There's no reason why acpi_ioremap_lock has to be a spinlock,
because all of the functions it is used in may sleep anyway and
there's no reason why it should be locked with interrupts off.
Use a mutex instead (that's going to allow us to put some more
operations under the lock later).
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

The functions acpi_os_map_generic_address() and
acpi_os_unmap_generic_address() are only used in drivers/acpi/osl.c,
so make them static and remove the extern definitions of them from
include/linux/acpi_io.h.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

This patch reverts one hunk of 677bd810
"ACPI video: remove output switching control", namely the removal of
probing for _DOS/_DOD when searching for video devices.

This is needed on some Fujitsu Laptops (at least S7110, P8010) for the
ACPI backlight interface to work, as an these machines, neither ROM nor
posting methods are available, and after removal of output switching,
none of the caps triggers, which prevents the backlight search from
being entered.

Tested on a Fujitsu Lifebook S7110 and Fujitsu Lifebook P8010.
This probably fixes https://bugzilla.kernel.org/show_bug.cgi?id=27312
for the people who have no entry in /sys/class/backlight.

This is the complete list of public (starting with "_") methods implemented
on the S7110, BIOS rev 1.34:

\_SB_.PCI0.GFX0._ADR
\_SB_.PCI0.GFX0._DOS
\_SB_.PCI0.GFX0._DOD
\_SB_.PCI0.GFX0.CRT._ADR
\_SB_.PCI0.GFX0.CRT._DCS
\_SB_.PCI0.GFX0.CRT._DGS
\_SB_.PCI0.GFX0.CRT._DSS
\_SB_.PCI0.GFX0.LCD._ADR
\_SB_.PCI0.GFX0.LCD._BCL
\_SB_.PCI0.GFX0.LCD._BCM
\_SB_.PCI0.GFX0.LCD._BQC
\_SB_.PCI0.GFX0.LCD._DCS
\_SB_.PCI0.GFX0.LCD._DGS
\_SB_.PCI0.GFX0.LCD._DSS
\_SB_.PCI0.GFX0.LCD._PS0
\_SB_.PCI0.GFX0.LCD._PS3
\_SB_.PCI0.GFX0.TV._ADR
\_SB_.PCI0.GFX0.TV._DCS
\_SB_.PCI0.GFX0.TV._DGS
\_SB_.PCI0.GFX0.TV._DSS
\_SB_.PCI0.GFX0.DVI._ADR
\_SB_.PCI0.GFX0.DVI._DCS
\_SB_.PCI0.GFX0.DVI._DGS
\_SB_.PCI0.GFX0.DVI._DSS
Signed-off-by: NMichael Karcher <kernel@mkarcher.dialup.fu-berlin.de>
Acked-by: NZhang Rui <rui.zhang@intel.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Commit 9630bdd9 (ACPI: Use GPE reference counting to support shared
GPEs) introduced a suspend regression where boxes resume immediately
after being suspended due to the lid or sleep button wakeup status
not being cleared properly.  This happens if the GPEs corresponding
to those devices are not enabled all the time, which apparently is
expected by some BIOSes.

To fix this problem, enable button and lid GPEs unconditionally
during initialization and keep them enabled all the time, regardless
of whether or not the ACPI button driver is used.

References: https://bugzilla.kernel.org/show_bug.cgi?id=27372Reported-and-tested-by: NFerenc Wágner <wferi@niif.hu>
Cc: stable@kernel.org
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

Some ACPI BIOSes define _PRW for the root object which causes
acpi_setup_gpe_for_wake() to crash when trying to dereference the
bogus device_node pointer.  Avoid the crash by checking if
wake_device is not the root object before attempting to set up the
"implicit notify" mechanism for it.

