don't compare excess byte during consistency check.
NeilBrown [Sun, 1 Apr 2012 15:39:05 +0000]
md/raid1,raid10: don't compare excess byte during consistency check.

When comparing two pages read from different legs of a mirror, only
compare the bytes that were read, not the whole page.

In most cases we read a whole page, but in some cases with
bad blocks or odd sizes devices we might read fewer than that.

This bug has been present "forever" but at worst it might cause
a report of two many mismatches and generate a little bit
extra resync IO, so there is no need to back-port to -stable

Reported-by: majianpeng <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>

8 years agomd/raid5: Fix a bug about judging if the operation is syncing or replacing
majianpeng [Sun, 1 Apr 2012 15:16:59 +0000]
md/raid5: Fix a bug about judging if the operation is syncing or replacing

When create a raid5 using assume-clean and echo check or repair to
sync_action.Then component disks did not operated IO but the raid
check/resync faster than normal.
Because the judgement in function analyse_stripe():
if (do_recovery ||
    sh->sector >= conf->mddev->recovery_cp)
s->syncing = 1;
s->replacing = 1;
When check or repair,the recovery_cp == MaxSectore,so syncing equal zero
not one.

This bug was introduced by commit 9a3e1101b827
    md/raid5:  detect and handle replacements during recovery.
so this patch is suitable for 3.3-stable.

Cc: stable@vger.kernel.org
Signed-off-by: majianpeng <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>

8 years agomd/raid1:Remove unnecessary rcu_dereference(conf->mirrors[i].rdev).
majianpeng [Sun, 1 Apr 2012 15:04:19 +0000]
md/raid1:Remove unnecessary rcu_dereference(conf->mirrors[i].rdev).

Because rde->nr_pending > 0,so can not remove this disk.
And in any case, we aren't holding rcu_read_lock()

Signed-off-by: majianpeng <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>

8 years agomd: Avoid OOPS when reshaping raid1 to raid0
Jes Sorensen [Sun, 1 Apr 2012 13:48:38 +0000]
md: Avoid OOPS when reshaping raid1 to raid0

raid1 arrays do not have the notion of chunk size. Calculate the
largest chunk sector size we can use to avoid a divide by zero OOPS
when aligning the size of the new array to the chunk size.

Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
Signed-off-by: NeilBrown <neilb@suse.de>

8 years agomd/raid5: fix handling of bad blocks during recovery.
NeilBrown [Sun, 1 Apr 2012 13:48:38 +0000]
md/raid5: fix handling of bad blocks during recovery.

1/ We can only treat a known-bad-block like a read-error if we
   have the data that belongs in that block.  So fix that test.

2/ If we cannot recovery a stripe due to insufficient data,
   don't tell "md_done_sync" that the sync failed unless we really
   did fail something.  If we successfully record bad blocks,
   that is success.

Reported-by: "majianpeng" <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>

8 years agomd/raid1: If md_integrity_register() failed,run() must free the mem
majianpeng [Sun, 1 Apr 2012 23:48:38 +0000]
md/raid1: If md_integrity_register() failed,run() must free the mem

Signed-off-by: majianpeng <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>

8 years agomd/raid0: If md_integrity_register() fails, raid0_run() must free the mem.
majianpeng [Sun, 1 Apr 2012 23:48:37 +0000]
md/raid0: If md_integrity_register() fails, raid0_run() must free the mem.

Signed-off-by: majianpeng <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>

8 years agomd/linear: If md_integrity_register() fails, linear_run() must free the mem.
majianpeng [Sun, 1 Apr 2012 23:48:37 +0000]
md/linear: If md_integrity_register() fails, linear_run() must free the mem.

Signed-off-by: majianpeng <majianpeng@gmail.com>
Signed-off-by: NeilBrown <neilb@suse.de>

8 years agoARM: fix more fallout from 9f97da78bf (Disintegrate asm/system.h for ARM)
Russell King [Sat, 31 Mar 2012 08:07:30 +0000]
ARM: fix more fallout from 9f97da78bf (Disintegrate asm/system.h for ARM)

arch/arm/kernel/io.c: In function '_memcpy_toio':
arch/arm/kernel/io.c:29: error: implicit declaration of function 'outer_sync'
arch/arm/mach-omap2/omap_l3_noc.c: In function 'l3_interrupt_handler':
arch/arm/mach-omap2/omap_l3_noc.c:100: error: implicit declaration of function 'outer_sync'
kernel/irq/generic-chip.c: In function 'irq_gc_mask_disable_reg':
kernel/irq/generic-chip.c:45: error: implicit declaration of function 'outer_sync'
kernel/sched/rt.c: In function 'rt_set_overload':
kernel/sched/rt.c:248: error: implicit declaration of function 'outer_sync'

