I originally submitted a patch to workaround this by pushing all Ejection
Requests and Device Checks onto the kacpi_hotplug queue.
http://marc.info/?l=linux-acpi&m=131678270930105&w=2
The patch is still insufficient in that Bus Checks also need to be added.
Rather than add all events, including non-PCI-hotplug events, to the
hotplug queue, mjg suggested that a better approach would be to modify
the acpiphp driver so only acpiphp events would be added to the
kacpi_hotplug queue.
It's a longer patch, but at least we maintain the benefit of having separate
queues in ACPI. This, of course, is still only a workaround the problem.
As Bjorn and mjg pointed out, we have to refactor a lot of this code to do
the right thing but at this point it is a better to have this code working.
The acpi core places all events on the kacpi_notify queue. When the acpiphp
driver is loaded and a PCI card with a PCI-to-PCI bridge is removed the
following call sequence occurs:
cleanup_p2p_bridge()
-> cleanup_bridge()
-> acpi_remove_notify_handler()
-> acpi_os_wait_events_complete()
-> flush_workqueue(kacpi_notify_wq)
which is the queue we are currently executing on and the process will hang.
Move all hotplug acpiphp events onto the kacpi_hotplug workqueue. In
handle_hotplug_event_bridge() and handle_hotplug_event_func() we can simply
push the rest of the work onto the kacpi_hotplug queue and then avoid the
deadlock.
Signed-off-by: Prarit Bhargava <prarit@redhat.com>
Cc: mjg@redhat.com
Cc: bhelgaas@google.com
Cc: linux-acpi@vger.kernel.org
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Linux supports some optional features, but it should notify the BIOS about
them via the _OSI method. Currently Linux doesn't notify any, which might
make such features not work because the BIOS doesn't know about them.
Jarosz has a system which needs this to make ACPI processor aggregator
device work.
Reported-by: "Jarosz, Sebastian" <sebastian.jarosz@intel.com>
Signed-off-by: Shaohua Li <shaohua.li@intel.com>
Acked-by: Matthew Garrett <mjg@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Len Brown <len.brown@intel.com>
There is a problem with putting the first kernel in EFI virtual mode,
it is that when the second kernel comes up it tries to initialize the
EFI again and once we have put EFI in virtual mode we can not really
do that.
Actually, EFI is not necessary for kdump, we can boot the second kernel
with "noefi" parameter, but the boot will mostly fail because 2nd kernel
cannot find RSDP.
In this situation, we introduced "acpi_rsdp=" kernel parameter, so that
kexec-tools can pass the "noefi acpi_rsdp=X" to the second kernel to
make kdump works. The physical address of the RSDP can be got from
sysfs(/sys/firmware/efi/systab).
Signed-off-by: Takao Indoh <indou.takao@jp.fujitsu.com>
Reviewed-by: WANG Cong <amwang@redhat.com>
Reviewed-by: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
Signed-off-by: Len Brown <len.brown@intel.com>
All ACPICA locks are allocated by the same function,
acpi_os_create_lock(), with the help of a local variable called
"lock". Thus, when lockdep is enabled, it uses "lock" as the
name of all those locks and regards them as instances of the same
lock, which causes it to report possible locking problems with them
when there aren't any.
To work around this problem, define acpi_os_create_lock() as a macro
and make it pass its argument to spin_lock_init(), so that lockdep
uses it as the name of the new lock. Define this macron in a
Linux-specific file, to minimize the resulting modifications of
the OS-independent ACPICA parts.
This change is based on an earlier patch from Andrea Righi and it
addresses a regression from 2.6.39 tracked as
https://bugzilla.kernel.org/show_bug.cgi?id=38152
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Reported-and-tested-by: Borislav Petkov <bp@alien8.de>
Tested-by: Andrea Righi <andrea@betterlinux.com>
Reviewed-by: Florian Mickler <florian@mickler.org>
Signed-off-by: Len Brown <len.brown@intel.com>
In principle acpi_os_install_interrupt_handler() may be called
multiple times for different interrupts, either from
acpi_ev_get_gpe_xrupt_block(), or from acpi_ev_install_sci_handler().
However, it always attempts to request the same interrupt,
acpi_gbl_FADT.sci_interrupt and it doesn't check whether or not this
interrupt has already been requested.
Modify this function so that it refuses to request interrupts other
than acpi_gbl_FADT.sci_interrupt and change
acpi_os_remove_interrupt_handler() so that it refuses to free such
interrupts. Use the observation that the only supported ACPI
interrupt must be equal to acpi_gbl_FADT.sci_interrupt and drop an
unnecessary variable acpi_irq_irq.
