Borislav Petkov [Sun, 16 Dec 2007 23:30:22 +0000 (00:30 +0100)]
Suspend: Clean up suspend_64.c
There's a freakishly long comment in suspend_64.c, shorten it.
Signed-off-by: Borislav Petkov <bbpetkov@yahoo.de> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Johannes Berg [Wed, 16 Jan 2008 04:17:00 +0000 (23:17 -0500)]
Suspend: Add config option to disable the freezer if architecture wants that
This patch makes the freezer optional for suspend to allow the
system to work (or not work) like the original PMU suspend.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Make acpi_sleep_prepare() static and cause it to print a message
specifying the ACPI system sleep state to be entered (helpful for
debugging the suspend/hibernation code).
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@suse.cz> Signed-off-by: Len Brown <len.brown@intel.com>
ACPI hibernation: Call _PTS before suspending devices
The ACPI 1.0 specification wants us to put devices into low power
states after executing the _PTS global control method, while ACPI
2.0 and later want us to do that in the reverse order. The current
hibernation code follows ACPI 2.0 in that respect which may cause some
ACPI 1.0x systems to hang during hibernation (ref.
http://bugzilla.kernel.org/show_bug.cgi?id=9528).
Make the hibernation code execute _PTS before putting devices into
low power states (ie. in accordance with ACPI 1.0x) with the
possibility to override that using the 'acpi_new_pts_ordering' kernel
command line option.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Hibernation: Introduce begin() and end() callbacks
Introduce global hibernation callback .end() and rename global
hibernation callback .start() to .begin(), in analogy with the
recent modifications of the global suspend callbacks.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
The ACPI 1.0 specification wants us to put devices into low power
states after executing the _PTS global control method, while ACPI
2.0 and later want us to do that in the reverse order. The current
suspend code follows ACPI 2.0 in that respect which causes some
ACPI 1.0x systems to hang during suspend (ref.
http://bugzilla.kernel.org/show_bug.cgi?id=9528).
Make the suspend code execute _PTS before putting devices into low
power states (ie. in accordance with ACPI 1.0x) and provide a command
line option to override the default if need be.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
The preparation to enter an ACPI system sleep state is now tied to
the disabling of GPEs, but the GPEs should not be disabled before
suspending devices. Since on ACPI 1.0x systems the _PTS global
control method should be executed before suspending devices, we
need to disable GPEs separately.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@suse.cz> Signed-off-by: Len Brown <len.brown@intel.com>
ACPI: Separate invocations of _GTS and _BFS from _PTS and _WAK
The execution of ACPI global control methods _GTS and _BFS is
currently tied to the preparation to enter a sleep state and to the
leaving of the sleep state, respectively. However, these functions
are called before disabling the nonboot CPUs and after enabling
them, respectively (in fact, on ACPI 1.0x systems the first of them
ought to be called before suspending devices), while according to the
ACPI specification, _GTS is to be executed right prior to entering
the system sleep state and _BFS is to be executed right after the
platfor firmware has returned control to the OS on wake up.
Move the execution of _GTS and _BFS to the right places.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@suse.cz> Signed-off-by: Len Brown <len.brown@intel.com>
On ACPI systems the target state set by acpi_pm_set_target() is
reset by acpi_pm_finish(), but that need not be called if the
suspend fails. All platforms that use the .set_target() global
suspend callback are affected by analogous issues.
For this reason, we need an additional global suspend callback that
will reset the target state regardless of whether or not the suspend
is successful. Also, it is reasonable to rename the .set_target()
callback, since it will be used for a different purpose on ACPI
systems (due to ACPI 1.0x code ordering requirements).
Introduce the global suspend callback .end() to be executed at the
end of the suspend sequence and rename the .set_target() global
suspend callback to .begin().
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Shaohua Li [Wed, 20 Jun 2007 01:17:58 +0000 (09:17 +0800)]
ACPI: clear GPE earily in resume to avoid warning
Wakeup GPE hasn't a handler. If system is waked up by such GPE like a
USB hotplug, I saw a lot of error reporting the GPE hasn't handler.
acpi_leave_sleep_state will clear the GPE but it's too late, we should
do it before interrupt is re-enabled.