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

8 years agoARM: fix bios32.c build warning
Russell King [Fri, 30 Mar 2012 20:03:54 +0000]
ARM: fix bios32.c build warning

arch/arm/kernel/bios32.c: In function 'pcibios_fixup_bus':
arch/arm/kernel/bios32.c:302: warning: unused variable 'root'

caused by 9f786d033 (arm/PCI: get rid of device resource fixups)

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

8 years agoARM: 7337/1: ptrace: fix ptrace_read_user for !CONFIG_MMU platforms
Will Deacon [Tue, 21 Feb 2012 11:29:03 +0000]
ARM: 7337/1: ptrace: fix ptrace_read_user for !CONFIG_MMU platforms

Commit 68b7f715 ("nommu: ptrace support") added definitions for
PT_TEXT_ADDR and friends, as well as adding ptrace support for reading
from these magic offsets.

Unfortunately, this has probably never worked, since ptrace_read_user
predicates reading on off < sizeof(struct user), returning -EIO

This patch moves the offset size check until after we have tried to
match it against either a magic value or an offset into pt_regs.

Cc: Paul Brook <paul@codesourcery.com>
Signed-off-by: Will Deacon <will.deacon@arm.com>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

8 years agoARM: fix missing bug.h include in arch/arm/kernel/insn.c
Russell King [Fri, 30 Mar 2012 10:51:46 +0000]
ARM: fix missing bug.h include in arch/arm/kernel/insn.c

arch/arm/kernel/insn.c: In function '__arm_gen_branch_thumb2':
arch/arm/kernel/insn.c:13: error: implicit declaration of function 'WARN_ON_ONCE'

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

8 years agoARM: sa11x0: fix build errors from DMA engine API updates
Russell King [Fri, 30 Mar 2012 10:44:15 +0000]
ARM: sa11x0: fix build errors from DMA engine API updates

The recent merge of the sa11x0 code into mainline had silent conflicts
with further development of the DMA engine API, leading to build errors
and warnings:

drivers/net/irda/sa1100_ir.c: In function 'sa1100_irda_dma_start':
drivers/net/irda/sa1100_ir.c:151: error: too few arguments to function 'chan->device->device_prep_slave_sg'
drivers/dma/sa11x0-dma.c: In function 'sa11x0_dma_probe':
drivers/dma/sa11x0-dma.c:950: warning: assignment from incompatible pointer type

Fix these.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

8 years agoserial/sunzilog: fix keyboard on SUN SPARCstation
Sam Ravnborg [Wed, 4 Apr 2012 07:35:06 +0000]
serial/sunzilog: fix keyboard on SUN SPARCstation

The keyboard on my SUN SPARCstation 5 no longer worked.

    The culprint was: d4e33fac2408d37f7b52e80ca2a89f9fb482914f
    ("serial: Kill off NO_IRQ")

Fix up logic for no irq / irq so the keyboard works again.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Cc: Alan Cox <alan@linux.intel.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agosparc: pgtable_64: change include order
Aaro Koskinen [Sun, 1 Apr 2012 08:54:38 +0000]
sparc: pgtable_64: change include order

Fix the following build breakage in v3.4-rc1:

