Bug 1852989: PM: sleep: Simplify suspend-to-idle control flow
17a77a7...
by
"Rafael J. Wysocki" <email address hidden>
ACPI: EC: Rework flushing of pending work
There is a race condition in the ACPI EC driver, between
__acpi_ec_flush_event() and acpi_ec_event_handler(), that may
cause systems to stay in suspended-to-idle forever after a wakeup
event coming from the EC.
Namely, acpi_s2idle_wake() calls acpi_ec_flush_work() to wait until
the delayed work resulting from the handling of the EC GPE in
acpi_ec_dispatch_gpe() is processed, and that function invokes
__acpi_ec_flush_event() which uses wait_event() to wait for
ec->nr_pending_queries to become zero on ec->wait, and that wait
queue may be woken up too early.
Suppose that acpi_ec_dispatch_gpe() has caused acpi_ec_gpe_handler()
to run, so advance_transaction() has been called and it has invoked
acpi_ec_submit_query() to queue up an event work item, so
ec->nr_pending_queries has been incremented (under ec->lock). The
work function of that work item, acpi_ec_event_handler() runs later
and calls acpi_ec_query() to process the event. That function calls
acpi_ec_transaction() which invokes acpi_ec_transaction_unlocked()
and the latter wakes up ec->wait under ec->lock, but it drops that
lock before returning.
When acpi_ec_query() returns, acpi_ec_event_handler() acquires
ec->lock and decrements ec->nr_pending_queries, but at that point
__acpi_ec_flush_event() (woken up previously) may already have
acquired ec->lock, checked the value of ec->nr_pending_queries (and
it would not have been zero then) and decided to go back to sleep.
Next, if ec->nr_pending_queries is equal to zero now, the loop
in acpi_ec_event_handler() terminates, ec->lock is released and
acpi_ec_check_event() is called, but it does nothing unless
ec_event_clearing is equal to ACPI_EC_EVT_TIMING_EVENT (which is
not the case by default). In the end, if no more event work items
have been queued up while executing acpi_ec_transaction_unlocked(),
there is nothing to wake up __acpi_ec_flush_event() again and it
sleeps forever, so the suspend-to-idle loop cannot make progress and
the system is permanently suspended.
To avoid this issue, notice that it actually is not necessary to
wait for ec->nr_pending_queries to become zero in every case in
which __acpi_ec_flush_event() is used.
First, during platform-based system suspend (not suspend-to-idle),
__acpi_ec_flush_event() is called by acpi_ec_disable_event() after
clearing the EC_FLAGS_QUERY_ENABLED flag, which prevents
acpi_ec_submit_query() from submitting any new event work items,
so calling flush_scheduled_work() and flushing ec_query_wq
subsequently (in order to wait until all of the queries in that
queue have been processed) would be sufficient to flush all of
the pending EC work in that case.
Second, the purpose of the flushing of pending EC work while
suspended-to-idle described above really is to wait until the
first event work item coming from acpi_ec_dispatch_gpe() is
complete, because it should produce system wakeup events if
that is a valid EC-based system wakeup, so calling
flush_scheduled_work() followed by flushing ec_query_wq is also
sufficient for that purpose.
Rework the code to follow the above observations.
Fixes: 56b9918490 ("PM: sleep: Simplify suspend-to-idle control flow")
Reported-by: Kenneth R. Crudup <email address hidden>
Tested-by: Kenneth R. Crudup <email address hidden>
Cc: 5.4+ <email address hidden> # 5.4+
Signed-off-by: Rafael J. Wysocki <email address hidden>
(cherry picked from commit 016b87ca5c8c6e9e87db442f04dc99609b11ed36)
Signed-off-by: You-Sheng Yang <email address hidden>
e1582c5...
by
"Rafael J. Wysocki" <email address hidden>
PM: sleep: Simplify suspend-to-idle control flow
After commit 33e4f80ee69b ("ACPI / PM: Ignore spurious SCI wakeups
from suspend-to-idle") the "noirq" phases of device suspend and
resume may run for multiple times during suspend-to-idle, if there
are spurious system wakeup events while suspended. However, this
is complicated and fragile and actually unnecessary.