The problem was introduced by commit bba63a29
(ACPICA: Implicit notify support) that added the wake_device argument
to acpi_setup_gpe_for_wake().
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

The functions acpi_os_read_memory() and acpi_os_write_memory() do
two wrong things.  First, they shouldn't call rcu_read_unlock()
before the looked up address is actually used for I/O, because in
that case the iomap it belongs to may be removed before the I/O
is done.  Second, if they have to create a new mapping, they should
check the returned virtual address and tell the caller that the
operation failed if it is NULL (in fact, I think they even should not
attempt to map an address that's not present in one of the existing
ACPI iomaps, because that may cause problems to happen when they are
called from nonpreemptible context and their callers ought to know
what they are doing and map the requisite memory regions beforehand).

Make these functions call rcu_read_unlock() when the I/O is complete
(or if it's necessary to map the given address "on the fly") and
return an error code if the requested physical address is not present
in the existing ACPI iomaps and cannot be mapped.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>

It turns out that some device drivers map pages from the ACPI NVS region
during resume using ioremap(), which conflicts with ioremap_cache() used
for mapping those pages by the NVS save/restore code in nvs.c.

Make the NVS pages mapped by the code in nvs.c be unmapped before device
drivers' resume routines run.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Commit ca9b600b ("ACPI / PM: Make suspend_nvs_save() use
acpi_os_map_memory()") attempted to prevent the code in osl.c and nvs.c
from using different ioremap() variants by making the latter use
acpi_os_map_memory() for mapping the NVS pages.  However, that also
requires acpi_os_unmap_memory() to be used for unmapping them, which
causes synchronize_rcu() to be executed many times in a row
unnecessarily and introduces substantial delays during resume on some
systems.

Instead of using acpi_os_map_memory() for mapping the NVS pages in nvs.c
introduce acpi_os_ioremap() calling ioremap_cache() and make the code in
both osl.c and nvs.c use it.
Reported-by: NJeff Chua <jeff.chua.linux@gmail.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

The meaning of CONFIG_EMBEDDED has long since been obsoleted; the option
is used to configure any non-standard kernel with a much larger scope than
only small devices.

This patch renames the option to CONFIG_EXPERT in init/Kconfig and fixes
references to the option throughout the kernel.  A new CONFIG_EMBEDDED
option is added that automatically selects CONFIG_EXPERT when enabled and
can be used in the future to isolate options that should only be
considered for embedded systems (RISC architectures, SLOB, etc).

Calling the option "EXPERT" more accurately represents its intention: only
expert users who understand the impact of the configuration changes they
are making should enable it.
Reviewed-by: NIngo Molnar <mingo@elte.hu>
Acked-by: NDavid Woodhouse <david.woodhouse@intel.com>
Signed-off-by: NDavid Rientjes <rientjes@google.com>
Cc: Greg KH <gregkh@suse.de>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Arnd Bergmann <arnd@arndb.de>
Cc: Robin Holt <holt@sgi.com>
Cc: <linux-arch@vger.kernel.org>
Signed-off-by: NAndrew Morton <akpm@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

This partially reverts commit da8aeb92
("ACPI / Battery: Update information on info notification and resume"),
which causes a hang on resume on at least some machines.

This bug was bisected on an ASUS EeePC 901, which hangs at resume time
if we do that "acpi_battery_refresh(battery)" in the battery resume
function.

Rafael suspects we'll still need to refresh the sysfs files upon resume,
but that that can be done from a PM notifier (that will run after
thawing user space).
Bisected-and-tested-by: NLinus Torvalds <torvalds@linux-foundation.org>
Cc: Matthew Garrett <mjg@redhat.com>
Cc: Len Brown <len.brown@intel.com>
Acked-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Signed-off-by: NBob Moore <robert.moore@intel.com>
Signed-off-by: NLin Ming <ming.m.lin@intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

History: This support changes a method to "serialized" on the fly if the
method generates an AE_ALREADY_EXISTS error, indicating the possibility
that it cannot handle reentrancy.

This fix repairs a couple of issues seen in the field, especially on
machines with many cores.

1) Delete method children only upon the exit of the last thread, so
as to not delete objects out from under running threads.

2) Set the "serialized" bit for the method only upon the exit of the
last thread, so as to not cause deadlock when running threads attempt
to exit.