CC arch/sparc/kernel/cpu.o
In file included from /home/aaro/git/linux/arch/sparc/include/asm/pgtable_64.h:15:0,
                 from /home/aaro/git/linux/arch/sparc/include/asm/pgtable.h:4,
                 from arch/sparc/kernel/cpu.c:15:
include/asm-generic/pgtable-nopud.h:13:16: error: unknown type name 'pgd_t'
include/asm-generic/pgtable-nopud.h:25:28: error: unknown type name 'pgd_t'
include/asm-generic/pgtable-nopud.h:26:27: error: unknown type name 'pgd_t'
include/asm-generic/pgtable-nopud.h:27:31: error: unknown type name 'pgd_t'
include/asm-generic/pgtable-nopud.h:28:30: error: unknown type name 'pgd_t'
include/asm-generic/pgtable-nopud.h:38:34: error: unknown type name 'pgd_t'
In file included from /home/aaro/git/linux/arch/sparc/include/asm/pgtable_64.h:783:0,
                 from /home/aaro/git/linux/arch/sparc/include/asm/pgtable.h:4,
                 from arch/sparc/kernel/cpu.c:15:
include/asm-generic/pgtable.h: In function 'pgd_none_or_clear_bad':
include/asm-generic/pgtable.h:258:2: error: implicit declaration of function 'pgd_none' [-Werror=implicit-function-declaration]
include/asm-generic/pgtable.h:260:2: error: implicit declaration of function 'pgd_bad' [-Werror=implicit-function-declaration]
include/asm-generic/pgtable.h: In function 'pud_none_or_clear_bad':
include/asm-generic/pgtable.h:269:6: error: request for member 'pgd' in something not a structure or union

Reported-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: David S. Miller <davem@davemloft.net>

8 years agoavr32: fix nop compile fails from system.h split up
Paul Gortmaker [Tue, 3 Apr 2012 23:14:04 +0000]
avr32: fix nop compile fails from system.h split up

To fix:

  In file included from kernel/exit.c:61:
  arch/avr32/include/asm/mmu_context.h: In function 'enable_mmu':
  arch/avr32/include/asm/mmu_context.h:135: error: implicit declaration of function 'nop'

It needs an include of the new file created in commit ae4739465866
("Disintegrate asm/system.h for AVR32"), but since that file only
contains "nop", and since other arch already have precedent of putting
nop in asm/barrier.h we should just delete the new file and put nop in

Suggested-and-acked-by: David Howells <dhowells@redhat.com>
Cc: Haavard Skinnemoen <hskinnemoen@gmail.com>
Cc: Hans-Christian Egtvedt <egtvedt@samfundet.no>
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agoRevert "nouveau/bios: Fix tracking of BIOS image data"
Linus Torvalds [Wed, 4 Apr 2012 15:16:25 +0000]
Revert "nouveau/bios: Fix tracking of BIOS image data"

This reverts commit d06221c0617ab6d0bc41c4980cefdd9c8cc9a1c1.

It turns out to trigger the "BUG_ON(!PageCompound(page))" in kfree(),
apparently because the code ends up trying to free somethng that was
never kmalloced in the first place.

BenH points out that the patch was untested and wasn't meant to go into
the upstream kernel that quickly in the first place.


Reported-by: Meelis Roos <mroos@linux.ee>
Requested-by: Dave Airlie <airlied@gmail.com>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

8 years agodrm/i915: treat src w & h as fixed point in sprite handling code
Jesse Barnes [Tue, 20 Mar 2012 17:59:09 +0000]
drm/i915: treat src w & h as fixed point in sprite handling code

This was missed when we converted the source values to 16.16 fixed point.

Cc: stable@vger.kernel.org
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: no-lvds quirk on MSI DC500
Anisse Astier [Wed, 7 Mar 2012 17:36:35 +0000]
drm/i915: no-lvds quirk on MSI DC500

This hardware doesn't have an LVDS, it's a desktop box. Fix incorrect
LVDS detection.

Cc: stable@kernel.org
Signed-off-by: Anisse Astier <anisse@astier.eu>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: Add lock on drm_helper_resume_force_mode
Sean Paul [Fri, 23 Mar 2012 12:52:58 +0000]
drm/i915: Add lock on drm_helper_resume_force_mode

i915_drm_thaw was not locking the mode_config lock when calling
drm_helper_resume_force_mode. When there were multiple wake sources,
this caused FDI training failure on SNB which in turn corrupted the

Signed-off-by: Sean Paul <seanpaul@chromium.org>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: don't leak struct_mutex lock on ppgtt init failures
Daniel Vetter [Mon, 2 Apr 2012 21:33:03 +0000]
drm/i915: don't leak struct_mutex lock on ppgtt init failures

Reported-by: Konstantin Belousov <kostikbel@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: disable ppgtt on snb when dmar is enabled
Daniel Vetter [Mon, 2 Apr 2012 08:08:35 +0000]
drm/i915: disable ppgtt on snb when dmar is enabled

Totally unexpected that this regressed. Luckily it sounds like we just
need to have dmar disable on the igfx, not the entire system. At least
that's what a few days of testing between Tony Vroon and me indicates.

