tty: Add driver unthrottle in ioctl(...,TCFLSH,..).

PATCH Superseded Ilya Zykov 2 participants 3 messages Nov 22, 2012 – Nov 23, 2012 TTY layer and serial drivers

lore

Share & tools

lore

Ilya Zykov

author

2012-11-22 23:03 UTC #0 lore raw

PATCH v2 1/1 tty: Add driver unthrottle in ioctl(...,TCFLSH,..).

Regression 'tty: fix "IRQ45: nobody cared"' Regression commit 7b292b4bf9a9d6098440d85616d6ca4c608b8304

Function reset_buffer_flags() also invoked during the ioctl(...,TCFLSH,..).
At the time of request we can have full buffers and throttled driver too. If we don't unthrottle driver, we can get forever throttled driver, because, after request, we will have empty buffers and throttled driver and there is no place to unthrottle driver.
It simple reproduce with "pty" pair then one side sleep on tty->write_wait, and other side do ioctl(...,TCFLSH,..). Then there is no place to do writers wake up.

About 'tty: fix "IRQ45: nobody cared"':
We don't call tty_unthrottle() if release last filp - ('tty->count == 0')
In other case it must be safely.

Unfortunately, many drivers indirectly invoke tty_unthrottle() before TTY LAYER decremented (tty->count).
This Patch help us catch bugs in tty's drivers and invoke tty_unthrottle()
in right moment only.

Signed-off-by Ilya Zykov <[email protected]>

---

drivers/tty/n_tty.c +8 -1

index 26f0d0e..f20b44a 100644
@@ -173,6 +173,13 @@ static void reset_buffer_flags(struct tty_struct *tty)
 {
 	unsigned long flags;
 
+	WARN_ON(tty->drvbug);
+	/*
+	 * We should not call this method from driver's close function.
+	 * It will be called by TTY layer later.
+	 */
+	if (tty->drvbug)
+		return;
 	spin_lock_irqsave(&tty->read_lock, flags);
 	tty->read_head = tty->read_tail = tty->read_cnt = 0;
 	spin_unlock_irqrestore(&tty->read_lock, flags);

@@ -184,6 +191,7 @@ static void reset_buffer_flags(struct tty_struct *tty)
 	tty->canon_head = tty->canon_data = tty->erasing = 0;
 	memset(&tty->read_flags, 0, sizeof tty->read_flags);
 	n_tty_set_room(tty);
+	check_unthrottle(tty);
 }
 
 /**

@@ -1585,7 +1593,6 @@ static int n_tty_open(struct tty_struct *tty)
 			return -ENOMEM;
 	}
 	reset_buffer_flags(tty);
-	tty_unthrottle(tty);
 	tty->column = 0;
 	n_tty_set_termios(tty, NULL);
 	tty->minimum_to_wake = 1;

drivers/tty/tty_io.c +2 -0

index b425c79..0cd2370 100644
@@ -1649,8 +1649,10 @@ int tty_release(struct inode *inode, struct file *filp)
 			tty_name(tty, buf), tty->count);
 #endif
 
+	tty->drvbug = 1;
 	if (tty->ops->close)
 		tty->ops->close(tty, filp);
+	tty->drvbug = 0;
 
 	tty_unlock();
 	/*

include/linux/tty.h +1 -1

index ed1e82e..01455b6 100644
@@ -281,7 +281,7 @@ struct tty_struct {
 	int count;
 	struct winsize winsize;		/* termios mutex */
 	unsigned char stopped:1, hw_stopped:1, flow_stopped:1, packet:1;
-	unsigned char low_latency:1, warned:1;
+	unsigned char low_latency:1, warned:1, drvbug:1;
 	unsigned char ctrl_status;	/* ctrl_lock */
 	unsigned int receive_room;	/* Bytes free for queue */

Ilya Zykov

author

2012-11-22 23:11 UTC #1 lore raw

PATCH v2 1/1 tty: Add driver unthrottle in ioctl(...,TCFLSH,..).