This change has been tested on Toshiba Portege R500 and HP nx6325
without introducing any visible problems.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Introduce function acpi_os_get_iomem() that may be used by its callers
to get a reference to an ACPI iomap.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
The reference counting of ACPI iomaps is carried out entirely under
acpi_ioremap_lock, so it is sufficient to use simple counters instead
of krefs for this purpose.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Modify acpi_os_map_memory() so that it doesn't call acpi_os_ioremap()
unconditionally every time it is executed (except when
acpi_gbl_permanent_mmap is unset), which pretty much defeats the
purpose of maintaining the list of ACPI iomaps in osl.c.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Make acpi_os_unmap_generic_address() use acpi_map_lookup() to find
the desired iomap and drop the reference to it directly (and
eventually remove it if necessary) instead of calling
acpi_os_unmap_memory(), which requires us to walk the list of ACPI
iomaps twice in a row (first, to get the virtual address associated
with the iomap and second, to get the iomap itself).
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
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: Rafael 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: Rafael 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: Rafael J. Wysocki <rjw@sisk.pl>
ACPI workqueues aren't used during memory reclaming. Use
alloc_workqueue() to create workqueues w/o rescuers.
If the purpose of the separation between kacpid_wq and kacpi_notify_wq
was to give notifications better response time, kacpi_notify_wq can be
dropped and kacpi_wq can be created with higher @max_active.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Len Brown <lenb@kernel.org>
Cc: linux-acpi@vger.kernel.org
Commit ca9b600be3 ("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: Jeff Chua <jeff.chua.linux@gmail.com>
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
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=26342
Signed-off-by: Thomas Renninger <trenn@suse.de>
CC: Matthew Wilcox <matthew@wil.cx>
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>
WARNING: drivers/acpi/acpi.o(.text+0xeda): Section mismatch in reference from the function acpi_os_initialize1() to the function .init.text:set_osi_linux()
The function acpi_os_initialize1() references
the function __init set_osi_linux().
Signed-off-by: Zhang Rui <rui.zhang@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
commit b0ed7a91(ACPICA/ACPI: Add new host interfaces for _OSI suppor)
introduced another regression that only one _OSI string can be added or
removed.
Now multiple _OSI strings can be added or removed, for example
acpi_osi=Linux acpi_osi=FreeBSD acpi_osi="!Windows 2006"
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
commit b0ed7a91(ACPICA/ACPI: Add new host interfaces for _OSI suppor)
introduced a regression that _OSI string setup fails.
There are 2 paths to setup _OSI string.
DMI:
acpi_dmi_osi_linux -> set_osi_linux -> acpi_osi_setup -> copy _OSI
string to osi_setup_string
Boot command line:
acpi_osi_setup -> copy _OSI string to osi_setup_string
Later, acpi_osi_setup_late will be called to handle osi_setup_string.
If _OSI string is "Linux" or "!Linux", then the call path is,
acpi_osi_setup_late -> acpi_cmdline_osi_linux -> set_osi_linux ->
acpi_osi_setup -> copy _OSI string to osi_setup_string
This actually never installs _OSI string(acpi_install_interface not
called), but just copy the _OSI string to osi_setup_string.
This patch fixes the regression.
Reported-and-tested-by: Lukas Hejtmanek <xhejtman@ics.muni.cz>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
This patch optimizes ACPI MMIO remappings by keeping track of the
remappings on a PAGE_SIZE granularity.
When an ioremap() occurs, the underlying infrastructure works on a 'page'
based granularity. As such, an ioremap() request for 1 byte for example,
will end up mapping in an entire (PAGE_SIZE) page. Huang Ying took
advantage of this in commit 15651291a2 by
checking if subsequent ioremap() requests reside within any of the list's
existing remappings still in place, and if so, incrementing a reference
count on the existing mapping as opposed to performing another ioremap().
Signed-off-by: Myron Stowe <myron.stowe@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Convert the simple locking introduced earlier for the ACPI MMIO
remappings list to an RCU based locking scheme.
Signed-off-by: Myron Stowe <myron.stowe@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
During ACPI initialization, pre-map fixed hardware registers that are
accessed during ACPI's 'system event' related IRQ handing.
