[PATCH] staging: vchiq: rework remove_event handling

From: Arnd Bergmann <arnd@arndb.de>
To: Eric Anholt <eric@anholt.net>,
	Stefan Wahren <stefan.wahren@i2se.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Arnd Bergmann <arnd@arndb.de>,
	Nicolas Saenz Julienne <nsaenzjulienne@suse.de>,
	Tara Null <tn@fastmail.net>,
	linux-rpi-kernel@lists.infradead.org,
	linux-arm-kernel@lists.infradead.org, devel@driverdev.osuosl.org,
	linux-kernel@vger.kernel.org
Subject: [PATCH] staging: vchiq: rework remove_event handling
Date: Mon, 10 Dec 2018 22:11:58 +0100	[thread overview]
Message-ID: <20181210211231.4084251-1-arnd@arndb.de> (raw)

I had started the removal of semaphores in this driver without knowing
that Nicolas Saenz Julienne also worked on this. In case of the "remote
event" infrastructure, my solution seemed significantly better, so I'm
proposing this as a change on top.

The problem with using either semaphores or completions here is that
it's an overly complex way of waking up a thread, and it looks like the
'count' of the semaphore can easily get out of sync, even though I found
it hard to come up with a specific example.

Changing it to a 'wait_queue_head_t' instead of a completion simplifies
this by letting us wait directly on the 'event->fired' variable that is
set by the videocore.

Another simplification is passing the wait queue directly into the helper
functions instead of going through the fragile logic of recording the
offset inside of a structure as part of a shared memory variable. This
also avoids one uncached memory read and should be faster.

Note that I'm changing it back to 'killable' after the previous patch
changed 'killable' to 'interruptible', apparently based on a misunderstanding
of the subtle down_interruptible() macro override in vchiq_killable.h.

Fixes: f27e47bc6b8b ("staging: vchiq: use completions instead of semaphores")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 .../interface/vchiq_arm/vchiq_core.c          | 63 +++++++------------
 .../interface/vchiq_arm/vchiq_core.h          | 12 ++--
 2 files changed, 30 insertions(+), 45 deletions(-)

diff --git a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.c b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.c
index 482b5daf6c0c..eda3004a0c6a 100644
--- a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.c
+++ b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.c
@@ -417,26 +417,23 @@ vchiq_set_conn_state(VCHIQ_STATE_T *state, VCHIQ_CONNSTATE_T newstate)
 }
 
 static inline void
-remote_event_create(REMOTE_EVENT_T *event)
+remote_event_create(wait_queue_head_t *wq, REMOTE_EVENT_T *event)
 {
 	event->armed = 0;
 	/* Don't clear the 'fired' flag because it may already have been set
 	** by the other side. */
+	init_waitqueue_head(wq);
 }
 
 static inline int
-remote_event_wait(VCHIQ_STATE_T *state, REMOTE_EVENT_T *event)
+remote_event_wait(wait_queue_head_t *wq, REMOTE_EVENT_T *event)
 {
 	if (!event->fired) {
 		event->armed = 1;
 		dsb(sy);
-		if (!event->fired) {
-			if (wait_for_completion_interruptible(
-					(struct completion *)
-					((char *)state + event->event))) {
-				event->armed = 0;
-				return 0;
-			}
+		if (wait_event_killable(*wq, event->fired)) {
+			event->armed = 0;
+			return 0;
 		}
 		event->armed = 0;
 		wmb();
@@ -447,26 +444,26 @@ remote_event_wait(VCHIQ_STATE_T *state, REMOTE_EVENT_T *event)
 }
 
 static inline void
-remote_event_signal_local(VCHIQ_STATE_T *state, REMOTE_EVENT_T *event)
+remote_event_signal_local(wait_queue_head_t *wq, REMOTE_EVENT_T *event)
 {
 	event->armed = 0;
-	complete((struct completion *)((char *)state + event->event));
+	wake_up_all(wq);
 }
 
 static inline void
-remote_event_poll(VCHIQ_STATE_T *state, REMOTE_EVENT_T *event)
+remote_event_poll(wait_queue_head_t *wq, REMOTE_EVENT_T *event)
 {
 	if (event->fired && event->armed)
-		remote_event_signal_local(state, event);
+		remote_event_signal_local(wq, event);
 }
 
 void
 remote_event_pollall(VCHIQ_STATE_T *state)
 {
-	remote_event_poll(state, &state->local->sync_trigger);
-	remote_event_poll(state, &state->local->sync_release);
-	remote_event_poll(state, &state->local->trigger);
-	remote_event_poll(state, &state->local->recycle);
+	remote_event_poll(&state->sync_trigger_event, &state->local->sync_trigger);
+	remote_event_poll(&state->sync_release_event, &state->local->sync_release);
+	remote_event_poll(&state->trigger_event, &state->local->trigger);
+	remote_event_poll(&state->recycle_event, &state->local->recycle);
 }
 
 /* Round up message sizes so that any space at the end of a slot is always big
@@ -550,7 +547,7 @@ request_poll(VCHIQ_STATE_T *state, VCHIQ_SERVICE_T *service, int poll_type)
 	wmb();
 
 	/* ... and ensure the slot handler runs. */
-	remote_event_signal_local(state, &state->local->trigger);
+	remote_event_signal_local(&state->trigger_event, &state->local->trigger);
 }
 