Reported-by: Tony Vroon <tony@linx.net>
Cc: Tony Vroon <tony@linx.net>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=43024
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: add Ivy Bridge GT2 Server entries
Eugeni Dodonov [Thu, 29 Mar 2012 23:55:48 +0000]
drm/i915: add Ivy Bridge GT2 Server entries

This adds PCI ID for IVB GT2 server variant which we were missing.

Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
[danvet: fix up conflict because the patch has been diffed against next. tsk.]
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: properly clear SSC1 bit in the pch refclock init code
Daniel Vetter [Fri, 30 Mar 2012 20:14:05 +0000]
drm/i915: properly clear SSC1 bit in the pch refclock init code

Noticed by staring at intel_reg_dumper diffs. Unfortunately it does
not seem to completely fix the bug.

Still, it's good to get this right, and maybe it helps someplace else.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=47117
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: apply CS reg readback trick against missed IRQ on snb
Daniel Vetter [Tue, 27 Mar 2012 07:31:24 +0000]
drm/i915: apply CS reg readback trick against missed IRQ on snb

Ben Widawsky reported missed IRQ issues and this patch here helps.

We have one other missed IRQ report still left on snb, reported by QA:


This is _not_ a regression due to the forcewake voodoo though, it
started showing up before that was applied and has been on-and-off for
the past few weeks. According to QA this patch does not help. But the
missed IRQ is always from the blt ring (despite running piglit, so
also render activity expected), so I'm hopefully that this is an issue
with the blt ring itself.

Tested-by: Ben Widawsky <ben@bwidawsk.net>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: quirk away broken OpRegion VBT
Daniel Vetter [Sat, 24 Mar 2012 22:51:30 +0000]
drm/i915: quirk away broken OpRegion VBT

Somehow the BIOS manages to screw things up when copying the VBT
around, because the one we scrap from the VBIOS rom actually works.

Cc: stable@kernel.org
Tested-by: Markus Heinz <markus.heinz@uni-dortmund.de>
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=28812
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: enable plain RC6 on Sandy Bridge by default
Eugeni Dodonov [Fri, 23 Mar 2012 14:57:19 +0000]
drm/i915: enable plain RC6 on Sandy Bridge by default

This is yet another chapter in the ongoing saga of bringing RC6 to Sandy
Bridge machines by default.

Now that we have discovered that RC6 issues are triggered by RC6+ state,
let's try to disable it by default. Plain RC6 is the one responsible for
most energy savings, and so far it haven't given any problems - at least,
none we are aware of.

So with this, when i915_enable_rc6=-1 (e.g., the default value), we'll
attempt to enable plain RC6 only on SNB. For Ivy Bridge, the behavior
stays the same as always - we enable both RC6 and deep RC6.

Note that while this exact patch does not has explicit tested-by's, the
equivalent settings were fixed in 3.3 kernel by a smaller patch. And it
has also received considerable testing through Canonical RC6 task-force
testing at https://wiki.ubuntu.com/Kernel/PowerManagementRC6. Up to date,
it looks like all the known issues are gone.

v2: improve description and reference a couple of open bugs related to
RC6 which seem to be fixed with this change.

References: https://bugs.freedesktop.org/show_bug.cgi?id=41682
References: https://bugs.freedesktop.org/show_bug.cgi?id=38567
References: https://bugs.freedesktop.org/show_bug.cgi?id=44867
Acked-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: allow to select rc6 modes via kernel parameter
Eugeni Dodonov [Fri, 23 Mar 2012 14:57:18 +0000]
drm/i915: allow to select rc6 modes via kernel parameter

This allows to select which rc6 modes are to be used via kernel parameter,
via a bitmask parameter. E.g.:

- to enable rc6, i915_enable_rc6=1
- to enable rc6 and deep rc6, i915_enable_rc6=3
- to enable rc6 and deepest rc6, use i915_enable_rc6=5
- to enable rc6, deep and deepest rc6, use i915_enable_rc6=7

Please keep in mind that the deepest RC6 state really should NOT be used
by default, as it could potentially worsen the issues with deep RC6. So do
enable it only when you know what you are doing. However, having it around
could help solving possible future rc6-related issues and their debugging
on user machines.

