Remove dcache_lock locking from hostfs filesystem, and move it into dcache
helpers. All that is required is a coherent path name. Protection from
concurrent modification of the namespace after path name generation is not
provided in current code, because dcache_lock is dropped before the path is
used.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Change d_hash so it may be called from lock-free RCU lookups. See similar
patch for d_compare for details.
For in-tree filesystems, this is just a mechanical change.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Change d_compare so it may be called from lock-free RCU lookups. This
does put significant restrictions on what may be done from the callback,
however there don't seem to have been any problems with in-tree fses.
If some strange use case pops up that _really_ cannot cope with the
rcu-walk rules, we can just add new rcu-unaware callbacks, which would
cause name lookup to drop out of rcu-walk mode.
For in-tree filesystems, this is just a mechanical change.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
smpfs and ncpfs want to update a live dentry name in-place. Rather than
have them open code the locking, provide a documented dcache API.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Use vfat's method for dealing with negative dentries to preserve case,
rather than overwrite dentry name in d_revalidate, which is a bit ugly
and also gets in the way of doing lock-free path walking.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Use vfat's method for dealing with negative dentries to preserve case,
rather than overwrite dentry name in d_revalidate, which is a bit ugly
and also gets in the way of doing lock-free path walking.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Change d_delete from a dentry deletion notification to a dentry caching
advise, more like ->drop_inode. Require it to be constant and idempotent,
and not take d_lock. This is how all existing filesystems use the callback
anyway.
This makes fine grained dentry locking of dput and dentry lru scanning
much simpler.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Remove redundant (and incorrect, since dcache RCU lookup) dentry locking
documentation and point to the canonical document.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Switching d_op on a live dentry is racy in general, so avoid it. In this case
it is a negative dentry, which is safer, but there are still concurrent ops
which may be called on d_op in that case (eg. d_revalidate). So in general
a filesystem may not do this. Fix configfs so as not to do this.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Switching d_op on a live dentry is racy in general, so avoid it. In this case
it is a negative dentry, which is safer, but there are still concurrent ops
which may be called on d_op in that case (eg. d_revalidate). So in general
a filesystem may not do this. Fix cgroupfs so as not to do this.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
The nr_unused counters count the number of objects on an LRU, and as such they
are synchronized with LRU object insertion and removal and scanning, and
protected under the LRU lock.
Making it per-cpu does not actually get any concurrency improvements because of
this lock, and summing the counter is much slower, and
incrementing/decrementing it costs more code size and is slower too.
These counters should stay per-LRU, which currently means global.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
d_validate has been broken for a long time.
kmem_ptr_validate does not guarantee that a pointer can be dereferenced
if it can go away at any time. Even rcu_read_lock doesn't help, because
the pointer might be queued in RCU callbacks but not executed yet.
So the parent cannot be checked, nor the name hashed. The dentry pointer
can not be touched until it can be verified under lock. Hashing simply
cannot be used.
Instead, verify the parent/child relationship by traversing parent's
d_child list. It's slow, but only ncpfs and the destaged smbfs care
about it, at this point.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Since ACPI buttons and lids can be configured to wake up the system
from sleep states, report wakeup events from these devices.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
Drop special ACPI wakeup flags, wakeup.state.enabled and
wakeup.flags.always_enabled, that aren't necessary any more after
we've started to use standard device wakeup flags for handling ACPI
wakeup devices.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
There are ACPI devices (buttons and the laptop lid) that can wake up
the system from sleep states and have no "physical" companion
devices. The ACPI subsystem uses two flags, wakeup.state.enabled and
wakeup.flags.always_enabled, for handling those devices, but they
are not accessible through the standard device wakeup infrastructure.
User space can only control them via the /proc/acpi/wakeup interface
that is not really convenient (e.g. the way in which devices are
enabled to wake up the system is not portable between different
systems, because it requires one to know the devices' "names" used in
the system's ACPI tables).