 /* Called from queue_message, by the slot handler and application threads,
@@ -1069,7 +1066,7 @@ queue_message_sync(VCHIQ_STATE_T *state, VCHIQ_SERVICE_T *service,
 		(mutex_lock_killable(&state->sync_mutex) != 0))
 		return VCHIQ_RETRY;
 
-	remote_event_wait(state, &local->sync_release);
+	remote_event_wait(&state->sync_release_event, &local->sync_release);
 
 	rmb();
 
@@ -1887,7 +1884,7 @@ slot_handler_func(void *v)
 	while (1) {
 		DEBUG_COUNT(SLOT_HANDLER_COUNT);
 		DEBUG_TRACE(SLOT_HANDLER_LINE);
-		remote_event_wait(state, &local->trigger);
+		remote_event_wait(&state->trigger_event, &local->trigger);
 
 		rmb();
 
@@ -1976,7 +1973,7 @@ recycle_func(void *v)
 		return -ENOMEM;
 
 	while (1) {
-		remote_event_wait(state, &local->recycle);
+		remote_event_wait(&state->recycle_event, &local->recycle);
 
 		process_free_queue(state, found, length);
 	}
@@ -1998,7 +1995,7 @@ sync_func(void *v)
 		int type;
 		unsigned int localport, remoteport;
 
-		remote_event_wait(state, &local->sync_trigger);
+		remote_event_wait(&state->sync_trigger_event, &local->sync_trigger);
 
 		rmb();
 
@@ -2193,11 +2190,6 @@ vchiq_init_state(VCHIQ_STATE_T *state, VCHIQ_SLOT_ZERO_T *slot_zero)
 
 	init_completion(&state->connect);
 	mutex_init(&state->mutex);
-	init_completion(&state->trigger_event);
-	init_completion(&state->recycle_event);
-	init_completion(&state->sync_trigger_event);
-	init_completion(&state->sync_release_event);
-
 	mutex_init(&state->slot_mutex);
 	mutex_init(&state->recycle_mutex);
 	mutex_init(&state->sync_mutex);
@@ -2229,24 +2221,17 @@ vchiq_init_state(VCHIQ_STATE_T *state, VCHIQ_SLOT_ZERO_T *slot_zero)
 	state->data_use_count = 0;
 	state->data_quota = state->slot_queue_available - 1;
 
-	local->trigger.event = offsetof(VCHIQ_STATE_T, trigger_event);
-	remote_event_create(&local->trigger);
+	remote_event_create(&state->trigger_event, &local->trigger);
 	local->tx_pos = 0;
-
-	local->recycle.event = offsetof(VCHIQ_STATE_T, recycle_event);
-	remote_event_create(&local->recycle);
+	remote_event_create(&state->recycle_event, &local->recycle);
 	local->slot_queue_recycle = state->slot_queue_available;
-
-	local->sync_trigger.event = offsetof(VCHIQ_STATE_T, sync_trigger_event);
-	remote_event_create(&local->sync_trigger);
-
-	local->sync_release.event = offsetof(VCHIQ_STATE_T, sync_release_event);
-	remote_event_create(&local->sync_release);
+	remote_event_create(&state->sync_trigger_event, &local->sync_trigger);
+	remote_event_create(&state->sync_release_event, &local->sync_release);
 
 	/* At start-of-day, the slot is empty and available */
 	((VCHIQ_HEADER_T *)SLOT_DATA_FROM_INDEX(state, local->slot_sync))->msgid
 		= VCHIQ_MSGID_PADDING;
-	remote_event_signal_local(state, &local->sync_release);
+	remote_event_signal_local(&state->sync_release_event, &local->sync_release);
 
 	local->debug[DEBUG_ENTRIES] = DEBUG_MAX;
 
diff --git a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.h b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.h
index b76281f7510e..aae2c59700bd 100644
--- a/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.h
+++ b/drivers/staging/vc04_services/interface/vchiq_arm/vchiq_core.h
@@ -37,6 +37,7 @@
 #include <linux/mutex.h>
 #include <linux/completion.h>
 #include <linux/kthread.h>
+#include <linux/wait.h>
 
 #include "vchiq_cfg.h"
 
@@ -262,8 +263,7 @@ typedef struct vchiq_bulk_queue_struct {
 typedef struct remote_event_struct {
 	int armed;
 	int fired;
-	/* Contains offset from the beginning of the VCHIQ_STATE_T structure */
-	u32 event;
+	u32 __unused;
 } REMOTE_EVENT_T;
 
 typedef struct opaque_platform_state_t *VCHIQ_PLATFORM_STATE_T;
@@ -426,16 +426,16 @@ struct vchiq_state_struct {
 	struct task_struct *sync_thread;
 
 	/* Local implementation of the trigger remote event */
-	struct completion trigger_event;
+	wait_queue_head_t trigger_event;
 
 	/* Local implementation of the recycle remote event */
-	struct completion recycle_event;
+	wait_queue_head_t recycle_event;
 
 	/* Local implementation of the sync trigger remote event */
-	struct completion sync_trigger_event;
+	wait_queue_head_t sync_trigger_event;
 
 	/* Local implementation of the sync release remote event */
-	struct completion sync_release_event;
+	wait_queue_head_t sync_release_event;
 
 	char *tx_data;
 	char *rx_data;
-- 
2.20.0