Note that this changes behavior - previously, value of 1 would enable both
RC6 and deep RC6. Now it should only enable RC6 and deep/deepest RC6
stages must be enabled manually.

v2: address Chris Wilson comments and clean up the code.

References: https://bugs.freedesktop.org/show_bug.cgi?id=42579
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: Mark untiled BLT commands as fenced on gen2/3
Chris Wilson [Wed, 21 Mar 2012 10:48:18 +0000]
drm/i915: Mark untiled BLT commands as fenced on gen2/3

The BLT commands on gen2/3 utilize the fence registers and so we cannot
modify any fences for the object whilst those commands are in flight.
Currently we marked tiled commands as occupying a fence, but forgot to
restrict the untiled commands from preventing a fence being assigned
before they were completed.

One side-effect is that we ten have to double check that a fence was
allocated for a fenced buffer during move-to-active.

Reported-by: Jiri Slaby <jirislaby@gmail.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=43427
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=47990
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Testcase: i-g-t/tests/gem_tiled_after_untiled_blt
Tested-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: stable@kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: properly restore the ppgtt page directory on resume
Daniel Vetter [Wed, 21 Mar 2012 23:14:43 +0000]
drm/i915: properly restore the ppgtt page directory on resume

The ppgtt page directory lives in a snatched part of the gtt pte
range. Which naturally gets cleared on hibernate when we pull the
power. Suspend to ram (which is what I've tested) works because
despite the fact that this is a mmio region, it is actually back by
system ram.

Fix this by moving the page directory setup code to the ppgtt init
code (which gets called on resume).

This fixes hibernate on my ivb and snb.

Reviewed-by: Ben Widawsky <ben@bwidawsk.net>
Signed-Off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm/i915: Sanitize BIOS debugging bits from PIPECONF
Chris Wilson [Thu, 22 Mar 2012 15:00:50 +0000]
drm/i915: Sanitize BIOS debugging bits from PIPECONF

Quoting the BSpec from time immemorial:

  PIPEACONF, bits 28:27: Frame Start Delay (Debug)

  Used to delay the frame start signal that is sent to the display planes.
  Care must be taken to insure that there are enough lines during VBLANK
  to support this setting.

An instance of the BIOS leaving these bits set was found in the wild,
where it caused our modesetting to go all squiffy and skewiff.

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=47271
Reported-and-tested-by: Eva Wang <evawang@linpus.com>
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=43012
Reported-and-tested-by: Carl Richell <carl@system76.com>
Cc: stable@kernel.org
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>

8 years agodrm: Validate requested virtual size against allocated fb size
Chris Wilson [Mon, 26 Mar 2012 20:15:53 +0000]
drm: Validate requested virtual size against allocated fb size

mplayer -vo fbdev tries to create a screen that is twice as tall as the
allocated framebuffer for "doublebuffering". By default, and all in-tree
users, only sufficient memory is allocated and mapped to satisfy the
smallest framebuffer and the virtual size is no larger than the actual.
For these users, we should therefore reject any userspace request to
create a screen that requires a buffer larger than the framebuffer
originally allocated.

References: https://bugs.freedesktop.org/show_bug.cgi?id=38138
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: stable@kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agodrm/radeon: Don't dereference possibly-NULL pointer.
Michel Dänzer [Thu, 29 Mar 2012 14:47:43 +0000]
drm/radeon: Don't dereference possibly-NULL pointer.

Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agomm, drm/udl: fixup vma flags on mmap
Konstantin Khlebnikov [Sat, 31 Mar 2012 09:29:25 +0000]
mm, drm/udl: fixup vma flags on mmap

There should be VM_MIXEDMAP, not VM_PFNMAP, because udl_gem_fault() inserts
pages via vm_insert_page(). Other drm/gem drivers already do this.

Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Cc: Dave Airlie <airlied@redhat.com>
Cc: dri-devel@lists.freedesktop.org
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agodrm/radeon/kms: fix fans after resume
Alex Deucher [Thu, 29 Mar 2012 23:04:08 +0000]
drm/radeon/kms: fix fans after resume

On pre-R600 asics, the SpeedFanControl table is not
executed as part of ASIC_Init as it is on newer asics.


Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agonouveau/bios: Fix tracking of BIOS image data
Benjamin Herrenschmidt [Mon, 2 Apr 2012 03:38:19 +0000]
nouveau/bios: Fix tracking of BIOS image data

The code tries various methods for retreiving the BIOS data. However
it doesn't clear the bios->data pointer between the iterations.

