CPU system maps are protected with reader/writer locks. The reader
lock, get_online_cpus(), assures that the maps are not updated while
holding the lock. The writer lock, cpu_hotplug_begin(), is used to
udpate the cpu maps along with cpu_maps_update_begin().
However, the ACPI processor handler updates the cpu maps without
holding the the writer lock.
acpi_map_lsapic() is called from acpi_processor_hotadd_init() to
update cpu_possible_mask and cpu_present_mask. acpi_unmap_lsapic()
is called from acpi_processor_remove() to update cpu_possible_mask.
Currently, they are either unprotected or protected with the reader
lock, which is not correct.
For example, the get_online_cpus() below is supposed to assure that
cpu_possible_mask is not changed while the code is iterating with
for_each_possible_cpu().
get_online_cpus();
for_each_possible_cpu(cpu) {
:
}
put_online_cpus();
However, this lock has no protection with CPU hotplug since the ACPI
processor handler does not use the writer lock when it updates
cpu_possible_mask. The reader lock does not serialize within the
readers.
This patch protects them with the writer lock with cpu_hotplug_begin()
along with cpu_maps_update_begin(), which must be held before calling
cpu_hotplug_begin(). It also protects arch_register_cpu() /
arch_unregister_cpu(), which creates / deletes a sysfs cpu device
interface. For this purpose it changes cpu_hotplug_begin() and
cpu_hotplug_done() to global and exports them in cpu.h.
Signed-off-by: Toshi Kani <toshi.kani@hp.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
try_offline_node() checks that all CPUs associated with the given
node have been removed by using cpu_present_bits. If all cpus
related to that node have been removed, try_offline_node() clears
the node information.
However, try_offline_node() called from acpi_processor_remove() never
clears the node information. For disabling cpu_present_bits,
acpi_unmap_lsapic() needs be called. Yet, acpi_unmap_lsapic() is
called after try_offline_node() has run. So when try_offline_node()
runs, the CPU's cpu_present_bits is always set.
Fix the issue by moving try_offline_node() after acpi_unmap_lsapic().
The problem fixed here was uncovered by commit cecdb19 "ACPI / scan:
Change the implementation of acpi_bus_trim()".
[rjw: Changelog]
Signed-off-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Acked-by: Toshi Kani <toshi.kani@hp.com>
Cc: 3.9+ <stable@vger.kernel.org> # 3.9+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
The __cpuinit type of throwaway sections might have made sense
some time ago when RAM was more constrained, but now the savings
do not offset the cost and complications. For example, the fix in
commit 5e427ec2d0 ("x86: Fix bit corruption at CPU resume time")
is a good example of the nasty type of bugs that can be created
with improper use of the various __init prefixes.
After a discussion on LKML[1] it was decided that cpuinit should go
the way of devinit and be phased out. Once all the users are gone,
we can then finally remove the macros themselves from linux/init.h.
This removes all the drivers/acpi uses of the __cpuinit macros
from all C files.
[1] https://lkml.org/lkml/2013/5/20/589
Cc: Len Brown <lenb@kernel.org>
Cc: "Rafael J. Wysocki" <rjw@sisk.pl>
Cc: linux-acpi@vger.kernel.org
Signed-off-by: Paul Gortmaker <paul.gortmaker@windriver.com>
In acpi_processor_add(), get_cpu_device() may return NULL in some cases
which is then passed to acpi_bind_one() and that will case a NULL
pointer dereference to occur.
Add a check to prevent that from happening.
[rjw: Changelog]
Signed-off-by: Hanjun Guo <hanjun.guo@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Make acpi_processor_add() pass the ACPI handle of the processor
namespace object to acpi_bind_one() instead of setting it directly
to allow acpi_bind_one() to catch possible bugs causing the ACPI
handle of the processor device to be set earlier.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Toshi Kani <toshi.kani@hp.com>
Commit ac212b6 (ACPI / processor: Use common hotplug infrastructure)
forgot about initializing the per-CPU 'processors' variables which
lead to ACPI cpuidle failure to use C-states and caused boot slowdown
on multi-CPU machines.
Fix the problem by adding per_cpu(processors, pr->id) initialization
to acpi_processor_add() and add make acpi_processor_remove() clean it
up as appropriate.
Also modify acpi_processor_stop() so that it doesn't clear
per_cpu(processors, pr->id) on processor driver removal which would
then cause problems to happen when the driver is loaded again.
This version of the patch contains fixes from Yinghai Lu.
Reported-and-tested-by: Yinghai Lu <yinghai@kernel.org>
Reported-and-tested-by: Daniel Lezcano <daniel.lezcano@linaro.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Split the ACPI processor driver into two parts, one that is
non-modular, resides in the ACPI core and handles the enumeration
and hotplug of processors and one that implements the rest of the
existing processor driver functionality.
The non-modular part uses an ACPI scan handler object to enumerate
processors on the basis of information provided by the ACPI namespace
and to hook up with the common ACPI hotplug infrastructure. It also
populates the ACPI handle of each processor device having a
corresponding object in the ACPI namespace, which allows the driver
proper to bind to those devices, and makes the driver bind to them
if it is readily available (i.e. loaded) when the scan handler's
.attach() routine is running.
There are a few reasons to make this change.
First, switching the ACPI processor driver to using the common ACPI
hotplug infrastructure reduces code duplication and size considerably,
even though a new file is created along with a header comment etc.
Second, since the common hotplug code attempts to offline devices
before starting the (non-reversible) removal procedure, it will abort
(and possibly roll back) hot-remove operations involving processors
if cpu_down() returns an error code for one of them instead of
continuing them blindly (if /sys/firmware/acpi/hotplug/force_remove
is unset). That is a more desirable behavior than what the current
code does.
Finally, the separation of the scan/hotplug part from the driver
proper makes it possible to simplify the driver's .remove() routine,
because it doesn't need to worry about the possible cleanup related
to processor removal any more (the scan/hotplug part is responsible
for that now) and can handle device removal and driver removal
symmetricaly (i.e. as appropriate).
Some user-visible changes in sysfs are made (for example, the
'sysdev' link from the ACPI device node to the processor device's
directory is gone and a 'physical_node' link is present instead
and a corresponding 'firmware_node' is present in the processor
device's directory, the processor driver is now visible under
/sys/bus/cpu/drivers/ and bound to the processor device), but
that shouldn't affect the functionality that users care about
(frequency scaling, C-states and thermal management).
Tested on my venerable Toshiba Portege R500.
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Toshi Kani <toshi.kani@hp.com>
Toshi Kani