Skip to contentSkip to conversation

FAILED: patch "[PATCH] n_tty: Access echo_* variables carefully." failed to apply to 4.4-stable tree

PATCH [email protected] 2 participants 3 messages Jul 4, 2018 – Jul 5, 2018

Subsystems (2)
lore
Share & tools
lore

[email protected]

author maintainer
#0 lore raw

The patch below does not apply to the 4.4-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

thanks,

greg k-h

------------------ original commit in Linus's tree ------------------

Quoted 1 line

From ebec3f8f5271139df618ebdf8427e24ba102ba94 Mon Sep 17 00:00:00 2001

From: Tetsuo Handa <[email protected]>
Date: Sat, 26 May 2018 09:53:14 +0900
Subject: [PATCH] n_tty: Access echo_* variables carefully.

syzbot is reporting stalls at __process_echoes() [1]. This is because since ldata->echo_commit < ldata->echo_tail becomes true for some reason, the discard loop is serving as almost infinite loop. This patch tries to avoid falling into ldata->echo_commit < ldata->echo_tail situation by making access to echo_* variables more carefully.

Since reset_buffer_flags() is called without output_lock held, it should not touch echo_* variables. And omit a call to reset_buffer_flags() from n_tty_open() by using vzalloc().

Since add_echo_byte() is called without output_lock held, it needs memory barrier between storing into echo_buf[] and incrementing echo_head counter. echo_buf() needs corresponding memory barrier before reading echo_buf[]. Lack of handling the possibility of not-yet-stored multi-byte operation might be the reason of falling into ldata->echo_commit < ldata->echo_tail situation, for if I do WARN_ON(ldata->echo_commit == tail + 1) prior to echo_buf(ldata, tail + 1), the WARN_ON() fires.

Also, explicitly masking with buffer for the former "while" loop, and use ldata->echo_commit > tail for the latter "while" loop.

[1] https://syzkaller.appspot.com/bug?id=17f23b094cd80df750e5b0f8982c521ee6bcbf40

drivers/tty/n_tty.c +24 -18
index b279f8730e04..431742201709 100644
@@ -143,6 +143,7 @@ static inline unsigned char *read_buf_addr(struct n_tty_data *ldata, size_t i)
 
 static inline unsigned char echo_buf(struct n_tty_data *ldata, size_t i)
 {
+	smp_rmb(); /* Matches smp_wmb() in add_echo_byte(). */
 	return ldata->echo_buf[i & (N_TTY_BUF_SIZE - 1)];
 }
 
@@ -318,9 +319,7 @@ static inline void put_tty_queue(unsigned char c, struct n_tty_data *ldata)
 static void reset_buffer_flags(struct n_tty_data *ldata)
 {
 	ldata->read_head = ldata->canon_head = ldata->read_tail = 0;
-	ldata->echo_head = ldata->echo_tail = ldata->echo_commit = 0;
 	ldata->commit_head = 0;
-	ldata->echo_mark = 0;
 	ldata->line_start = 0;
 
 	ldata->erasing = 0;
@@ -619,12 +618,19 @@ static size_t __process_echoes(struct tty_struct *tty)
 	old_space = space = tty_write_room(tty);
 
 	tail = ldata->echo_tail;
-	while (ldata->echo_commit != tail) {
+	while (MASK(ldata->echo_commit) != MASK(tail)) {
 		c = echo_buf(ldata, tail);
 		if (c == ECHO_OP_START) {
 			unsigned char op;
 			int no_space_left = 0;
 
+			/*
+			 * Since add_echo_byte() is called without holding
+			 * output_lock, we might see only portion of multi-byte
+			 * operation.
+			 */
+			if (MASK(ldata->echo_commit) == MASK(tail + 1))
+				goto not_yet_stored;
 			/*
 			 * If the buffer byte is the start of a multi-byte
 			 * operation, get the next byte, which is either the
@@ -636,6 +642,8 @@ static size_t __process_echoes(struct tty_struct *tty)
 				unsigned int num_chars, num_bs;
 
 			case ECHO_OP_ERASE_TAB:
+				if (MASK(ldata->echo_commit) == MASK(tail + 2))
+					goto not_yet_stored;
 				num_chars = echo_buf(ldata, tail + 2);
 