Regression 'tty: fix "IRQ45: nobody cared"' Regression commit 7b292b4bf9a9d6098440d85616d6ca4c608b8304

About 'tty: fix "IRQ45: nobody cared"':
We don't call tty_unthrottle() if release last filp - ('tty->count == 0')
In other case it must be safely.

Signed-off-by Ilya Zykov <[email protected]>

---

drivers/tty/n_tty.c +8 -1

index 26f0d0e..f20b44a 100644
@@ -173,6 +173,13 @@ static void reset_buffer_flags(struct tty_struct *tty)
 {
 	unsigned long flags;
 
+	WARN_ON(tty->drvbug);
+	/*
+	 * We should not call this method from driver's close function.
+	 * It will be called by TTY layer later.
+	 */
+	if (tty->drvbug)
+		return;
 	spin_lock_irqsave(&tty->read_lock, flags);
 	tty->read_head = tty->read_tail = tty->read_cnt = 0;
 	spin_unlock_irqrestore(&tty->read_lock, flags);

@@ -184,6 +191,7 @@ static void reset_buffer_flags(struct tty_struct *tty)
 	tty->canon_head = tty->canon_data = tty->erasing = 0;
 	memset(&tty->read_flags, 0, sizeof tty->read_flags);
 	n_tty_set_room(tty);
+	check_unthrottle(tty);
 }
 
 /**

@@ -1585,7 +1593,6 @@ static int n_tty_open(struct tty_struct *tty)
 			return -ENOMEM;
 	}
 	reset_buffer_flags(tty);
-	tty_unthrottle(tty);
 	tty->column = 0;
 	n_tty_set_termios(tty, NULL);
 	tty->minimum_to_wake = 1;

drivers/tty/tty_io.c +2 -0

index b425c79..0cd2370 100644
@@ -1649,8 +1649,10 @@ int tty_release(struct inode *inode, struct file *filp)
 			tty_name(tty, buf), tty->count);
 #endif
 
+	tty->drvbug = 1;
 	if (tty->ops->close)
 		tty->ops->close(tty, filp);
+	tty->drvbug = 0;
 
 	tty_unlock();
 	/*

include/linux/tty.h +1 -1

index ed1e82e..01455b6 100644
@@ -281,7 +281,7 @@ struct tty_struct {
 	int count;
 	struct winsize winsize;		/* termios mutex */
 	unsigned char stopped:1, hw_stopped:1, flow_stopped:1, packet:1;
-	unsigned char low_latency:1, warned:1;
+	unsigned char low_latency:1, warned:1, drvbug:1;
 	unsigned char ctrl_status;	/* ctrl_lock */
 	unsigned int receive_room;	/* Bytes free for queue */

Alan Cox

2012-11-23 13:48 UTC #2 lore raw

In reply to Ilya Zykov

Quoted 11 lines · On Fri, 23 Nov 2012 03:02:56 +0400 Ilya Zykov <[email protected]> wrote:

Regression 'tty: fix "IRQ45: nobody cared"' Regression commit 7b292b4bf9a9d6098440d85616d6ca4c608b8304
Function reset_buffer_flags() also invoked during the ioctl(...,TCFLSH,..).
At the time of request we can have full buffers and throttled driver too. If we don't unthrottle driver, we can get forever throttled driver, because, after request, we will have empty buffers and throttled driver and there is no place to unthrottle driver.
It simple reproduce with "pty" pair then one side sleep on tty->write_wait, and other side do ioctl(...,TCFLSH,..). Then there is no place to do writers wake up.

Quoted 8 lines

About 'tty: fix "IRQ45: nobody cared"':
We don't call tty_unthrottle() if release last filp - ('tty->count == 0')
In other case it must be safely.
Unfortunately, many drivers indirectly invoke tty_unthrottle() before TTY LAYER decremented (tty->count).
This Patch help us catch bugs in tty's drivers and invoke tty_unthrottle()
in right moment only.

This isn't really what I meant by fixing the problem. Simply declaring the known universe to disagree with your view and adding a WARN isn't a fix.

Alan