[PATCH 08/67] serial: fix wakup races in the mrst_max3110 driver

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
From: Greg Kroah-Hartman
Date: Thursday, August 5, 2010 - 3:22 pm

From: Arjan van de Ven <arjan@linux.intel.com>

The mrst_max3110 driver had a set of unsafe wakeup sequences
along the following line:

if (!atomic_read(&foo)) {
  atomic_set(&foo, 1);
  wake_up(worker_thread);
}
and the worker thread would do

if (atomic_read(&foo)) {
  do_work();
  atomic_set(&foo, 0);
}

which can result in various missed wakups due to test-then-set races,
as well as due to clear-after-work instead of clear-before-work.

This patch fixes these races by using the proper bit test-and-set operations,
and by doing clear-before-work.

Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
Signed-off-by: Alan Cox <alan@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
---
 drivers/serial/mrst_max3110.c |   46 ++++++++++++++++-------------------------
 1 files changed, 18 insertions(+), 28 deletions(-)

diff --git a/drivers/serial/mrst_max3110.c b/drivers/serial/mrst_max3110.c
index 0341853..f6ad1ec 100644
--- a/drivers/serial/mrst_max3110.c
+++ b/drivers/serial/mrst_max3110.c
@@ -48,6 +48,10 @@
 
 #define PR_FMT	"mrst_max3110: "
 
+#define UART_TX_NEEDED 1
+#define CON_TX_NEEDED  2
+#define BIT_IRQ_PENDING    3
+
 struct uart_max3110 {
 	struct uart_port port;
 	struct spi_device *spi;
@@ -63,15 +67,13 @@ struct uart_max3110 {
 	u8 clock;
 	u8 parity, word_7bits;
 
-	atomic_t uart_tx_need;
+	unsigned long uart_flags;
 
 	/* console related */
 	struct circ_buf con_xmit;
-	atomic_t con_tx_need;
 
 	/* irq related */
 	u16 irq;
-	atomic_t irq_pending;
 };
 
 /* global data structure, may need be removed */
@@ -176,10 +178,9 @@ static void serial_m3110_con_putchar(struct uart_port *port, int ch)
 		xmit->head = (xmit->head + 1) & (PAGE_SIZE - 1);
 	}
 
-	if (!atomic_read(&max->con_tx_need)) {
-		atomic_set(&max->con_tx_need, 1);
+
+	if (!test_and_set_bit(CON_TX_NEEDED, &max->uart_flags))
 		wake_up_process(max->main_thread);
-	}
 }
 
 /*
@@ -318,10 +319,8 @@ static void serial_m3110_start_tx(struct uart_port *port)
 	struct uart_max3110 *max =
 		container_of(port, struct uart_max3110, port);
 
-	if (!atomic_read(&max->uart_tx_need)) {
-		atomic_set(&max->uart_tx_need, 1);
+	if (!test_and_set_bit(UART_TX_NEEDED, &max->uart_flags))
 		wake_up_process(max->main_thread);
-	}
 }
 
 static void receive_chars(struct uart_max3110 *max, unsigned char *str, int len)
@@ -392,32 +391,23 @@ static int max3110_main_thread(void *_max)
 	pr_info(PR_FMT "start main thread\n");
 
 	do {
-		wait_event_interruptible(*wq, (atomic_read(&max->irq_pending) ||
-					       atomic_read(&max->con_tx_need) ||
-					     atomic_read(&max->uart_tx_need)) ||
-					     kthread_should_stop());
+		wait_event_interruptible(*wq, max->uart_flags || kthread_should_stop());
 
 		mutex_lock(&max->thread_mutex);
 
-#ifdef CONFIG_MRST_MAX3110_IRQ
-		if (atomic_read(&max->irq_pending)) {
+		if (test_and_clear_bit(BIT_IRQ_PENDING, &max->uart_flags))
 			max3110_console_receive(max);
-			atomic_set(&max->irq_pending, 0);
-		}
-#endif
 
 		/* first handle console output */
-		if (atomic_read(&max->con_tx_need)) {
+		if (test_and_clear_bit(CON_TX_NEEDED, &max->uart_flags))
 			send_circ_buf(max, xmit);
-			atomic_set(&max->con_tx_need, 0);
-		}
 
 		/* handle uart output */
-		if (atomic_read(&max->uart_tx_need)) {
+		if (test_and_clear_bit(UART_TX_NEEDED, &max->uart_flags))
 			transmit_char(max);
-			atomic_set(&max->uart_tx_need, 0);
-		}
+
 		mutex_unlock(&max->thread_mutex);
+
 	} while (!kthread_should_stop());
 
 	return ret;
@@ -430,10 +420,9 @@ static irqreturn_t serial_m3110_irq(int irq, void *dev_id)
 