3) Cleanup the use of the AML "MethodFlags" and internal method flags
so that there is no longer any confustion between the two.
Reported-by: NDana Myers <dana.myers@oracle.com>
Signed-off-by: NLin Ming <ming.m.lin@intel.com>
Signed-off-by: NBob Moore <robert.moore@intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

Prevents issues if the namespace is changing underneath the
debugger.  Especially temporary nodes, since the debugger displays
these also.
Signed-off-by: NBob Moore <robert.moore@intel.com>
Signed-off-by: NLin Ming <ming.m.lin@intel.com>
Reviewed-by: NRafael Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Fixes a race condition between method execution and namespace
walks that can possibly fault. Problem was apparently introduced
in version 20100528 as a result of a performance optimization
that reduces the number of namespace walks upon method exit
by using the delete_namespace_subtree function instead of the
delete_namespace_by_owner function used previously. Bug is in
the delete_namespace_subtree function.
Signed-off-by: NDana Myers <dana.myers@oracle.com>
Signed-off-by: NBob Moore <robert.moore@intel.com>
Signed-off-by: NLin Ming <ming.m.lin@intel.com>
Reviewed-by: NRafael Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

We will leak the memory allocated to 'local_gpe_event_info' if
'acpi_ut_acquire_mutex()' fails or if 'acpi_ev_valid_gpe_event()' fails in
drivers/acpi/acpica/evgpe.c::acpi_ev_asynch_execute_gpe_method().
Signed-off-by: NJesper Juhl <jj@chaosbits.net>
Reviewed-by: NRafael Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Commit 415e12b2 ("PCI/ACPI: Request _OSC control once for each root
bridge (v3)") put the acpi_hest_init() call in acpi_pci_root_init() into
a wrong place, presumably because the author confused acpi_pci_disabled
with acpi_disabled.  Bring the code ordering in acpi_pci_root_init()
back to sanity.

Additionally, make sure that hest_disable is set when acpi_disabled is
set, which is going to prevent acpi_hest_parse(), that still may be
executed for acpi_disabled=1 through aer_acpi_firmware_first(), from
crashing because of uninitialized hest_tab.
Reported-and-tested-by: NAndres Salomon <dilinger@queued.net>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Commit 9cd03144 (ACPI / ACPICA: Fix global lock acquisition) forgot to
initialize the spinlock it added.  Fix that.
Reported-and-tested-by: NSergey Senozhatsky <sergey.senozhatsky@gmail.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Tested-by: NLinus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: NLinus Torvalds <torvalds@linux-foundation.org>

Move the evaluation of acpi_pci_osc_control_set() (to request control of
PCI Express native features) into acpi_pci_root_add() to avoid calling
it many times for the same root complex with the same arguments.
Additionally, check if all of the requisite _OSC support bits are set
before calling acpi_pci_osc_control_set() for a given root complex.

References: https://bugzilla.kernel.org/show_bug.cgi?id=20232Reported-by: NOzan Caglayan <ozan@pardus.org.tr>
Tested-by: NOzan Caglayan <ozan@pardus.org.tr>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NJesse Barnes <jbarnes@virtuousgeek.org>

The recent rework of the NVS saving/restoring code introduced two
build issues for !CONFIG_ACPI, a warning in drivers/acpi/internal.h
and an error in arch/x86/kernel/e820.c.

Fix them by providing suitable static inline definitions of the
relevant functions.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Acked-by: NRandy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

CPUIDLE_FLAG_SHALLOW
CPUIDLE_FLAG_BALANCED
CPUIDLE_FLAG_DEEP
CPUIDLE_FLAG_CHECK_BM

were set by acpi_processor_setup_cpuidle(),
but never used by cpuidle or by acpi_idle.
So stop setting them.
Signed-off-by: NLen Brown <len.brown@intel.com>

Having four variables for the same thing:
  idle_halt, idle_nomwait, force_mwait and boot_option_idle_overrides
is rather confusing and unnecessary complex.

if idle= boot param is passed, only set up one variable:
boot_option_idle_overrides

Introduces following functional changes/fixes:
  - intel_idle driver does not register if any idle=xy
    boot param is passed.
  - processor_idle.c will also not register a cpuidle driver
    and get active if idle=halt is passed.
    Before a cpuidle driver with one (C1, halt) state got registered
    Now the default_idle function will be used which finally uses
    the same idle call to enter sleep state (safe_halt()), but
    without registering a whole cpuidle driver.

