From e0359f1551b8d4a8d00704699c07fabb11a07cf1 Mon Sep 17 00:00:00 2001 From: "Rafael J. Wysocki" Date: Thu, 22 Feb 2024 10:35:54 +0100 Subject: Revert "ACPI: EC: Use a spin lock without disabing interrupts" Commit eb9299beadbd ("ACPI: EC: Use a spin lock without disabing interrupts") introduced an unexpected user-visible change in behavior, which is a significant CPU load increase when the EC is in use. This most likely happens due to increased spinlock contention and so reducing this effect would require a major rework of the EC driver locking. There is no time for this in the current cycle, so revert commit eb9299beadbd. Closes: https://bugzilla.kernel.org/show_bug.cgi?id=218511 Reported-by: Dieter Mummenschanz Signed-off-by: Rafael J. Wysocki --- drivers/acpi/ec.c | 112 ++++++++++++++++++++++++++++++++---------------------- 1 file changed, 66 insertions(+), 46 deletions(-) diff --git a/drivers/acpi/ec.c b/drivers/acpi/ec.c index dbdee2924594..02255795b800 100644 --- a/drivers/acpi/ec.c +++ b/drivers/acpi/ec.c @@ -525,10 +525,12 @@ static void acpi_ec_clear(struct acpi_ec *ec) static void acpi_ec_enable_event(struct acpi_ec *ec) { - spin_lock(&ec->lock); + unsigned long flags; + + spin_lock_irqsave(&ec->lock, flags); if (acpi_ec_started(ec)) __acpi_ec_enable_event(ec); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); /* Drain additional events if hardware requires that */ if (EC_FLAGS_CLEAR_ON_RESUME) @@ -544,9 +546,11 @@ static void __acpi_ec_flush_work(void) static void acpi_ec_disable_event(struct acpi_ec *ec) { - spin_lock(&ec->lock); + unsigned long flags; + + spin_lock_irqsave(&ec->lock, flags); __acpi_ec_disable_event(ec); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); /* * When ec_freeze_events is true, we need to flush events in @@ -567,9 +571,10 @@ void acpi_ec_flush_work(void) static bool acpi_ec_guard_event(struct acpi_ec *ec) { + unsigned long flags; bool guarded; - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, flags); /* * If firmware SCI_EVT clearing timing is "event", we actually * don't know when the SCI_EVT will be cleared by firmware after @@ -585,29 +590,31 @@ static bool acpi_ec_guard_event(struct acpi_ec *ec) guarded = ec_event_clearing == ACPI_EC_EVT_TIMING_EVENT && ec->event_state != EC_EVENT_READY && (!ec->curr || ec->curr->command != ACPI_EC_COMMAND_QUERY); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); return guarded; } static int ec_transaction_polled(struct acpi_ec *ec) { + unsigned long flags; int ret = 0; - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, flags); if (ec->curr && (ec->curr->flags & ACPI_EC_COMMAND_POLL)) ret = 1; - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); return ret; } static int ec_transaction_completed(struct acpi_ec *ec) { + unsigned long flags; int ret = 0; - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, flags); if (ec->curr && (ec->curr->flags & ACPI_EC_COMMAND_COMPLETE)) ret = 1; - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); return ret; } @@ -749,6 +756,7 @@ static int ec_guard(struct acpi_ec *ec) static int ec_poll(struct acpi_ec *ec) { + unsigned long flags; int repeat = 5; /* number of command restarts */ while (repeat--) { @@ -757,14 +765,14 @@ static int ec_poll(struct acpi_ec *ec) do { if (!ec_guard(ec)) return 0; - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, flags); advance_transaction(ec, false); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); } while (time_before(jiffies, delay)); pr_debug("controller reset, restart transaction\n"); - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, flags); start_transaction(ec); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); } return -ETIME; } @@ -772,10 +780,11 @@ static int ec_poll(struct acpi_ec *ec) static int acpi_ec_transaction_unlocked(struct acpi_ec *ec, struct transaction *t) { + unsigned long tmp; int ret = 0; /* start transaction */ - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, tmp); /* Enable GPE for command processing (IBF=0/OBF=1) */ if (!acpi_ec_submit_flushable_request(ec)) { ret = -EINVAL; @@ -786,11 +795,11 @@ static int acpi_ec_transaction_unlocked(struct acpi_ec *ec, ec->curr = t; ec_dbg_req("Command(%s) started", acpi_ec_cmd_string(t->command)); start_transaction(ec); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, tmp); ret = ec_poll(ec); - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, tmp); if (t->irq_count == ec_storm_threshold) acpi_ec_unmask_events(ec); ec_dbg_req("Command(%s) stopped", acpi_ec_cmd_string(t->command)); @@ -799,7 +808,7 @@ static int acpi_ec_transaction_unlocked(struct acpi_ec *ec, acpi_ec_complete_request(ec); ec_dbg_ref(ec, "Decrease command"); unlock: - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, tmp); return ret; } @@ -927,7 +936,9 @@ EXPORT_SYMBOL(ec_get_handle); static void acpi_ec_start(struct acpi_ec *ec, bool