In some cases, the shadow() method will fail and not update bios->data
at all, which will cause us to "score" the old data and incorrectly
attribute that score to the new method. This can cause double frees
later when disposing of the unused data.

Additionally, we were not freeing the data for methods that fail the
score test (we only freed when a "best" is superseeded, not when the
new method has a lower score than the exising "best"). Fix that as well.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agonouveau: Fix crash when pci_ram_rom() returns a size of 0
Benjamin Herrenschmidt [Mon, 2 Apr 2012 03:37:13 +0000]
nouveau: Fix crash when pci_ram_rom() returns a size of 0

From b15b244d6e6e20964bd4b85306722cb60c3c0809 Mon Sep 17 00:00:00 2001
From: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Date: Mon, 2 Apr 2012 13:28:18 +1000

Under some circumstances, pci_map_rom() can return a valid mapping
but a size of 0 (if it cannot find an image in the header).

This causes nouveau to try to kmalloc() a 0 sized pointer and
dereference it, which crashes.

Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Acked-by: Ben Skeggs <bskeggs@redhat.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agodrm/nouveau: select POWER_SUPPLY
Dave Airlie [Mon, 2 Apr 2012 09:45:49 +0000]
drm/nouveau: select POWER_SUPPLY

Ben H. reported that building nouveau into the kernel and power supply
as a module was broken.

Just have nouveau select it, like radeon does.

Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Signed-off-by: Dave Airlie <airlied@redhat.com>

8 years agodrm/nouveau: inform userspace of relaxed kernel subchannel requirements
Ben Skeggs [Thu, 29 Mar 2012 10:24:34 +0000]
drm/nouveau: inform userspace of relaxed kernel subchannel requirements

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

8 years agoRevert "drm/nouveau: inform userspace of new kernel subchannel requirements"
Ben Skeggs [Thu, 29 Mar 2012 10:21:32 +0000]
Revert "drm/nouveau: inform userspace of new kernel subchannel requirements"

This reverts commit a81f15499887d3f9f24ec70bb9b7e778942a6b7b.

Gah, we have a released userspace component using fixed subc assignment
that conflicts with this.  To avoid breaking ABI this needs to be

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

8 years agodrm/nouveau: oops, create m2mf for nvd9 too
Ben Skeggs [Tue, 27 Mar 2012 04:41:04 +0000]
drm/nouveau: oops, create m2mf for nvd9 too

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

8 years agomicroblaze: Fix ret_from_fork declaration
Michal Simek [Mon, 2 Apr 2012 10:55:47 +0000]
microblaze: Fix ret_from_fork declaration

ret_from_fork is used by noMMU system too.

It should be the part of patch
"Disintegrate asm/system.h for Microblaze"
(sha1: c40d04df152a1111c5bbcb632278394dabd2b73d)

Signed-off-by: Michal Simek <monstr@monstr.eu>

8 years agomicroblaze: Do not use tlb_skip in early_printk
Michal Simek [Mon, 2 Apr 2012 10:50:54 +0000]
microblaze: Do not use tlb_skip in early_printk

tlb_skip is valid only for MMU system.

Signed-off-by: Michal Simek <monstr@monstr.eu>

8 years agomicroblaze: Add missing headers caused by disintegration asm/system.h
Michal Simek [Fri, 30 Mar 2012 09:29:38 +0000]
microblaze: Add missing headers caused by disintegration asm/system.h

It should be the part of patch
"Disintegrate asm/system.h for Microblaze"
(sha1: c40d04df152a1111c5bbcb632278394dabd2b73d)

Signed-off-by: Michal Simek <monstr@monstr.eu>

8 years agomicroblaze: Fix stack usage in PAGE_SIZE copy_tofrom_user
Michal Simek [Fri, 30 Mar 2012 06:21:38 +0000]
microblaze: Fix stack usage in PAGE_SIZE copy_tofrom_user

If access to user space failed we need to reconstruct
stack pointer and restore all register.