Signed-off-by: Shaohua Li <shaohua.li@intel.com> Acked-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Johannes Berg [Sat, 8 Dec 2007 01:14:00 +0000 (02:14 +0100)]
Suspend: Clean up Kconfig (V2)
This cleans up the suspend Kconfig and removes the need to
declare centrally which architectures support suspend. All
architectures that currently support suspend are modified
accordingly.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Acked-by: Russell King <rmk@arm.linux.org.uk> Acked-by: Paul Mackerras <paulus@samba.org> Acked-by: Ralf Baechle <ralf@linux-mips.org> Acked-by: Paul Mundt <lethal@linux-sh.org> Cc: Pavel Machek <pavel@suse.cz> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Johannes Berg [Sat, 8 Dec 2007 01:12:39 +0000 (02:12 +0100)]
Hibernation: Clean up Kconfig (V2)
This cleans up the hibernation Kconfig and removes the need to
declare centrally which architectures support hibernation. All
architectures that currently support hibernation are modified
accordingly.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Acked-by: Paul Mackerras <paulus@samba.org> Cc: Pavel Machek <pavel@suse.cz> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Suspend: Fix compilation warning for CONFIG_SUSPEND unset
Suspend: Make debug facility depend on CONFIG_SUSPEND
Make the new suspend debug facility code depend on CONFIG_SUSPEND,
as appropriate, to remove the compiler warning printed when CONFIG_PM is set
and CONFIG_SUSPEND is not set.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Len Brown <len.brown@intel.com>
Alan Stern [Mon, 19 Nov 2007 22:49:18 +0000 (23:49 +0100)]
PM: Convert PM notifiers to out-of-line code
This patch (as1008b) converts the PM notifier routines from inline
calls to out-of-line code. It also prevents pm_chain_head from
being created when CONFIG_PM_SLEEP isn't enabled, and EXPORTs the
notifier registration and unregistration routines.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Johannes Berg [Mon, 19 Nov 2007 22:46:16 +0000 (23:46 +0100)]
PM: Make PM_TRACE more architecture independent
When trying to debug a suspend failure I started implementing
PM_TRACE for powerpc. I then noticed that I'm debugging a suspend
failure and so PM_TRACE isn't useful at all, but thought that
nonetheless this could be useful in the future.
Basically, to support PM_TRACE, you add a Kconfig option that
selects PM_TRACE and provides the infrastructure as per the
help text of PM_TRACE.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Make it possible to test the hibernation core code with the help of the
/sys/power/pm_test attribute introduced for suspend testing in the previous
patch.
Writing an appropriate string to this file causes the hibernation code to work
in one of the test modes defined as follows:
freezer
- test the freezing of processes
devices
- test the freezing of processes and suspending of devices
platform
- test the freezing of processes, suspending of devices and platform global
control methods(*)
processors
- test the freezing of processes, suspending of devices, platform global
control methods(*) and the disabling of nonboot CPUs
core
- test the freezing of processes, suspending of devices, platform global
control methods(*), the disabling of nonboot CPUs and suspending of
platform/system devices
(*) - the platform global control methods are only available on ACPI systems
and are only tested if the hibernation mode is set to "platform"
Then, if a hibernation is started by normal means, the hibernation core will
perform its normal operations up to the point indicated by given test level.
