Received-SPF: pass (google.com: domain of linux-kernel+bounces-123904-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
From: Elizabeth Figura <zfigura@codeweavers.com>
To: Arnd Bergmann <arnd@arndb.de>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Jonathan Corbet <corbet@lwn.net>,
	Shuah Khan <shuah@kernel.org>
Cc: linux-kernel@vger.kernel.org,
	linux-api@vger.kernel.org,
	wine-devel@winehq.org,
	=?UTF-8?q?Andr=C3=A9=20Almeida?= <andrealmeid@igalia.com>,
	Wolfram Sang <wsa@kernel.org>,
	Arkadiusz Hiler <ahiler@codeweavers.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Andy Lutomirski <luto@kernel.org>,
	linux-doc@vger.kernel.org,
	linux-kselftest@vger.kernel.org,
	Randy Dunlap <rdunlap@infradead.org>,
	Elizabeth Figura <zfigura@codeweavers.com>
Subject: [PATCH v3 16/30] ntsync: Introduce alertable waits.
Date: Thu, 28 Mar 2024 19:06:07 -0500
Message-ID: <20240329000621.148791-17-zfigura@codeweavers.com>
In-Reply-To: <20240329000621.148791-1-zfigura@codeweavers.com>
References: <20240329000621.148791-1-zfigura@codeweavers.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

NT waits can optionally be made "alertable". This is a special channel for
thread wakeup that is mildly similar to SIGIO. A thread has an internal single
bit of "alerted" state, and if a thread is alerted while an alertable wait, the
wait will return a special value, consume the "alerted" state, and will not
consume any of its objects.

Alerts are implemented using events; the user-space NT emulator is expected to
create an internal ntsync event for each thread and pass that event to wait
functions.

Signed-off-by: Elizabeth Figura <zfigura@codeweavers.com>
---
 drivers/misc/ntsync.c       | 68 ++++++++++++++++++++++++++++++++-----
 include/uapi/linux/ntsync.h |  2 +-
 2 files changed, 60 insertions(+), 10 deletions(-)

diff --git a/drivers/misc/ntsync.c b/drivers/misc/ntsync.c
index a03c6fceb518..19fd70ac3f50 100644
--- a/drivers/misc/ntsync.c
+++ b/drivers/misc/ntsync.c
@@ -819,25 +819,32 @@ static int setup_wait(struct ntsync_device *dev,
 		      const struct ntsync_wait_args *args, bool all,
 		      struct ntsync_q **ret_q)
 {
+	int fds[NTSYNC_MAX_WAIT_COUNT + 1];
 	const __u32 count = args->count;
-	int fds[NTSYNC_MAX_WAIT_COUNT];
 	struct ntsync_q *q;
+	__u32 total_count;
 	__u32 i, j;
 
 	if (!args->owner)
 		return -EINVAL;
 
-	if (args->pad || args->pad2 || (args->flags & ~NTSYNC_WAIT_REALTIME))
+	if (args->pad || (args->flags & ~NTSYNC_WAIT_REALTIME))
 		return -EINVAL;
 
 	if (args->count > NTSYNC_MAX_WAIT_COUNT)
 		return -EINVAL;
 
+	total_count = count;
+	if (args->alert)
+		total_count++;
+
 	if (copy_from_user(fds, u64_to_user_ptr(args->objs),
 			   array_size(count, sizeof(*fds))))
 		return -EFAULT;
+	if (args->alert)
+		fds[count] = args->alert;
 
-	q = kmalloc(struct_size(q, entries, count), GFP_KERNEL);
+	q = kmalloc(struct_size(q, entries, total_count), GFP_KERNEL);
 	if (!q)
 		return -ENOMEM;
 	q->task = current;
@@ -847,7 +854,7 @@ static int setup_wait(struct ntsync_device *dev,
 	q->ownerdead = false;
 	q->count = count;
 
-	for (i = 0; i < count; i++) {
+	for (i = 0; i < total_count; i++) {
 		struct ntsync_q_entry *entry = &q->entries[i];
 		struct ntsync_obj *obj = get_obj(dev, fds[i]);
 