 				/*
@@ -730,7 +738,8 @@ static size_t __process_echoes(struct tty_struct *tty)
 	/* If the echo buffer is nearly full (so that the possibility exists
 	 * of echo overrun before the next commit), then discard enough
 	 * data at the tail to prevent a subsequent overrun */
-	while (ldata->echo_commit - tail >= ECHO_DISCARD_WATERMARK) {
+	while (ldata->echo_commit > tail &&
+	       ldata->echo_commit - tail >= ECHO_DISCARD_WATERMARK) {
 		if (echo_buf(ldata, tail) == ECHO_OP_START) {
 			if (echo_buf(ldata, tail + 1) == ECHO_OP_ERASE_TAB)
 				tail += 3;
@@ -740,6 +749,7 @@ static size_t __process_echoes(struct tty_struct *tty)
 			tail++;
 	}
 
+ not_yet_stored:
 	ldata->echo_tail = tail;
 	return old_space - space;
 }
@@ -750,6 +760,7 @@ static void commit_echoes(struct tty_struct *tty)
 	size_t nr, old, echoed;
 	size_t head;
 
+	mutex_lock(&ldata->output_lock);
 	head = ldata->echo_head;
 	ldata->echo_mark = head;
 	old = ldata->echo_commit - ldata->echo_tail;
@@ -758,10 +769,12 @@ static void commit_echoes(struct tty_struct *tty)
 	 * is over the threshold (and try again each time another
 	 * block is accumulated) */
 	nr = head - ldata->echo_tail;
-	if (nr < ECHO_COMMIT_WATERMARK || (nr % ECHO_BLOCK > old % ECHO_BLOCK))
+	if (nr < ECHO_COMMIT_WATERMARK ||
+	    (nr % ECHO_BLOCK > old % ECHO_BLOCK)) {
+		mutex_unlock(&ldata->output_lock);
 		return;
+	}
 
-	mutex_lock(&ldata->output_lock);
 	ldata->echo_commit = head;
 	echoed = __process_echoes(tty);
 	mutex_unlock(&ldata->output_lock);
@@ -812,7 +825,9 @@ static void flush_echoes(struct tty_struct *tty)
 
 static inline void add_echo_byte(unsigned char c, struct n_tty_data *ldata)
 {
-	*echo_buf_addr(ldata, ldata->echo_head++) = c;
+	*echo_buf_addr(ldata, ldata->echo_head) = c;
+	smp_wmb(); /* Matches smp_rmb() in echo_buf(). */
+	ldata->echo_head++;
 }
 
 /**
@@ -1881,30 +1896,21 @@ static int n_tty_open(struct tty_struct *tty)
 	struct n_tty_data *ldata;
 
 	/* Currently a malloc failure here can panic */
-	ldata = vmalloc(sizeof(*ldata));
+	ldata = vzalloc(sizeof(*ldata));
 	if (!ldata)
-		goto err;
+		return -ENOMEM;
 
 	ldata->overrun_time = jiffies;
 	mutex_init(&ldata->atomic_read_lock);
 	mutex_init(&ldata->output_lock);
 
 	tty->disc_data = ldata;
-	reset_buffer_flags(tty->disc_data);
-	ldata->column = 0;
-	ldata->canon_column = 0;
-	ldata->num_overrun = 0;
-	ldata->no_room = 0;
-	ldata->lnext = 0;
 	tty->closing = 0;
 	/* indicate buffer work may resume */
 	clear_bit(TTY_LDISC_HALTED, &tty->flags);
 	n_tty_set_termios(tty, NULL);
 	tty_unthrottle(tty);
-
 	return 0;
-err:
-	return -ENOMEM;
 }
 
 static inline int input_available_p(struct tty_struct *tty, int poll)
#1 lore raw
Quoted 6 lines · Greg Kroah-Hartman

The patch below does not apply to the 4.4-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

OK. "ldata->minimum_to_wake = 1;" line was there (which will be removed in 4.6-rc1). Updated patch for 4.4-stable is attached below.