Next, it will wait for 5 seconds and carry out the resume operations needed to
transition the system back to the fully functional state.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Len Brown <len.brown@intel.com>
Introduce sysfs attribute /sys/power/pm_test allowing one to test the suspend
core code. Namely, writing one of the strings:
freezer
devices
platform
processors
core
to this file causes the suspend code to work in one of the test modes defined as
follows:
freezer
- test the freezing of processes
devices
- test the freezing of processes and suspending of devices
platform
- test the freezing of processes, suspending of devices and platform global
control methods(*)
processors
- test the freezing of processes, suspending of devices, platform global
control methods and the disabling of nonboot CPUs
core
- test the freezing of processes, suspending of devices, platform global
control methods, the disabling of nonboot CPUs and suspending of
platform/system devices
(*) These are ACPI global control methods on ACPI systems
Then, if a suspend is started by normal means, the suspend core will perform
its normal operations up to the point indicated by given test level. Next, it
will wait for 5 seconds and carry out the resume operations needed to transition
the system back to the fully functional state.
Writing "none" to /sys/power/pm_test turns the testing off.
When open for reading, /sys/power/pm_test contains a space-separated list of all
available tests (including "none" that represents the normal functionality) in
which the current test level is indicated by square brackets.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Len Brown <len.brown@intel.com>
Alan Stern [Mon, 19 Nov 2007 22:38:25 +0000 (23:38 +0100)]
Hibernation: Add PM_RESTORE_PREPARE and PM_POST_RESTORE notifiers (rev. 2)
Add PM_RESTORE_PREPARE and PM_POST_RESTORE notifiers to the PM core, to be used
in analogy with the existing PM_HIBERNATION_PREPARE and PM_POST_HIBERNATION
notifiers.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: "Rafael J. Wysocki" <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Adrian Bunk [Mon, 19 Nov 2007 22:36:20 +0000 (23:36 +0100)]
Hibernation: Move function prototypes to header
This patch moves the prototypes of count_highmem_pages() and
restore_highmem() to kernel/power/power.h
Signed-off-by: Adrian Bunk <bunk@kernel.org> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Some code in acpi_hibernation_finish() was moved to acpi_hibernation_leave(),
but the old copy had been left (it's harmless, but also useless). Remove it.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Len Brown <len.brown@intel.com>
Hibernation: Correct definitions of some ioctls (rev. 2)
Three ioctl numbers belonging to the hibernation userland interface,
SNAPSHOT_ATOMIC_SNAPSHOT, SNAPSHOT_AVAIL_SWAP, SNAPSHOT_GET_SWAP_PAGE,
are defined in a wrong way (eg. not portable). Provide new ioctl numbers for
these ioctls and mark the existing ones as deprecated.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Len Brown <len.brown@intel.com>
Hibernation: Rework platform support ioctls (rev. 2)
Modify the hibernation userland interface by adding two new ioctls to it,
SNAPSHOT_PLATFORM_SUPPORT and SNAPSHOT_POWER_OFF, that can be used,
respectively, to switch the hibernation platform support on/off and to make the
kernel transition the system to the hibernation state (eg. ACPI S4) using the
platform (eg. ACPI) driver.
These ioctls are intended to replace the misdesigned SNAPSHOT_PMOPS ioctl,
which from now is regarded as obsolete and will be removed in the future.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Len Brown <len.brown@intel.com>
Add a new ioctl, SNAPSHOT_GET_IMAGE_SIZE, returning the size of the (just
created) hibernation image, to the hibernation userland interface.
This ioctl is necessary so that the userland utilities using the interface need
not access the hibernation image header, owned by the kernel, in order to obtain
the size of the image.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Pavel Machek <pavel@ucw.cz> Signed-off-by: Len Brown <len.brown@intel.com>
Ingo Molnar [Fri, 1 Feb 2008 22:45:18 +0000 (23:45 +0100)]
x86: fix bootup crash in native_read_tsc()
fix bootup crash in native_read_tsc() that was reported on an Athlon-XP
and bisected. The correct feature boundary for X86_FEATURE_MFENCE_RDTSC
is not XMM but XMM2.