@@ -897,9 +904,9 @@ static void try_wake_any_obj(struct ntsync_obj *obj)
 static int ntsync_wait_any(struct ntsync_device *dev, void __user *argp)
 {
 	struct ntsync_wait_args args;
+	__u32 i, total_count;
 	struct ntsync_q *q;
 	int signaled;
-	__u32 i;
 	int ret;
 
 	if (copy_from_user(&args, argp, sizeof(args)))
@@ -909,9 +916,13 @@ static int ntsync_wait_any(struct ntsync_device *dev, void __user *argp)
 	if (ret < 0)
 		return ret;
 
+	total_count = args.count;
+	if (args.alert)
+		total_count++;
+
 	/* queue ourselves */
 
-	for (i = 0; i < args.count; i++) {
+	for (i = 0; i < total_count; i++) {
 		struct ntsync_q_entry *entry = &q->entries[i];
 		struct ntsync_obj *obj = entry->obj;
 
@@ -920,9 +931,15 @@ static int ntsync_wait_any(struct ntsync_device *dev, void __user *argp)
 		spin_unlock(&obj->lock);
 	}
 
-	/* check if we are already signaled */
+	/*
+	 * Check if we are already signaled.
+	 *
+	 * Note that the API requires that normal objects are checked before
+	 * the alert event. Hence we queue the alert event last, and check
+	 * objects in order.
+	 */
 
-	for (i = 0; i < args.count; i++) {
+	for (i = 0; i < total_count; i++) {
 		struct ntsync_obj *obj = q->entries[i].obj;
 
 		if (atomic_read(&q->signaled) != -1)
@@ -939,7 +956,7 @@ static int ntsync_wait_any(struct ntsync_device *dev, void __user *argp)
 
 	/* and finally, unqueue */
 
-	for (i = 0; i < args.count; i++) {
+	for (i = 0; i < total_count; i++) {
 		struct ntsync_q_entry *entry = &q->entries[i];
 		struct ntsync_obj *obj = entry->obj;
 
@@ -999,6 +1016,14 @@ static int ntsync_wait_all(struct ntsync_device *dev, void __user *argp)
 		 */
 		list_add_tail(&entry->node, &obj->all_waiters);
 	}
+	if (args.alert) {
+		struct ntsync_q_entry *entry = &q->entries[args.count];
+		struct ntsync_obj *obj = entry->obj;
+
+		spin_lock_nest_lock(&obj->lock, &dev->wait_all_lock);
+		list_add_tail(&entry->node, &obj->any_waiters);
+		spin_unlock(&obj->lock);
+	}
 
 	/* check if we are already signaled */
 
@@ -1006,6 +1031,21 @@ static int ntsync_wait_all(struct ntsync_device *dev, void __user *argp)
 
 	spin_unlock(&dev->wait_all_lock);
 
+	/*
+	 * Check if the alert event is signaled, making sure to do so only
+	 * after checking if the other objects are signaled.
+	 */
+
+	if (args.alert) {
+		struct ntsync_obj *obj = q->entries[args.count].obj;
+
+		if (atomic_read(&q->signaled) == -1) {
+			spin_lock(&obj->lock);
+			try_wake_any_obj(obj);
+			spin_unlock(&obj->lock);
+		}
+	}
+
 	/* sleep */
 
 	ret = ntsync_schedule(q, &args);
@@ -1028,6 +1068,16 @@ static int ntsync_wait_all(struct ntsync_device *dev, void __user *argp)
 
 		put_obj(obj);
 	}
+	if (args.alert) {
+		struct ntsync_q_entry *entry = &q->entries[args.count];
+		struct ntsync_obj *obj = entry->obj;
+
+		spin_lock_nest_lock(&obj->lock, &dev->wait_all_lock);
+		list_del(&entry->node);
+		spin_unlock(&obj->lock);
+
+		put_obj(obj);
+	}
 
 	spin_unlock(&dev->wait_all_lock);
 
diff --git a/include/uapi/linux/ntsync.h b/include/uapi/linux/ntsync.h
index 80f36de46a75..f21dbac42164 100644
--- a/include/uapi/linux/ntsync.h
+++ b/include/uapi/linux/ntsync.h
@@ -37,8 +37,8 @@ struct ntsync_wait_args {
 	__u32 owner;
 	__u32 index;
 	__u32 flags;
+	__u32 alert;
 	__u32 pad;
-	__u32 pad2;
 };
 
 #define NTSYNC_MAX_WAIT_COUNT 64
-- 
2.43.0