ACPI's 'system event' handing accesses specific fixed hardware
registers; namely PM1a event, PM1b event, GPE0, and GPE1 register
blocks which are declared within the FADT. If these registers are
backed by MMIO, as opposed to I/O port space, accessing them within
interrupt context will cause a panic as acpi_os_read_memory()
depends on ioremap() in such cases - BZ 18012.
By utilizing the functionality provided in the previous two patches -
ACPI: Maintain a list of ACPI memory mapped I/O remappings, and, ACPI:
Add interfaces for ioremapping/iounmapping ACPI registers - accesses
to ACPI MMIO areas will now be safe from within interrupt contexts (IRQ
and/or NMI) provided the area was pre-mapped. This solves BZ 18012.
ACPI "System Event" reference(s):
ACPI Specification, Revision 4.0, Section 3 "ACPI Overview",
3.8 "System Events", 5.6 "ACPI Event Programming Model".
Reference: https://bugzilla.kernel.org/show_bug.cgi?id=18012
Reported-by: <bjorn.helgaas@hp.com>
Signed-off-by: Myron Stowe <myron.stowe@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Add remapping and unmapping interfaces for ACPI registers that are
backed by memory mapped I/O (MMIO). These interfaces, along with
the MMIO remapping list, enable accesses of such registers from within
interrupt context.
ACPI Generic Address Structure (GAS) reference (ACPI's fixed/generic
hardware registers use the GAS format):
ACPI Specification, Revision 4.0, Section 5.2.3.1, "Generic Address
Structure".
Signed-off-by: Myron Stowe <myron.stowe@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
For memory mapped I/O (MMIO) remappings, add a list to maintain the
remappings and augment the corresponding mapping and unmapping interface
routines (acpi_os_map_memory() and acpi_os_unmap_memory()) to
dynamically add to, and delete from, the list.
The current ACPI I/O accessing methods - acpi_read() and acpi_write() -
end up calling ioremap() when accessing MMIO. This prevents use of these
methods within interrupt context (IRQ and/or NMI), since ioremap() may
block to allocate memory. Maintaining a list of MMIO remappings enables
accesses to such areas from within interrupt context provided they have
been pre-mapped.
Signed-off-by: Myron Stowe <myron.stowe@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
The size used for I/O remapping MMIO read and write accesses has not
accounted for the basis of ACPI's Generic Address Structure (GAS)
'Register Bit Width' field which is bits, not bytes. This patch
adjusts the ioremap() 'size' argument accordingly.
ACPI "Generic Register" reference:
ACPI Specification, Revision 4.0, Section 5.2.3.1, "Generic Address
Structure".
Signed-off-by: Myron Stowe <myron.stowe@hp.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Found by running make namespacecheck on linux-next
Signed-off-by: Stephen Hemminger <shemminger@vyatta.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Linux-2.6.22 initiated a dmesg complaint when it saw BIOS that invoked
OSI(Linux). Linux-2.6.23 continued that complaint and started our
policy of ignoring the bogus BIOS request.
Past-time for Linux to label that complaint with FW_BUG.
Signed-off-by: Len Brown <len.brown@intel.com>
This function is not OS-dependent and has been replaced by
acpi_hw_derive_pci_id, which is now in the ACPICA core code. Local
implementations of acpi_os_derive_pci_id are no longer necessary and
are removed. ACPICA BZ 857.
http://www.acpica.org/bugzilla/show_bug.cgi?id=857
Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Adds install/remove interfaces so that the host can dynamically
alter the global _OSI table. Also adds support for _OSI handlers.
Additional support: new debugger command (osi), and test support in
the acpiexec utility. Adds new file, utilities/utosi.c.
ACPICA bugzilla 836.
The Linux OSL _OSI code is also changed.
acpi_osi_setup can't call acpi_install/remove_interface because ACPICA
is not initialized yet at this early time.
So we just save the osi string in acpi_osi_setup and will handle it
later in a new function acpi_osi_setup_late.
http://www.acpica.org/bugzilla/show_bug.cgi?id=836
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Bob Moore <robert.moore@intel.com
Signed-off-by: Len Brown <len.brown@intel.com>
Prototype in acpiosxf.h had the output value pointer as a (u32 *).
Should be a (u64 *).
Signed-off-by: Bob Moore <robert.moore@intel.com>
Signed-off-by: Lin Ming <ming.m.lin@intel.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Some minor improvements in error handling, but overall it was mostly dead
code.