This patch fixed problem introduces by:
"microblaze: Add loop unrolling for PAGE in copy_tofrom_user"
(sha1: ebe211254bfa6295f4ab0b33c7c881bdfabbab60)

Signed-off-by: Michal Simek <monstr@monstr.eu>

8 years agomicroblaze: Fix tlb_skip variable on noMMU system
Michal Simek [Wed, 28 Mar 2012 08:31:36 +0000]
microblaze: Fix tlb_skip variable on noMMU system

TLBs are available only for MMU systems.

Error log:
arch/microblaze/kernel/setup.c: In function 'debugfs_tlb':
arch/microblaze/kernel/setup.c:217: error: 'tlb_skip' undeclared (first use in this function)
arch/microblaze/kernel/setup.c:217: error: (Each undeclared identifier is reported only once
arch/microblaze/kernel/setup.c:217: error: for each function it appears in.)
make[1]: *** [arch/microblaze/kernel/setup.o] Error 1

Signed-off-by: Michal Simek <monstr@monstr.eu>

8 years agomicroblaze: Fix __futex_atomic_op macro register usage
Michal Simek [Tue, 27 Mar 2012 08:10:15 +0000]
microblaze: Fix __futex_atomic_op macro register usage

Old Microblaze toolchain supported "b" contstrains for
all register but it always points to general purpose reg.
New Microblaze toolchain is more strict in this
and general purpose register should be used there "r".

Signed-off-by: Michal Simek <monstr@monstr.eu>

8 years agom68k/q40: Add missing platform check before registering platform devices
Geert Uytterhoeven [Sun, 18 Mar 2012 12:20:27 +0000]
m68k/q40: Add missing platform check before registering platform devices

On multi-platform kernels, the Q40/Q60 platform devices should be
registered when running on Q40/Q60 only. Else it may crash later.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>

8 years agom68k/mac: Add missing platform check before registering platform devices
Geert Uytterhoeven [Sun, 18 Mar 2012 12:21:38 +0000]
m68k/mac: Add missing platform check before registering platform devices

On multi-platform kernels, the Mac platform devices should be registered
when running on Mac only. Else it may crash later.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: stable@vger.kernel.org

8 years agom68k: include asm/cmpxchg.h in our m68k atomic.h
Greg Ungerer [Fri, 30 Mar 2012 05:52:09 +0000]
m68k: include asm/cmpxchg.h in our m68k atomic.h

After commit 9ffc93f203c18a70623f21950f1dd473c9ec48cd ("Remove all

CC init/main.o
In file included from include/linux/mm.h:15:0,
                 from include/linux/ring_buffer.h:5,
                 from include/linux/ftrace_event.h:4,
                 from include/trace/syscall.h:6,
                 from include/linux/syscalls.h:78,
                 from init/main.c:16:
include/linux/debug_locks.h: In function ‘__debug_locks_off’:
include/linux/debug_locks.h:16:2: error: implicit declaration of function ‘xchg’

There is no indirect inclusions of the new asm/cmpxchg.h for m68k here.
Looking at most other architectures they include asm/cmpxchg.h in their
asm/atomic.h. M68k currently does not do this. Including this in atomic.h
fixes all m68k build problems.

Signed-off-by: Greg Ungerer <gerg@uclinux.org>
Acked-by: David Howells <dhowells@redhat.com>
Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>

8 years agodrm: base prime/dma-buf support (v5)
Dave Airlie [Fri, 25 Nov 2011 15:21:02 +0000]
drm: base prime/dma-buf support (v5)

This adds the basic drm dma-buf interface layer, called PRIME. This
commit doesn't add any driver support, it is simply and agreed upon starting
point so we can work towards merging driver support for the next merge window.

Current drivers with work done are nouveau, i915, udl, exynos and omap.

The main APIs exposed to userspace allow translating a 32-bit object handle
to a file descriptor, and a file descriptor to a 32-bit object handle.

The flags value is currently limited to O_CLOEXEC.

Daniel Vetter: lots of review
Rob Clark: cleaned up lots of the internals and did lifetime review.

v2: rename some functions after Chris preferred a green shed
v3: Fix Ville pointed out using buffer + kmalloc
v4: add locking as per ickle review
v5: allow re-exporting the original dma-buf (Daniel)

Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Reviewed-by: Rob Clark <rob.clark@linaro.org>
Reviewed-by: Sumit Semwal <sumit.semwal@linaro.org>
Reviewed-by: Inki Dae <inki.dae@samsung.com>
Acked-by: Ben Widawsky <benjamin.widawsky@intel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>