That means idle= param will always avoid cpuidle drivers to register
with one exception (same behavior as before):
idle=nomwait
may still register acpi_idle cpuidle driver, but C1 will not use
mwait, but hlt. This can be a workaround for IO based deeper sleep
states where C1 mwait causes problems.
Signed-off-by: NThomas Renninger <trenn@suse.de>
cc: x86@kernel.org
Signed-off-by: NLen Brown <len.brown@intel.com>

printk("%pR",...)
is for formatting struct resource only.
But the list built up in drivers/acpi/osl.c uses it's own struct:
struct acpi_res_list {}

Without this patch you can see wrongly formatted resources (SMRG is of IO type):

ACPI: resource 0000:00:1f.3 [io  0x0400-0x041f] conflicts with AC
      PI region SMRG [mem 0x00000400-0x0000040f 64bit pref disabled]

https://bugzilla.kernel.org/show_bug.cgi?id=26342Signed-off-by: NThomas Renninger <trenn@suse.de>
CC: Matthew Wilcox <matthew@wil.cx>
Signed-off-by: NLen Brown <len.brown@intel.com>

As pointed out by Linus CONFIG_X86 in drivers/acpi/numa.c is
ugly.

Builds and boots on ia64 (both normally and with maxcpus=8 to limit
the number of cpus).
Signed-off-by: NTony Luck <tony.luck@intel.com>
Acked-by: NYinghai Lu <yinghai@kernel.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Wu Fengguang <fengguang.wu@intel.com>
Cc: Bjorn Helgaas <bjorn.helgaas@hp.com>
Cc: Len Brown <len.brown@intel.com>
LKML-Reference: <4D2D6B5D.4080208@kernel.org>
Signed-off-by: NIngo Molnar <mingo@elte.hu>

A notification event 0x81 from an ACPI battery device requires us to
re-read the battery information structure.  Follow this requirement
and remove and re-create the battery's attibutes in sysfs so that
they reflect the reporting units used by the battery at the moment
(those units may actually change sometimes at run time, which happens
on some Thinkpads).

The approach used in this patch was suggested by Matthew Garrett.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Reported-by: NMatthew Garrett <mjg@redhat.com>
Signed-off-by: NLen Brown <len.brown@intel.com>

The wake_capable ACPI device flag is not necessary, because it is
only used in scan.c for recording the information that _PRW is
present for the given device.  That information is only used by
acpi_add_single_object() to decide whether or not to call
acpi_bus_get_wakeup_device_flags(), so the flag may be dropped
if the _PRW check is moved to acpi_bus_get_wakeup_device_flags().
Moreover, acpi_bus_get_wakeup_device_flags() always returns 0,
so it really should be void.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Before evaluating _PRW for devices that are reported as inactive or
not present by their _STA control methods we should check if those
methods are actually present (otherwise the evaulation of _PRW will
obviously fail and a scary message will be printed unnecessarily).
Reported-by: NAndreas Mohr <andi@lisas.de>
Reported-by: NMaciej Rutecki <maciej.rutecki@gmail.com>
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

It certainly is not a good idea to execute _ON or _OFF and _STA
for the same power resource at the same time which may happen in
some circumstances in theory.  To prevent that from happening,
read the power state of each power resource under its mutex, as
that will prevent the state from being changed at the same time.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Rename acpi_power_off_device() to acpi_power_off() in analogy with
acpi_power_on().
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Since acpi_bus_set_power() should not use __acpi_bus_get_power() to
update the device's device->power.state field before changing its
power state (this may cause device->power.state to be inconsistent
with the device power resources' reference counters), remove this
call from it.  In consequence, the acpi_power_nocheck variable is not
necessary any more, so it can be dropped along with the DMI table
used for setting that variable for HP Pavilion 05.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

