linux-kernel - [PATCH v2] tty: serial: fsl_lpuart: fix race on RX DMA shutdown

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

Message-Id: <20230309134302.74940-2-alexander.sverdlin@siemens.com>
Date:   Thu,  9 Mar 2023 14:43:02 +0100
From:   "A. Sverdlin" <alexander.sverdlin@...mens.com>
To:     NXP Linux Team <linux-imx@....com>
Cc:     Alexander Sverdlin <alexander.sverdlin@...mens.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Jiri Slaby <jirislaby@...nel.org>,
        linux-serial@...r.kernel.org, Dong Aisheng <aisheng.dong@....com>,
        Shawn Guo <shawnguo@...nel.org>,
        Sascha Hauer <s.hauer@...gutronix.de>,
        Pengutronix Kernel Team <kernel@...gutronix.de>,
        Fabio Estevam <festevam@...il.com>, linux-i2c@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: [PATCH v2] tty: serial: fsl_lpuart: fix race on RX DMA shutdown

From: Alexander Sverdlin <alexander.sverdlin@...mens.com>

>From time to time DMA completion can come in the middle of DMA shutdown:

<process ctx>:				<IRQ>:
lpuart32_shutdown()
  lpuart_dma_shutdown()
    del_timer_sync()
					lpuart_dma_rx_complete()
					  lpuart_copy_rx_to_tty()
					    mod_timer()
    lpuart_dma_rx_free()

When the timer fires a bit later, sport->dma_rx_desc is NULL:

Unable to handle kernel NULL pointer dereference at virtual address 0000000000000004
pc : lpuart_copy_rx_to_tty+0xcc/0x5bc
lr : lpuart_timer_func+0x1c/0x2c
Call trace:
 lpuart_copy_rx_to_tty
 lpuart_timer_func
 call_timer_fn
 __run_timers.part.0
 run_timer_softirq
 __do_softirq
 __irq_exit_rcu
 irq_exit
 handle_domain_irq
 gic_handle_irq
 call_on_irq_stack
 do_interrupt_handler
 ...

To fix this fold del_timer_sync() into lpuart_dma_rx_free() after
dmaengine_terminate_sync() to make sure timer will not be re-started in
lpuart_copy_rx_to_tty() <= lpuart_dma_rx_complete().

Fixes: 4a8588a1cf86 ("serial: fsl_lpuart: delete timer on shutdown")
Signed-off-by: Alexander Sverdlin <alexander.sverdlin@...mens.com>
---
 drivers/tty/serial/fsl_lpuart.c | 11 +++--------
 1 file changed, 3 insertions(+), 8 deletions(-)

Changelog:
v2: added "Fixes:" tag

diff --git a/drivers/tty/serial/fsl_lpuart.c b/drivers/tty/serial/fsl_lpuart.c
index e945f41b93d43..47c267ee22e04 100644
--- a/drivers/tty/serial/fsl_lpuart.c
+++ b/drivers/tty/serial/fsl_lpuart.c
@@ -1354,6 +1354,7 @@ static void lpuart_dma_rx_free(struct uart_port *port)
 	struct dma_chan *chan = sport->dma_rx_chan;
 
 	dmaengine_terminate_sync(chan);
+	del_timer_sync(&sport->lpuart_timer);
 	dma_unmap_sg(chan->device->dev, &sport->rx_sgl, 1, DMA_FROM_DEVICE);
 	kfree(sport->rx_ring.buf);
 	sport->rx_ring.tail = 0;
@@ -1813,7 +1814,6 @@ static int lpuart32_startup(struct uart_port *port)
 static void lpuart_dma_shutdown(struct lpuart_port *sport)
 {
 	if (sport->lpuart_dma_rx_use) {
-		del_timer_sync(&sport->lpuart_timer);
 		lpuart_dma_rx_free(&sport->port);
 		sport->lpuart_dma_rx_use = false;
 	}
@@ -1973,10 +1973,8 @@ lpuart_set_termios(struct uart_port *port, struct ktermios *termios,
 	 * Since timer function acqures sport->port.lock, need to stop before
 	 * acquring same lock because otherwise del_timer_sync() can deadlock.
 	 */
-	if (old && sport->lpuart_dma_rx_use) {
-		del_timer_sync(&sport->lpuart_timer);
+	if (old && sport->lpuart_dma_rx_use)
 		lpuart_dma_rx_free(&sport->port);
-	}
 
 	spin_lock_irqsave(&sport->port.lock, flags);
 
@@ -2210,10 +2208,8 @@ lpuart32_set_termios(struct uart_port *port, struct ktermios *termios,
 	 * Since timer function acqures sport->port.lock, need to stop before
 	 * acquring same lock because otherwise del_timer_sync() can deadlock.
 	 */
-	if (old && sport->lpuart_dma_rx_use) {
-		del_timer_sync(&sport->lpuart_timer);
+	if (old && sport->lpuart_dma_rx_use)
 		lpuart_dma_rx_free(&sport->port);
-	}
 
 	spin_lock_irqsave(&sport->port.lock, flags);
 
@@ -3014,7 +3010,6 @@ static int lpuart_suspend(struct device *dev)
 			 * cannot resume as expected, hence gracefully release the
 			 * Rx DMA path before suspend and start Rx DMA path on resume.
 			 */
-			del_timer_sync(&sport->lpuart_timer);
 			lpuart_dma_rx_free(&sport->port);
 
 			/* Disable Rx DMA to use UART port as wakeup source */
-- 
2.34.1