linux-kernel - [PATCH] serial: amba-pl011: do not time out prematurely when draining tx fifo

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-Id: <20220408233503.7251-1-LinoSanfilippo@gmx.de>
Date:   Sat,  9 Apr 2022 01:35:02 +0200
From:   Lino Sanfilippo <LinoSanfilippo@....de>
To:     gregkh@...uxfoundation.org
Cc:     linux@...linux.org.uk, jirislaby@...nel.org,
        linux-serial@...r.kernel.org, linux-kernel@...r.kernel.org,
        p.rosenberger@...bus.com, lukas@...ner.de,
        Lino Sanfilippo <LinoSanfilippo@....de>, stable@...r.kernel.org
Subject: [PATCH] serial: amba-pl011: do not time out prematurely when draining tx fifo

The current timeout for draining the tx fifo in RS485 mode is calculated by
multiplying the time it takes to transmit one character (with the given
baud rate) with the maximal number of characters in the tx queue.

This timeout is too short for two reasons:
First when calculating the time to transmit one character integer division
is used which may round down the result in case of a remainder of the
division.

Fix this by rounding up the division result.

Second the hardware may need additional time (e.g for first putting the
characters from the fifo into the shift register) before the characters are
actually put onto the wire.

To be on the safe side double the current maximum number of iterations
that are used to wait for the queue draining.

Fixes: 8d479237727c ("serial: amba-pl011: add RS485 support")
Cc: stable@...r.kernel.org
Signed-off-by: Lino Sanfilippo <LinoSanfilippo@....de>
---
 drivers/tty/serial/amba-pl011.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/tty/serial/amba-pl011.c b/drivers/tty/serial/amba-pl011.c
index 51ecb050ae40..4d11a3e547f9 100644
--- a/drivers/tty/serial/amba-pl011.c
+++ b/drivers/tty/serial/amba-pl011.c
@@ -1255,13 +1255,18 @@ static inline bool pl011_dma_rx_running(struct uart_amba_port *uap)
 
 static void pl011_rs485_tx_stop(struct uart_amba_port *uap)
 {
+	/*
+	 * To be on the safe side only time out after twice as many iterations
+	 * as fifo size.
+	 */
+	const int MAX_TX_DRAIN_ITERS = uap->port.fifosize * 2;
 	struct uart_port *port = &uap->port;
 	int i = 0;
 	u32 cr;
 
 	/* Wait until hardware tx queue is empty */
 	while (!pl011_tx_empty(port)) {
-		if (i == port->fifosize) {
+		if (i > MAX_TX_DRAIN_ITERS) {
 			dev_warn(port->dev,
 				 "timeout while draining hardware tx queue\n");
 			break;
@@ -2052,7 +2057,7 @@ pl011_set_termios(struct uart_port *port, struct ktermios *termios,
 	 * with the given baud rate. We use this as the poll interval when we
 	 * wait for the tx queue to empty.
 	 */
-	uap->rs485_tx_drain_interval = (bits * 1000 * 1000) / baud;
+	uap->rs485_tx_drain_interval = DIV_ROUND_UP(bits * 1000 * 1000, baud);
 
 	pl011_setup_status_masks(port, termios);
 

base-commit: 1a3b1bba7c7a5eb8a11513cf88427cb9d77bc60a
-- 
2.35.1