The main reason for doing this is that on some systems the EC may
signal system wakeup events (power button events, for example) as
well as events that should not cause the system to resume (spurious
system wakeup events). Thus, in order to determine whether or not
a given event signaled by the EC while suspended is a proper system
wakeup one, the EC GPE needs to be dispatched and to start with that
was achieved by allowing the ACPI SCI action handler to run, which
was only possible after calling resume_device_irqs().
However, dispatching the EC GPE this way turned out to take too much
time in some cases and some EC events might be missed due to that, so
commit 68e22011856f ("ACPI: EC: Dispatch the EC GPE directly on
s2idle wake") started to dispatch the EC GPE right after a wakeup
event has been detected, so in fact the full ACPI SCI action handler
doesn't need to run any more to deal with the wakeups coming from the
EC.
Use this observation to simplify the suspend-to-idle control flow
so that the "noirq" phases of device suspend and resume are each
run only once in every suspend-to-idle cycle, which is reported to
significantly reduce power drawn by some systems when suspended to
idle (by allowing them to reach a deep platform-wide low-power state
through the suspend-to-idle flow). [What appears to happen is that
the "noirq" resume of devices after a spurious EC wakeup brings some
devices into a state in which they prevent the platform from reaching
the deep low-power state going forward, even after a subsequent
"noirq" suspend phase, and on some systems the EC triggers such
wakeups already when the "noirq" suspend of devices is running for
the first time in the given suspend/resume cycle, so the platform
cannot reach the deep low-power state at all.]
First, make acpi_s2idle_wake() use the acpi_ec_dispatch_gpe() return
value to determine whether or not the wakeup may have been triggered
by the EC (in which case the system wakeup is canceled and ACPI
events are processed in order to determine whether or not the event
is a proper system wakeup one) and use rearm_wake_irq() (introduced
by a previous change) in it to rearm the ACPI SCI for system wakeup
detection in case the system will remain suspended.
Second, drop acpi_s2idle_sync(), which is not needed any more, and
the corresponding global platform suspend-to-idle callback.
Next, drop the pm_wakeup_pending() check (which is an optimization
only) from __device_suspend_noirq() to prevent it from returning
errors on system wakeups occurring before the "noirq" phase of
device suspend is complete (as in the case of suspend-to-idle it is
not known whether or not these wakeups are suprious at that point),
in order to avoid having to carry out a "noirq" resume of devices
on a spurious system wakeup.
Finally, change the code flow in s2idle_loop() to (1) run the
"noirq" suspend of devices once before starting the loop, (2) check
for spurious EC wakeups (via the platform ->wake callback) for the
first time before calling s2idle_enter(), and (3) run the "noirq"
resume of devices once after leaving the loop.
Signed-off-by: Rafael J. Wysocki <email address hidden>
Acked-by: Thomas Gleixner <email address hidden>
(cherry picked from commit 56b991849009f5def0443bfb2f48c8321d888e15)
Signed-off-by: You-Sheng Yang <email address hidden>
192be13...
by
"Rafael J. Wysocki" <email address hidden>
ACPI: PM: Set s2idle_wakeup earlier and clear it later
The role of the s2idle_wakeup variable is to cause
acpi_pm_wakeup_event() and acpi_pm_notify_handler() to
increment pm_abort_suspend and trigger a wakeup from
suspend-to-idle in case the ACPI SCI wakeup was canceled
by acpi_s2idle_wake().
However, for this purpose it need not be set in acpi_s2idle_wake()
and cleared in acpi_s2idle_sync(), respectively. In fact, it
may be set as early as in acpi_s2idle_prepare() and cleared as
late as in acpi_s2idle_restore(), so do that to allow subsequent
changes to be simpler.
This change is not expected to alter functionality.
