When an ECDT table is present the EC address space handler gets registered
on the root node. So to unregister it properly the unregister call also
must be done on the root node.

Store the ACPI handle used for the acpi_install_address_space_handler()
call and use te same handle for the acpi_remove_address_space_handler()
call.

Reported-by: Rafael J. Wysocki <rafael@kernel.org>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Instead of having acpi_device_add() defined as a wrapper around
__acpi_device_add(), export acpi_tie_acpi_dev() so it can be called
directly by acpi_add_power_resource(), fold acpi_device_add() into the
latter and rename __acpi_device_add() to acpi_device_add().

No intentional functional impact.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Punit Agrawal <punit.agrawal@bytedance.com>

The initialization of ACPI device objects is split between
acpi_init_device_object() and __acpi_device_add() that initializes
the dev field in struct acpi_device.  The "release" function pointer
is passed to __acpi_device_add() for this reason.

However, that split is artificial and all of the initialization can
be carried out by acpi_init_device_object(), so rearrange the code
to that end.  In particular, make acpi_init_device_object() take the
"release" pointer as an argument, along with the "type" which is
related to it, instead of __acpi_device_add().

No intentional functional impact.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Punit Agrawal <punit.agrawal@bytedance.com>

As recommended in include/linux/idr.h, use ida_alloc() instead of
ida_simple_get() for creating unique device object names and for
symmetry replace ida_simple_remove() with ida_free() (and fix up
the related overly long code line while at it).

Also drop the ACPI_MAX_DEVICE_INSTANCES limit that is not necessary
any more and may not be sufficient for future platforms.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>

The only caller of acpi_scan_init(), acpi_init(), doesn't check its
return value, so turn it into a void function.

This avoids complaints from the Smatch static checker that the
function should return a negative error code when it fails, which
is not really a problem in this particular case.

No intentional functional impact.

Link: https://lore.kernel.org/linux-acpi/20220106082317.GA9123@kili/
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>

The EC driver uses a relatively simple state machine for the event
work handling, but it is not really straightforward to figure out.

The states are as follows:

 "Ready": The event handling work can be submitted.

  In this state, the EC_FLAGS_QUERY_PENDING flag is clear.

 "In progress": The event handling work is pending or is being
                processed.  It cannot be submitted again.

  In ths state, the EC_FLAGS_QUERY_PENDING flag is set and both the
  events_to_process count is nonzero and the EC_FLAGS_QUERY_GUARDING
  flag is clear.

 "Complete": The event handling work has been completed, but it still
             cannot be submitted again.

  In ths state, the EC_FLAGS_QUERY_PENDING flag is set and the
  events_to_process count is zero or the EC_FLAGS_QUERY_GUARDING
  flag is set.

The state changes from "Ready" to "In progress" when new event is
detected by advance_transaction() and acpi_ec_submit_event() is
called by it.

Next, the state can change from "In progress" directly to "Ready" in
the following situations:

 * ec_event_clearing is ACPI_EC_EVT_TIMING_STATUS and the state of
   an ACPI_EC_COMMAND_QUERY transaction becomes ACPI_EC_COMMAND_POLL.

 * ec_event_clearing is ACPI_EC_EVT_TIMING_QUERY and the state of
   an ACPI_EC_COMMAND_QUERY transaction becomes
   ACPI_EC_COMMAND_COMPLETE.

 * ec_event_clearing is either ACPI_EC_EVT_TIMING_STATUS or
   ACPI_EC_EVT_TIMING_QUERY and there are no more events to
   process (ie. ec->events_to_process becomes 0).

If ec_event_clearing is ACPI_EC_EVT_TIMING_EVENT, however, the
state must change from "In progress" to "Complete" before it
can change to "Ready".  The changes from "In progress" to
"Complete" in that case occur in the following situations:

 * The state of an ACPI_EC_COMMAND_QUERY transaction becomes
   ACPI_EC_COMMAND_COMPLETE.

 * There are no more events to process (ie. ec->events_to_process
   becomes 0).

Finally, the state changes from "Complete" to "Ready" when
advance_transaction() is invoked when the state is "Complete" and
the state of the current transaction is not ACPI_EC_COMMAND_POLL.

To make this state machine visible in the code, add a new
event_state field to struct acpi_ec and modify the code to use
it istead the EC_FLAGS_QUERY_PENDING and EC_FLAGS_QUERY_GUARDING
flags.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Notice that it is not necessary to queue up the event work again
if the while () loop in acpi_ec_event_handler() is still running
which is the case if nr_pending_queries is greater than 0 at the
beginning of acpi_ec_submit_event() and modify the code to avoid
doing that.

While at it, rename nr_pending_queries in struct acpi_ec to
events_to_process which actually matches the role of that field
and change its data type to unsigned int which is sufficient.

No expected functional impact.

Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

The flushing of pending work in the EC driver uses drain_workqueue()
to flush the event handling work that can requeue itself via
advance_transaction(), but this is problematic, because that
work may also be requeued from the query workqueue.

Namely, if an EC transaction is carried out during the execution of
a query handler, it involves calling advance_transaction() which
may queue up the event handling work again.  This causes the kernel
to complain about attempts to add a work item to the EC event
workqueue while it is being drained and worst-case it may cause a
valid event to be skipped.

To avoid this problem, introduce two new counters, events_in_progress
and queries_in_progress, incremented when a work item is queued on
the event workqueue or the query workqueue, respectively, and
decremented at the end of the corresponding work function, and make
acpi_ec_dispatch_gpe() the workqueues in a loop until the both of
these counters are zero (or system wakeup is pending) instead of
calling acpi_ec_flush_work().

At the same time, change __acpi_ec_flush_work() to call
flush_workqueue() instead of drain_workqueue() to flush the event
workqueue.

While at it, use the observation that the work item queued in
acpi_ec_query() cannot be pending at that time, because it is used
only once, to simplify the code in there.

Additionally, clean up a comment in acpi_ec_query() and adjust white
space in acpi_ec_event_processor().

Fixes: f0ac20c3f613 ("ACPI: EC: Fix flushing of pending work")
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

Revert commit c10383e8ddf4 ("ACPI: scan: Release PM resources blocked
by unused objects"), because it causes boot issues to appear on some
platforms.

Reported-by: Kyle D. Pelton <kyle.d.pelton@intel.com>
Reported-by: Saranya Gopal <saranya.gopal@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>

On some systems the ACPI namespace contains device objects that are
not used in certain configurations of the system.  If they start off
in the D0 power state configuration, they will stay in it until the
system reboots, because of the lack of any mechanism possibly causing
their configuration to change.  If that happens, they may prevent
some power resources from being turned off or generally they may
prevent the platform from getting into the deepest low-power states
thus causing some energy to be wasted.

Address this issue by changing the configuration of unused ACPI
device objects to the D3cold power state one after carrying out
the ACPI-based enumeration of devices.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=214091
Link: https://lore.kernel.org/linux-acpi/20211007205126.11769-1-mario.limonciello@amd.com/
Reported-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Tested-by: Mario Limonciello <mario.limonciello@amd.com>