Signed-off-by: Ingo Molnar <mingo@elte.hu> Acked-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6: (100 commits)
ide: move hwif_register() call out of ide_probe_port()
ide: factor out code for tuning devices from ide_probe_port()
ide: move handling of I/O resources out of ide_probe_port()
ide: make probe_hwif() return an error value
ide: use ide_remove_port_from_hwgroup in init_irq()
ide: prepare init_irq() for using ide_remove_port_from_hwgroup()
ide: factor out code removing port from hwgroup from ide_unregister()
ide: I/O resources are released too early in ide_unregister()
ide: cleanup ide_system_bus_speed()
ide: remove needless zeroing of hwgroup fields from init_irq()
ide: remove unused ide_hwgroup_t fields
ide_platform: remove struct hwif_prop
ide: remove hwif->present manipulations from hwif_init()
ide: move wait_hwif_ready() documentation in the right place
ide: fix handling of busy I/O resources in probe_hwif()
<linux/hdsmart.h> is not used by kernel code
ide: don't include <linux/hdsmart.h>
ide-floppy: cleanup header
ide: update/add my Copyrights
ide: delete filenames/versions from comments
...
ide: move hwif_register() call out of ide_probe_port()
* Add BUG_ON(hwif->present) at the start of ide_probe_port().
* Move hwif_register() call (along with setting hwif->present) from
ide_probe_port() to ide_device_add_all().
As a result the port will be registered with the device tree _after_:
- probing both devices (if both are present)
- port reset (if hwif->reset is set)
- restoring local IRQs state and re-enabling port IRQ
* Remove 'if (hwif != hwif_prop.hwif)' check from plat_ide_remove()
(pdev->dev.driver_data and hwif_prop.hwif are set at the same time
in plat_ide_probe() and are identical).
* Use hwif->index for ide_unregister() in plat_ide_remove().
* Use local variables instead of hwif_prop.plat_ide[_alt]_mapbase in
plat_ide_probe() and remove no longer needed struct hwif_prop.
* pci_map_sg() -> dma_map_sg() in ide_build_sglist().
* pci_unmap_sg() -> dma_unmap_sg() in ide_destroy_dmatable().
There should be no functionality changes caused by this patch except
for blackfin arch whose dma_[un]map_sg() implementation differs from
pci_[un]map_sg() one (on s390 arch there is no PCI, on avr32 and h8300
archs PCI is currently unsupported, on m32r arch PCI support depends
on BROKEN, on m68k arch PCI support depends on HADES which in turn
depends on BROKEN, on all other archs dma_[un]map_sg() functionality
matches with pci_[un]map_sg() one).
* Use hwif->dev instead of state->dev in icside_build_sglist(),
icside_dma_end(), icside_dma_start() and icside_dma_setup().
* Remove no longer needed 'dev' field from struct icside_state.
Cc: Russell King <rmk@arm.linux.org.uk> Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
* Rename 'simplex_stat' variable to 'dma_stat' in ide_get_or_set_dma_base().
* Factor out code for forcing host out of "simplex" mode from
ide_get_or_set_dma_base() to ide_pci_clear_simplex() helper.
* Add IDE_HFLAG_CLEAR_SIMPLEX host flag and set it in alim15x3 (for M5229),
amd74xx (for AMD 7409), cmd64x (for CMD643), generic (for Netcell) and
serverworks (for CSB5) host drivers.
* Make ide_get_or_set_dma_base() test for IDE_HFLAG_CLEAR_SIMPLEX host flag
instead of checking dev->device (BTW the code was buggy because it didn't
check for dev->vendor, luckily none of these PCI Device IDs was used by
some other vendor for PCI IDE controller).
* Remove no longer needed assertion from amd74xx_probe().
* Factor out cable detection for AMD7409 to amd7409_cable_detect() and for
chipsets >= AMD7411 to amd7411_cable_detect().
* Use dev->vendor and dev->device instead of amd_config->udma_mask when
selecting cable detection method and checking for broken FIFO support in
init_chipset_amd74xx().
* Remove no longer needed AMD_BAD_FIFO define.
* Add 'swdma' parameter for setting .swdma_mask to DECLARE_AMD_DEV() macro.
* Add 'udma' parameter for setting .udma_mask to DECLARE_{AMD,NV}_DEV() macro.
