Skip to contentSkip to conversation

serial: off by one

PATCH In review Roel Kluin 3 participants 3 messages Jun 3, 2009 – Jun 4, 2009

lore
Share & tools
lore

Roel Kluin

author
#0 lore raw

PATCH 1/1 serial: off by one

In zs_console_putchar() occurs:

	if (zs_transmit_drain(zport, irq))
		write_zsdata(zport, ch);

However if in zs_transmit_drain() no empty Tx Buffer occurs, limit reaches -1 => true, and the write still occurs.

This patch changes postfix to prefix decrements in this and similar functions to prevent similar mistakes in the future. This decreases the iterations with one but the chosen loop count was arbitrary anyway.

In sunhv limit reaches -1, not 0, so the test is off by one.

---
 drivers/serial/sb1250-duart.c |    6 +++---
 drivers/serial/sunhv.c        |    2 +-
 drivers/serial/zs.c           |    6 +++---
 3 files changed, 7 insertions(+), 7 deletions(-)
drivers/serial/sb1250-duart.c +3 -3
index a4fb343..319e8b8 100644
@@ -204,7 +204,7 @@ static int sbd_receive_drain(struct sbd_port *sport)
 {
 	int loops = 10000;
 
-	while (sbd_receive_ready(sport) && loops--)
+	while (sbd_receive_ready(sport) && --loops)
 		read_sbdchn(sport, R_DUART_RX_HOLD);
 	return loops;
 }
@@ -218,7 +218,7 @@ static int __maybe_unused sbd_transmit_drain(struct sbd_port *sport)
 {
 	int loops = 10000;
 
-	while (!sbd_transmit_ready(sport) && loops--)
+	while (!sbd_transmit_ready(sport) && --loops)
 		udelay(2);
 	return loops;
 }
@@ -232,7 +232,7 @@ static int sbd_line_drain(struct sbd_port *sport)
 {
 	int loops = 10000;
 
-	while (!sbd_transmit_empty(sport) && loops--)
+	while (!sbd_transmit_empty(sport) && --loops)
 		udelay(2);
 	return loops;
 }
drivers/serial/sunhv.c +1 -1
index a94a2ab..1df5325 100644
@@ -461,7 +461,7 @@ static void sunhv_console_write_paged(struct console *con, const char *s, unsign
 					break;
 				udelay(1);
 			}
-			if (limit <= 0)
+			if (limit < 0)
 				break;
 			page_bytes -= written;
 			ra += written;
drivers/serial/zs.c +3 -3
index 9e6a873..d8c2809 100644
@@ -231,7 +231,7 @@ static int zs_receive_drain(struct zs_port *zport)
 {
 	int loops = 10000;
 
-	while ((read_zsreg(zport, R0) & Rx_CH_AV) && loops--)
+	while ((read_zsreg(zport, R0) & Rx_CH_AV) && --loops)
 		read_zsdata(zport);
 	return loops;
 }
@@ -241,7 +241,7 @@ static int zs_transmit_drain(struct zs_port *zport, int irq)
 	struct zs_scc *scc = zport->scc;
 	int loops = 10000;
 
-	while (!(read_zsreg(zport, R0) & Tx_BUF_EMP) && loops--) {
+	while (!(read_zsreg(zport, R0) & Tx_BUF_EMP) && --loops) {
 		zs_spin_unlock_cond_irq(&scc->zlock, irq);
 		udelay(2);
 		zs_spin_lock_cond_irq(&scc->zlock, irq);
@@ -254,7 +254,7 @@ static int zs_line_drain(struct zs_port *zport, int irq)
 	struct zs_scc *scc = zport->scc;
 	int loops = 10000;
 
-	while (!(read_zsreg(zport, R1) & ALL_SNT) && loops--) {
+	while (!(read_zsreg(zport, R1) & ALL_SNT) && --loops) {
 		zs_spin_unlock_cond_irq(&scc->zlock, irq);
 		udelay(2);
 		zs_spin_lock_cond_irq(&scc->zlock, irq);
Acked-by on v1 1/1
#1 lore raw

From: Roel Kluin <[email protected]>
Date: Thu, 04 Jun 2009 01:20:33 +0200

Quoted 8 lines

In zs_console_putchar() occurs:

	if (zs_transmit_drain(zport, irq))
		write_zsdata(zport, ch);

However if in zs_transmit_drain() no empty Tx Buffer occurs, limit reaches -1 => true, and the write still occurs.

Quoted 7 lines

This patch changes postfix to prefix decrements in this and similar functions to prevent similar mistakes in the future. This decreases the iterations with one but the chosen loop count was arbitrary anyway.

In sunhv limit reaches -1, not 0, so the test is off by one.

Acked-by on v1 1/1
#2 lore raw
In reply to Roel Kluin
Quoted 8 lines · Roel Kluin

In zs_console_putchar() occurs:

	if (zs_transmit_drain(zport, irq))
		write_zsdata(zport, ch);

However if in zs_transmit_drain() no empty Tx Buffer occurs, limit reaches -1 => true, and the write still occurs.

Quoted 7 lines

This patch changes postfix to prefix decrements in this and similar functions to prevent similar mistakes in the future. This decreases the iterations with one but the chosen loop count was arbitrary anyway.

In sunhv limit reaches -1, not 0, so the test is off by one.

Maciej