Signed-off-by: Rafael J. Wysocki <email address hidden>
Acked-by: Thomas Gleixner <email address hidden>
(cherry picked from commit 41275eb5c7181febdfaa63c3a0ad9b7acdadcd52)
Signed-off-by: You-Sheng Yang <email address hidden>
c3fef4c...
by
"Rafael J. Wysocki" <email address hidden>
ACPI: EC: Return bool from acpi_ec_dispatch_gpe()
On some systems, if suspend-to-idle is used, the EC may signal system
wakeup events (power button events, for example) as well as events
that should not cause the system to resume and acpi_ec_dispatch_gpe()
needs to be called to determine whether or not the system should
resume then. In particular, if acpi_ec_dispatch_gpe() doesn't detect
any EC events at all, the system should remain suspended, so it is
useful to know when that is the case.
For this reason, make acpi_ec_dispatch_gpe() return a bool value
indicating whether or not any EC events have been detected by it.
Signed-off-by: Rafael J. Wysocki <email address hidden>
Acked-by: Thomas Gleixner <email address hidden>
(cherry picked from commit 9089f16e053afc5e18feaeb9f64cc7c90d6bd687)
Signed-off-by: You-Sheng Yang <email address hidden>
475193c...
by
"Rafael J. Wysocki" <email address hidden>
ACPICA: Return u32 from acpi_dispatch_gpe()
In some cases it is useful to know whether or not the
acpi_ev_detect_gpe() called by acpi_dispatch_gpe() has found
the GPE to be active, so return the return value of it (whose
data type is u32) from latter.
Signed-off-by: Rafael J. Wysocki <email address hidden>
Acked-by: Thomas Gleixner <email address hidden>
(cherry picked from commit 6921de898ba8f2ec91cfea70e7160b89c477382e)
Signed-off-by: You-Sheng Yang <email address hidden>
b32b7ff...
by
"Rafael J. Wysocki" <email address hidden>
PCI: irq: Introduce rearm_wake_irq()
Introduce a new function, rearm_wake_irq(), allowing a wakeup IRQ
to be armed for systen wakeup detection again without running any
action handlers associated with it after it has been armed for
wakeup detection and triggered.
That is useful for IRQs, like ACPI SCI, that may deliver wakeup
as well as non-wakeup interrupts when armed for systen wakeup
detection. In those cases, it may be possible to determine whether
or not the delivered interrupt is a systen wakeup one without
running the entire action handler (or handlers, if the IRQ is
shared) for the IRQ, and if the interrupt turns out to be a
non-wakeup one, the IRQ can be rearmed with the help of the
new function.
Signed-off-by: Rafael J. Wysocki <email address hidden>
Acked-by: Thomas Gleixner <email address hidden>
(cherry picked from commit 3a79bc63d90750f737ab9d7219bd3091d2fd6d84)
Signed-off-by: You-Sheng Yang <email address hidden>
The ARMv8 64-bit architecture supports execute-only user permissions by
clearing the PTE_USER and PTE_UXN bits, practically making it a mostly
privileged mapping but from which user running at EL0 can still execute.
The downside, however, is that the kernel at EL1 inadvertently reading
such mapping would not trip over the PAN (privileged access never)
protection.
Revert the relevant bits from commit cab15ce604e5 ("arm64: Introduce
execute-only page access permissions") so that PROT_EXEC implies
PROT_READ (and therefore PTE_USER) until the architecture gains proper
support for execute-only user mappings.
Align SOF HDA implementation with snd-hda-intel driver and enable
sync_write flag for all supported Intel platforms in SOF. When set,
a sync is issued after each verb write.
Sync after write has helped to overcome intermittent delays in
system resume flow on Intel Coffee Lake systems, and most recently
probe errors related to the HDMI codec on Ice Lake systems.
Matches the snd-hda-intel driver change done in commit 2756d9143aa5
("ALSA: hda - Fix intermittent CORB/RIRB stall on Intel chips").
Signed-off-by: Kai Vehmanen <email address hidden>
Signed-off-by: Pierre-Louis Bossart <email address hidden>
Link: https://<email address hidden>
Signed-off-by: Mark Brown <email address hidden>
(cherry picked from commit f3416e7144f5d4ba0fc5dcef6ebfff891266c46a)
Signed-off-by: Hui Wang <email address hidden>
Acked-by: Stefan Bader <email address hidden>
Acked-by: Kleber Sacilotto de Souza <email address hidden>
Signed-off-by: Kleber Sacilotto de Souza <email address hidden>