* Keep a copy of a current amd74xx_chipsets[] entry in amd74xx_probe()
in order to fix ->swdma_mask on early AMD7409 revisions and ->udma_mask
on Serenade mainboards.
* Remove no longer needed fixups from init_chipset_amd74xx()
and AMD_CHECK_{SWDMA,SERENADE} defines.
* Move printing banner message from init_chipset_amd74xx() to amd74xx_probe(),
also remove incorrect comment while at it.
* Use hwif->ultra_mask instead of amd_config->udma_mask in amd_set_drive().
* Add 'udma_mask' argument to amd_set_speed() and pass UDMA mask from
amd_set_drive() instead of using amd_config->udma_mask.
* Move amd_config->base from AMD_* defines to users of these defines and add
0x40 the defined values. Then add amd_offset() inline helper for selecting
offset from 0x40 base (needed for nVidia controllers) and finally use it in
amd_set_speed(), amd7411_cable_detect() and init_chipset_amd74xx() instead
of amd_config->base.
* Remove no longer needed AMD_BAD_SWDMA define, ->{swdma,ultra}_mask setup
from init_hwif_amd74xx(), amd_{config,chipset} variables and amd_ide_chips
table.
* Fix init_chipset_amd74xx() comment.
* Bump driver version.
There should be no functionality changes caused by this patch.
Sergei Shtylyov [Fri, 1 Feb 2008 22:09:30 +0000 (23:09 +0100)]
ide: ide_setup_dma() assumes 8 ports
According to http://marc.info/?l=linux-ide&m=114346138611631, the drivers must
always register 8 DMA ports with ide_setup_dma(), so its last argument is not
needed. While at it, kill some useless parens in that function...
Signed-off-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
ide-{floppy,tape}: remove debug code for dumping identify data
IDE core supports dumping raw identify data in hdparm friendly format now
so verbose identify dumping in ide-{floppy,tape}.c device drivers (done iff
IDE{FLOPPY,TAPE}_DEBUG_INFO is defined to '1' and it is '0' by default)
is no longer nedeed.
* Add ide_dump_identify() debug helper for dumping raw identify data in
the hdparm friendly format (== the identify data can be extracted from
dmesg output and passed to hdparm --Istdin).
* Dump identify data in ide-probe.c::do_identify() if DEBUG is enabled.
Acked-by: Sergei Shtylyov <sshtylyov@ru.mvista.com> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
ide-cd: merge cdrom_write_check_ireason() and cdrom_read_check_ireason()
Add 'rw' parameter to cdrom_read_check_ireason(), make it handle
both read and write checking, rename it to ide_cd_check_ireason(),
finally remove no longer needed cdrom_write_check_ireason().
There should be no functionality changes caused by this patch.
ide-cd: prepare cdrom_rw_intr() and cdrom_newpc_intr() to be merged
In cdrom_newpc_intr():
* cleanup variables in the 'transfer data' loop
In cdrom_rw_intr():
* rename 'sectors_to_transfer' to 'thislen'
* rename 'this_transfer' to 'blen'
* keep number of bytes (instead of sectors) in 'thislen' and 'blen'
* call 'xferfunc' only once for 'blen'
* cache 'rq->buffer' in 'ptr' variable
* check for 'rq->bio' before setting 'ptr' and 'blen'
* check for 'ptr' instead of 'rq->current_nr_sectors'
There should be no functionality changes caused by this patch.
ide-cd: merge cdrom_start_read_continuation() and cdrom_start_rw_cont()
* Add handling of write requests to cdrom_start_read_continuation(),
rename it to cdrom_start_rw_cont() and remove no longer needed
cdrom_start_write_cont().
ide-cd: merge cdrom_pc_intr() and cdrom_newpc_intr()
Add handling of REQ_TYPE_{SENSE,ATA_PC} requests to cdrom_newpc_intr()
(please note that these requests never have 'bio' attached to them
and they never use DMA), then remove no longer needed cdrom_pc_intr().
There should be no functionality changes caused by this patch.
projects / linux-2.6 / log