Signed-off-by: Andi Kleen <ak@linux.intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Len Brown <len.brown@intel.com>
* 'acpica' of git://git.kernel.org/pub/scm/linux/kernel/git/lenb/linux-acpi-2.6: (27 commits)
ACPI / ACPICA: Simplify acpi_ev_initialize_gpe_block()
ACPI / ACPICA: Fail acpi_gpe_wakeup() if ACPI_GPE_CAN_WAKE is unset
ACPI / ACPICA: Do not execute _PRW methods during initialization
ACPI: Fix bogus GPE test in acpi_bus_set_run_wake_flags()
ACPICA: Update version to 20100702
ACPICA: Fix for Alias references within Package objects
ACPICA: Fix lint warning for 64-bit constant
ACPICA: Remove obsolete GPE function
ACPICA: Update debug output components
ACPICA: Add support for WDDT - Watchdog Descriptor Table
ACPICA: Drop acpi_set_gpe
ACPICA: Use low-level GPE enable during GPE block initialization
ACPI / EC: Do not use acpi_set_gpe
ACPI / EC: Drop suspend and resume routines
ACPICA: Remove wakeup GPE reference counting which is not used
ACPICA: Introduce acpi_gpe_wakeup()
ACPICA: Rename acpi_hw_gpe_register_bit
ACPICA: Update version to 20100528
ACPICA: Add signatures for undefined tables: ATKG, GSCI, IEIT
ACPICA: Optimization: Reduce the number of namespace walks
...
Currently, during initialization ACPICA walks the entire ACPI
namespace in search of any device objects with assciated _PRW
methods. All of the _PRW methods found are executed in the process
to extract the GPE information returned by them, so that the GPEs in
question can be marked as "able to wakeup" (more precisely, the
ACPI_GPE_CAN_WAKE flag is set for them). The only purpose of this
exercise is to avoid enabling the CAN_WAKE GPEs automatically, even
if there are _Lxx/_Exx methods associated with them. However, it is
both costly and unnecessary, because the host OS has to execute the
_PRW methods anyway to check which devices can wake up the system
from sleep states. Moreover, it then uses full information
returned by _PRW, including the GPE information, so it can take care
of disabling the GPEs if necessary.
Remove the code that walks the namespace and executes _PRW from
ACPICA and modify comments to reflect that change. Make
acpi_bus_set_run_wake_flags() disable GPEs for wakeup devices
so that they don't cause spurious wakeup events to be signaled.
This not only reduces the complexity of the ACPICA initialization
code, but in some cases it should reduce the kernel boot time as
well.
Unfortunately, for this purpose we need a new ACPICA function,
acpi_gpe_can_wake(), to be called by the host OS in order to disable
the GPEs that can wake up the system and were previously enabled by
acpi_ev_initialize_gpe_block() or acpi_ev_update_gpes() (such a GPE
should be disabled only once, because the initialization code enables
it only once, but it may be pointed to by _PRW for multiple devices
and that's why the additional function is necessary).
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Len Brown <len.brown@intel.com>
ACPI works need to be executed on cpu0 and acpi/osl.c achieves this by
creating singlethread workqueue and then binding it to cpu0 from a
work which is quite unorthodox. Make it create regular workqueues and
use queue_work_on() instead. This is in preparation of concurrency
managed workqueue and the extra workers won't be a problem after it's
implemented.
Signed-off-by: Tejun Heo <tj@kernel.org>
When the user passes the kernel parameter acpi_enforce_resources=lax,
the ACPI resources are no longer protected, so a native driver can
make use of them. In that case, we do not want the asus_atk0110 to be
loaded. Unfortunately, this driver loads automatically due to its
MODULE_DEVICE_TABLE, so the user ends up with two drivers loaded for
the same device - this is bad.
So I suggest that we prevent the asus_atk0110 driver from loading if
acpi_enforce_resources=lax.
Signed-off-by: Jean Delvare <khali@linux-fr.org>
Acked-by: Luca Tettamanti <kronos.it@gmail.com>
Cc: Len Brown <lenb@kernel.org>
By default, ACPI resource conflict messages are logged at level
KERN_ERR. This is a rather high level for a message that is more a
warning than an indication of a real kernel error. Also, KERN_ERR level
messages can appear over some boot splash screens, and this message is
not serious enough to warrant such treatment. Thus, the log level has
been reduced to KERN_WARNING.
[lenb changed to KERN_WARNING rather than all the way to KERN_INFO]
Also, cleanup message to use %pR resource printing format.
Signed-off-by: Chase Douglas <chase.douglas@canonical.com>
Signed-off-by: Len Brown <len.brown@intel.com>
Prarit Bhargava