Quoted 1 line

From 24d0c82c23c67b8985de01b1337937299846ac1d Mon Sep 17 00:00:00 2001

From: Tetsuo Handa <[email protected]>
Date: Thu, 5 Jul 2018 13:28:22 +0900
Subject: [PATCH] n_tty: Access echo_* variables carefully.

commit ebec3f8f5271139df618ebdf8427e24ba102ba94 upstream.

syzbot is reporting stalls at __process_echoes() [1]. This is because since ldata->echo_commit < ldata->echo_tail becomes true for some reason, the discard loop is serving as almost infinite loop. This patch tries to avoid falling into ldata->echo_commit < ldata->echo_tail situation by making access to echo_* variables more carefully.

Since reset_buffer_flags() is called without output_lock held, it should not touch echo_* variables. And omit a call to reset_buffer_flags() from n_tty_open() by using vzalloc().

Since add_echo_byte() is called without output_lock held, it needs memory barrier between storing into echo_buf[] and incrementing echo_head counter. echo_buf() needs corresponding memory barrier before reading echo_buf[]. Lack of handling the possibility of not-yet-stored multi-byte operation might be the reason of falling into ldata->echo_commit < ldata->echo_tail situation, for if I do WARN_ON(ldata->echo_commit == tail + 1) prior to echo_buf(ldata, tail + 1), the WARN_ON() fires.

Also, explicitly masking with buffer for the former "while" loop, and use ldata->echo_commit > tail for the latter "while" loop.

[1] https://syzkaller.appspot.com/bug?id=17f23b094cd80df750e5b0f8982c521ee6bcbf40

---
 drivers/tty/n_tty.c | 42 ++++++++++++++++++++++++------------------
 1 file changed, 24 insertions(+), 18 deletions(-)
drivers/tty/n_tty.c +24 -18
index 1d4d20e..b1ec202 100644
@@ -147,6 +147,7 @@ static inline unsigned char *read_buf_addr(struct n_tty_data *ldata, size_t i)
 
 static inline unsigned char echo_buf(struct n_tty_data *ldata, size_t i)
 {
+	smp_rmb(); /* Matches smp_wmb() in add_echo_byte(). */
 	return ldata->echo_buf[i & (N_TTY_BUF_SIZE - 1)];
 }
 
@@ -324,9 +325,7 @@ static inline void put_tty_queue(unsigned char c, struct n_tty_data *ldata)
 static void reset_buffer_flags(struct n_tty_data *ldata)
 {
 	ldata->read_head = ldata->canon_head = ldata->read_tail = 0;
-	ldata->echo_head = ldata->echo_tail = ldata->echo_commit = 0;
 	ldata->commit_head = 0;
-	ldata->echo_mark = 0;
 	ldata->line_start = 0;
 
 	ldata->erasing = 0;
@@ -647,13 +646,20 @@ static size_t __process_echoes(struct tty_struct *tty)
 	old_space = space = tty_write_room(tty);
 
 	tail = ldata->echo_tail;
-	while (ldata->echo_commit != tail) {
+	while (MASK(ldata->echo_commit) != MASK(tail)) {
 		c = echo_buf(ldata, tail);
 		if (c == ECHO_OP_START) {
 			unsigned char op;
 			int no_space_left = 0;
 
 			/*
+			 * Since add_echo_byte() is called without holding
+			 * output_lock, we might see only portion of multi-byte
+			 * operation.
+			 */
+			if (MASK(ldata->echo_commit) == MASK(tail + 1))
+				goto not_yet_stored;
+			/*
 			 * If the buffer byte is the start of a multi-byte
 			 * operation, get the next byte, which is either the
 			 * op code or a control character value.
@@ -664,6 +670,8 @@ static size_t __process_echoes(struct tty_struct *tty)
 				unsigned int num_chars, num_bs;
 
 			case ECHO_OP_ERASE_TAB:
+				if (MASK(ldata->echo_commit) == MASK(tail + 2))
+					goto not_yet_stored;
 				num_chars = echo_buf(ldata, tail + 2);
 