There are no more users of acpi_bus_get_power(), so it can be
dropped.  Moreover, it should be dropped, because it modifies
the device->power.state field of an ACPI device without updating
the reference counters of the device's power resources, which is
wrong.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Use the new function acpi_bus_update_power() for manipulating power
resources used by ACPI fan devices, which allows them to be put into
the right state during initialization and resume.  Consequently,
remove the flags.force_power_state field from struct acpi_device,
which is not necessary any more.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Depending on the organization of the ACPI namespace, power resource
device objects may generally be scanned after the "regular" device
objects that they are referred from through _PRn.  This, in turn, may
cause acpi_bus_get_power_flags() to attempt to access them through
acpi_bus_init_power() before they are registered (and initialized by
acpi_power_driver).  [This is not a theoretical issue, it actually
happens for one PnP device on my testbed HP nx6325.]

To fix this problem, make acpi_bus_get_power_flags() attempt to
register power resource devices as soon as they have been found in
the _PRn output for any other devices.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

The ACPI device driver used for handling power resources,
acpi_power_driver, creates a struct acpi_power_resource object for
each ACPI device representing a power resource.  These objects are
then used when setting and reading the power states of devices using
the corresponding power resources.  Unfortunately, acpi_power_driver
is registered after acpi_scan_init() that may add devices using the
power resources before acpi_power_driver has a chance to create
struct acpi_power_resource objects for them (specifically, the power
resources may be referred to during the scanning process through
acpi_bus_get_power() before they have been initialized).

As the first step towards fixing this issue, move the registration
of acpi_power_driver into acpi_scan_init() so that power resource
devices can be initialized by it as soon as they have been found in
the namespace.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Add function acpi_bus_update_power() for reading the actual power
state of an ACPI device and updating its device->power.state field
in such a way that its power resources' reference counters will
remain consistent with that field.

For this purpose introduce __acpi_bus_set_power() setting the
power state of an ACPI device without updating its
device->power.state field and make acpi_bus_set_power() and
acpi_bus_update_power() use it (acpi_bus_set_power() retains the
current behavior for now).
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Add function acpi_bus_init_power() for getting the initial power
state of an ACPI device and reference counting its power resources
as appropriate.

Make acpi_bus_get_power_flags() use the new function instead of
acpi_bus_get_power() that updates device->power.state without
reference counting the device's power resources.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

It sometimes is necessary to get the power state of an ACPI device
without updating its device->power.state field, for example to
avoid inconsistencies between device->power.state and the reference
counters of the device's power resources.  For this purpose introduce
__acpi_bus_get_power() that will return the given device's power
state via a pointer (instead of modifying device->power.state)
and make acpi_bus_get_power() use it.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Introduce function acpi_power_on_resources() that reference counts
and possibly turns on ACPI power resources for a given device and
a given power state of it.

This function will be used for reference counting device power
resources during initialization.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

ACPI device power resources should be reference counted during
device initialization, so that their reference counters are always
up to date.  It is convenient to do that with the help of a function
that will reference count and possibly turn on power resources in
a given list, so introduce that function, acpi_power_on_list().
For symmetry, introduce acpi_power_off_list() for performing the
reverse operation and use the both of them to simplify
acpi_power_transition().
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

acpi_power_get_inferred_state() should not update
device->power.state behind the back of its caller, so make it return
the state via a pointer instead.
Signed-off-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLen Brown <len.brown@intel.com>

Modify/add some comments to minimize ACPICA/linux GPE code divergence.
Acked-by: NRafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: NLin Ming <ming.m.lin@intel.com>
Signed-off-by: NLen Brown <len.brown@intel.com>