To address this problem, use standard device wakeup flags instead of
the special ACPI flags for handling those devices. In particular,
use device_set_wakeup_capable() to mark the ACPI wakeup devices
during initialization and use device_set_wakeup_enable() to allow
or disallow them to wake up the system from sleep states. Rework
the /proc/acpi/wakeup interface to take these changes into account.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
If a device is enabled to wake up the system from sleep states via
/proc/acpi/wakeup and there are other devices associated with the
same wakeup GPE, all of these devices are automatically enabled to
wake up the system. This isn't correct, because the fact the GPE is
shared need not imply that wakeup power has to be enabled for all the
devices at the same time (i.e. it is possible that one device will
have its wakeup power enabled and it will wake up the system from a
sleep state if the shared wakeup GPE is enabled, while another device
having its wakeup power disabled will not wake up the system even
though the GPE is enabled). Rework acpi_system_write_wakeup_device()
so that it only enables wakeup for one device at a time.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
There are two problems with the ACPICA's current implementation of
the global lock acquisition. First, acpi_ev_global_lock_handler(),
which in fact is an interface to the outside of the kernel, doesn't
validate its input, so it only works correctly if the other side
(i.e. the ACPI firmware) is fully specification-compliant (as far
as the global lock is concerned). Unfortunately, that's known not
to be the case on some systems (i.e. we get spurious global lock
signaling interrupts without the pending flag set on some systems).
Second, acpi_ev_global_lock_handler() attempts to acquire the global
lock on behalf of a thread waiting for it without checking if there
actually is such a thread. Both of these shortcomings need to be
addressed to prevent all possible race conditions from happening.
Rework acpi_ev_global_lock_handler() so that it doesn't try to
acquire the global lock and make it signal the availability of the
global lock to the waiting thread instead. Make sure that the
availability of the global lock can only be signaled when there
is a thread waiting for it and that it can't be signaled more than
once in a row (to keep acpi_gbl_global_lock_semaphore in balance).
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
Although the temporary boot-time ACPI table mappings
were set up with CPU caching enabled, the permanent table
mappings and AML run-time region memory accesses were
set up with ioremap(), which on x86 is a synonym for
ioremap_nocache().
Changing this to ioremap_cache() improves performance as
seen when accessing the tables via acpidump,
or /sys/firmware/acpi/tables. It should also improve
AML run-time performance.
No change on ia64.
Reported-by: Jack Steiner <steiner@sgi.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
It turns out that the NVS memory region that suspend_nvs_save()
attempts to map has been already mapped by acpi_os_map_memory(), so
suspend_nvs_save() should better use acpi_os_map_memory() for mapping
memory to avoid conflicts.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
The file information and the list of include in drivers/acpi/nvs.c
are outdated, so update them.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
The saving of the ACPI NVS area during hibernation and suspend and
restoring it during the subsequent resume is entirely specific to
ACPI, so move it to drivers/acpi and drop the CONFIG_SUSPEND_NVS
configuration option which is redundant.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
Move them to radeon_asic.h together with the other asic
specific stuff.
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This is a copy-and-paste bug. We should be comparing 4 characters here
instead of 3.
Signed-off-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
When ioremap() fails (which might happen for some reason), we nicely
oops in suspend_nvs_save() due to NULL dereference by memcpy() in there.
Fail gracefully instead.
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
The vbios only partially initializes the memory controller on
NI, so now we need to load the MC ucode in the driver and set
the default clocks once the ucode is loaded.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The MC ucode is no longer loaded by the vbios
tables as on previous asics. It now must be loaded
by the driver.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The hardware supports advanced user defined color management
but at the moment, there is no infrastructure in place to take
advantage of it so for now we just support the legacy LUTs.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
This adds some new NI (northern islands) specific display
register defines.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
NI chips no longer load the MC ucode in the asic_init sequence so
the asic comes up in a basic mode with low engine/memory clocks and
a voltage. Once the MC ucode is loaded by the driver the card
can be programmed to it's proper default clocks and voltage. As such
the default clocks in the firmware info table as the post clocks, not
the default running clocks. Track the default post clocks and default
running clocks separately to handle this.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Some systems disable the vbios on secondary cards or cards that
have been posted. This code re-enabled the vbios so the driver
can load it.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
Needed to tell which DIG encoders are HBR2 capable for DP 1.2.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The DVOOutputControl table was removed for DCE5.
DVOEncoderControl now handles everything.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
The DAC1OutputControl table was removed for DCE5.
DAC1EncoderControl now handles everything.
Signed-off-by: Alex Deucher <alexdeucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