 				/*
@@ -758,7 +766,8 @@ static size_t __process_echoes(struct tty_struct *tty)
 	/* If the echo buffer is nearly full (so that the possibility exists
 	 * of echo overrun before the next commit), then discard enough
 	 * data at the tail to prevent a subsequent overrun */
-	while (ldata->echo_commit - tail >= ECHO_DISCARD_WATERMARK) {
+	while (ldata->echo_commit > tail &&
+	       ldata->echo_commit - tail >= ECHO_DISCARD_WATERMARK) {
 		if (echo_buf(ldata, tail) == ECHO_OP_START) {
 			if (echo_buf(ldata, tail + 1) == ECHO_OP_ERASE_TAB)
 				tail += 3;
@@ -768,6 +777,7 @@ static size_t __process_echoes(struct tty_struct *tty)
 			tail++;
 	}
 
+ not_yet_stored:
 	ldata->echo_tail = tail;
 	return old_space - space;
 }
@@ -778,6 +788,7 @@ static void commit_echoes(struct tty_struct *tty)
 	size_t nr, old, echoed;
 	size_t head;
 
+	mutex_lock(&ldata->output_lock);
 	head = ldata->echo_head;
 	ldata->echo_mark = head;
 	old = ldata->echo_commit - ldata->echo_tail;
@@ -786,10 +797,12 @@ static void commit_echoes(struct tty_struct *tty)
 	 * is over the threshold (and try again each time another
 	 * block is accumulated) */
 	nr = head - ldata->echo_tail;
-	if (nr < ECHO_COMMIT_WATERMARK || (nr % ECHO_BLOCK > old % ECHO_BLOCK))
+	if (nr < ECHO_COMMIT_WATERMARK ||
+	    (nr % ECHO_BLOCK > old % ECHO_BLOCK)) {
+		mutex_unlock(&ldata->output_lock);
 		return;
+	}
 
-	mutex_lock(&ldata->output_lock);
 	ldata->echo_commit = head;
 	echoed = __process_echoes(tty);
 	mutex_unlock(&ldata->output_lock);
@@ -840,7 +853,9 @@ static void flush_echoes(struct tty_struct *tty)
 
 static inline void add_echo_byte(unsigned char c, struct n_tty_data *ldata)
 {
-	*echo_buf_addr(ldata, ldata->echo_head++) = c;
+	*echo_buf_addr(ldata, ldata->echo_head) = c;
+	smp_wmb(); /* Matches smp_rmb() in echo_buf(). */
+	ldata->echo_head++;
 }
 
 /**
@@ -1920,31 +1935,22 @@ static int n_tty_open(struct tty_struct *tty)
 	struct n_tty_data *ldata;
 
 	/* Currently a malloc failure here can panic */
-	ldata = vmalloc(sizeof(*ldata));
+	ldata = vzalloc(sizeof(*ldata));
 	if (!ldata)
-		goto err;
+		return -ENOMEM;
 
 	ldata->overrun_time = jiffies;
 	mutex_init(&ldata->atomic_read_lock);
 	mutex_init(&ldata->output_lock);
 
 	tty->disc_data = ldata;
-	reset_buffer_flags(tty->disc_data);
-	ldata->column = 0;
-	ldata->canon_column = 0;
 	ldata->minimum_to_wake = 1;
-	ldata->num_overrun = 0;
-	ldata->no_room = 0;
-	ldata->lnext = 0;
 	tty->closing = 0;
 	/* indicate buffer work may resume */
 	clear_bit(TTY_LDISC_HALTED, &tty->flags);
 	n_tty_set_termios(tty, NULL);
 	tty_unthrottle(tty);
-
 	return 0;
-err:
-	return -ENOMEM;
 }
 
 static inline int input_available_p(struct tty_struct *tty, int poll)
Signature · 1 line
1.8.3.1
author maintainer
#2 lore raw
Quoted 10 lines · Tetsuo Handa
Quoted 6 lines · Greg Kroah-Hartman

The patch below does not apply to the 4.4-stable tree.
If someone wants it applied there, or to any other stable or longterm tree, then please email the backport, including the original git commit id to <[email protected]>.

OK. "ldata->minimum_to_wake = 1;" line was there (which will be removed in 4.6-rc1). Updated patch for 4.4-stable is attached below.

Thanks, now applied.

greg k-h