resuming) { - spin_lock(&ec->lock); + unsigned long flags; + + spin_lock_irqsave(&ec->lock, flags); if (!test_and_set_bit(EC_FLAGS_STARTED, &ec->flags)) { ec_dbg_drv("Starting EC"); /* Enable GPE for event processing (SCI_EVT=1) */ @@ -937,28 +948,31 @@ static void acpi_ec_start(struct acpi_ec *ec, bool resuming) } ec_log_drv("EC started"); } - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); } static bool acpi_ec_stopped(struct acpi_ec *ec) { + unsigned long flags; bool flushed; - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, flags); flushed = acpi_ec_flushed(ec); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); return flushed; } static void acpi_ec_stop(struct acpi_ec *ec, bool suspending) { - spin_lock(&ec->lock); + unsigned long flags; + + spin_lock_irqsave(&ec->lock, flags); if (acpi_ec_started(ec)) { ec_dbg_drv("Stopping EC"); set_bit(EC_FLAGS_STOPPED, &ec->flags); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); wait_event(ec->wait, acpi_ec_stopped(ec)); - spin_lock(&ec->lock); + spin_lock_irqsave(&ec->lock, flags); /* Disable GPE for event processing (SCI_EVT=1) */ if (!suspending) { acpi_ec_complete_request(ec); @@ -969,25 +983,29 @@ static void acpi_ec_stop(struct acpi_ec *ec, bool suspending) clear_bit(EC_FLAGS_STOPPED, &ec->flags); ec_log_drv("EC stopped"); } - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); } static void acpi_ec_enter_noirq(struct acpi_ec *ec) { - spin_lock(&ec->lock); + unsigned long flags; + + spin_lock_irqsave(&ec->lock, flags); ec->busy_polling = true; ec->polling_guard = 0; ec_log_drv("interrupt blocked"); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); } static void acpi_ec_leave_noirq(struct acpi_ec *ec) { - spin_lock(&ec->lock); + unsigned long flags; + + spin_lock_irqsave(&ec->lock, flags); ec->busy_polling = ec_busy_polling; ec->polling_guard = ec_polling_guard; ec_log_drv("interrupt unblocked"); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); } void acpi_ec_block_transactions(void) @@ -1119,9 +1137,9 @@ static void acpi_ec_event_processor(struct work_struct *work) ec_dbg_evt("Query(0x%02x) stopped", handler->query_bit); - spin_lock(&ec->lock); + spin_lock_irq(&ec->lock); ec->queries_in_progress--; - spin_unlock(&ec->lock); + spin_unlock_irq(&ec->lock); acpi_ec_put_query_handler(handler); kfree(q); @@ -1184,12 +1202,12 @@ static int acpi_ec_submit_query(struct acpi_ec *ec) */ ec_dbg_evt("Query(0x%02x) scheduled", value); - spin_lock(&ec->lock); + spin_lock_irq(&ec->lock); ec->queries_in_progress++; queue_work(ec_query_wq, &q->work); - spin_unlock(&ec->lock); + spin_unlock_irq(&ec->lock); return 0; @@ -1205,14 +1223,14 @@ static void acpi_ec_event_handler(struct work_struct *work) ec_dbg_evt("Event started"); - spin_lock(&ec->lock); + spin_lock_irq(&ec->lock); while (ec->events_to_process) { - spin_unlock(&ec->lock); + spin_unlock_irq(&ec->lock); acpi_ec_submit_query(ec); - spin_lock(&ec->lock); + spin_lock_irq(&ec->lock); ec->events_to_process--; } @@ -1229,11 +1247,11 @@ static void acpi_ec_event_handler(struct work_struct *work) ec_dbg_evt("Event stopped"); - spin_unlock(&ec->lock); + spin_unlock_irq(&ec->lock); guard_timeout = !!ec_guard(ec); - spin_lock(&ec->lock); + spin_lock_irq(&ec->lock); /* Take care of SCI_EVT unless someone else is doing that. */ if (guard_timeout && !ec->curr) @@ -1246,7 +1264,7 @@ static void acpi_ec_event_handler(struct work_struct *work) ec->events_in_progress--; - spin_unlock(&ec->lock); + spin_unlock_irq(&ec->lock); } static void clear_gpe_and_advance_transaction(struct acpi_ec *ec, bool interrupt) @@ -1271,11 +1289,13 @@ static void clear_gpe_and_advance_transaction(struct acpi_ec *ec, bool interrupt static void acpi_ec_handle_interrupt(struct acpi_ec *ec) { - spin_lock(&ec->lock); + unsigned long flags; + + spin_lock_irqsave(&ec->lock, flags); clear_gpe_and_advance_transaction(ec, true); - spin_unlock(&ec->lock); + spin_unlock_irqrestore(&ec->lock, flags); } static u32 acpi_ec_gpe_handler(acpi_handle gpe_device, @@ -2085,7 +2105,7 @@ bool acpi_ec_dispatch_gpe(void) * Dispatch the EC GPE in-band, but do not report wakeup in any case * to allow the caller to process events properly after that. */ - spin_lock(&first_ec->lock); + spin_lock_irq(&first_ec->lock); if (acpi_ec_gpe_status_set(first_ec)) { pm_pr_dbg("ACPI EC GPE status set\n"); @@ -2094,7 +2114,7 @@ bool acpi_ec_dispatch_gpe(void) work_in_progress = acpi_ec_work_in_progress(first_ec); } - spin_unlock(&first_ec->lock); + spin_unlock_irq(&first_ec->lock); if (!work_in_progress) return false; @@ -2107,11 +2127,11 @@ bool acpi_ec_dispatch_gpe(void) pm_pr_dbg("ACPI EC work flushed\n"); - spin_lock(&first_ec->lock); + spin_lock_irq(&first_ec->lock); work_in_progress = acpi_ec_work_in_progress(first_ec); - spin_unlock(&first_ec->lock); + spin_unlock_irq(&first_ec->lock); } while (work_in_progress && !pm_wakeup_pending()); return false; -- cgit