 	/* max3110's irq is a falling edge, not level triggered,
 	 * so no need to disable the irq */
-	if (!atomic_read(&max->irq_pending)) {
-		atomic_inc(&max->irq_pending);
+	if (!test_and_set_bit(BIT_IRQ_PENDING, &max->uart_flags))
 		wake_up_process(max->main_thread);
-	}
+
 	return IRQ_HANDLED;
 }
 #else
@@ -753,7 +742,8 @@ static int serial_m3110_probe(struct spi_device *spi)
 	max->baud = 0;
 
 	max->cur_conf = 0;
-	atomic_set(&max->irq_pending, 0);
+	max->uart_flags = 0;
+
 	/* Check if reading configuration register returns something sane */
 
 	res = RC_TAG;
-- 
1.7.2

--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
[GIT PATCH] TTY patches for 2.6.36, Greg KH, (Thu Aug 5, 2:35 pm)
[PATCH 01/67] n_gsm.c: removed duplicated #includes, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 02/67] serial: There's no config CONSOLE, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 03/67] vt: clean up the code - use kernel library, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 04/67] serial: add UART_CAP_EFR and UART_CAP_SLEEP ..., Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 05/67] mrst_max3110: add UART driver for Max3110 on ..., Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 06/67] max3110 sanity check a register, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 07/67] serial: replace open coded mutex with a real ..., Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 08/67] serial: fix wakup races in the mrst_max3110 ..., Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 09/67] tty: Remove Hayes ESP ioctls, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 10/67] tty: remove remaining Hayes ESP ioctls, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 11/67] tty: Add EXTPROC support for LINEMODE, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 12/67] vt/console: try harder to print output when ..., Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 13/67] stallion: prune lock_kernel calls, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 14/67] istallion: use bit ops for the board flags, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 15/67] riscom8: kill use of lock_kernel, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 16/67] isicom: kill off the BKL, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 17/67] rocket: kill BKL, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 18/67] synclink: kill the big kernel lock, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 19/67] cyclades: Kill off BKL usage, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 20/67] epca: Kill the big kernel lock, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 21/67] specialix: Kill the BKL, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 22/67] tty: Fix the digi acceleport driver NULL checks, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 23/67] synclink: reworking locking a bit, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 24/67] tty: serial - fix various misuses/mishandlin ..., Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 25/67] tty: serial - fix tty back references in termios, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 26/67] tty: serial - fix tty referencing in set_ldisc, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 27/67] vc: Locking clean up, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 28/67] tty: Make vt's have a tty_port, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 29/67] tty: Move the vt_tty field from the vc_data ..., Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 30/67] serial: Change the wait for carrier locking, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 31/67] serial: add port helpers, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 32/67] serial: trim locking on the helpers, Greg Kroah-Hartman, (Thu Aug 5, 3:22 pm)
[PATCH 33/67] serial: Use block_til_ready helper, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 34/67] serial: fix termios settings in open, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 35/67] tty: replace BKL with a new tty_lock, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 36/67] tty: never hold BTM while getting tty_mutex, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 37/67] tty: fix console_sem lock order, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 38/67] cdc-acm: remove dead code, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 39/67] tty: introduce wait_event_interruptible_tty, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 40/67] tty: reorder ldisc locking, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 41/67] tty: untangle locking of wait_until_sent, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 42/67] tty: remove tty_lock_nested, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 43/67] tty: implement BTM as mutex instead of BKL, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 44/67] tty: release BTM while sleeping in block_til ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 45/67] 8250: fix set_ldisc operation, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 46/67] tty: avoid recursive BTM in pty_close, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 47/67] serial: max3107: introduce a max3107 driver, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 48/67] serial: max3107: Abstract out the platform s ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 49/67] tty_io: remove casts from void*, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 50/67] vt: Fix warning: statement with no effect du ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 51/67] serial: crisv10: formatting of pointers in p ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 53/67] serial: fix missing bit coverage of ASYNC_FLAGS, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 54/67] serial: general fixes in the serial_rs485 st ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 55/67] serial: mcf: don't take spinlocks in already ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 56/67] serial: MMIO32 support for 8250_early.c, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 57/67] timbuart: use __devinit and __devexit macros ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 58/67] serial: 68328serial.c: remove dead (ALMA_ANS ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 59/67] serial: add support for OX16PCI958 card, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 60/67] mxser: remove unnesesary NULL check, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 61/67] hsu: driver for Medfield High Speed UART device, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 62/67] hsu: add a periodic timer to check dma rx ch ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 63/67] hsu: some code cleanup, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 64/67] hsu: call PCI pm hooks in suspend/resume fun ..., Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
[PATCH 65/67] serial: max3107: Fix gpiolib support, Greg Kroah-Hartman, (Thu Aug 5, 3:23 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Linus Torvalds, (Fri Aug 6, 11:40 am)
Re: [GIT PATCH] TTY patches for 2.6.36, Greg KH, (Fri Aug 6, 11:51 am)
Re: [GIT PATCH] TTY patches for 2.6.36, Greg KH, (Fri Aug 6, 12:37 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Linus Torvalds, (Fri Aug 6, 12:38 pm)
[PATCH retry] tty: implement BTM as mutex instead of BKL, Arnd Bergmann, (Fri Aug 6, 12:40 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Arnd Bergmann, (Fri Aug 6, 12:40 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Alan Cox, (Fri Aug 6, 12:45 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Arnd Bergmann, (Fri Aug 6, 12:58 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Linus Torvalds, (Fri Aug 6, 1:11 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Arnd Bergmann, (Fri Aug 6, 1:19 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Greg KH, (Fri Aug 6, 1:22 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Andy Whitcroft, (Mon Aug 9, 9:06 am)
Re: [GIT PATCH] TTY patches for 2.6.36, Arnd Bergmann, (Mon Aug 9, 11:38 am)
Re: [GIT PATCH] TTY patches for 2.6.36, Alan Cox, (Mon Aug 9, 12:24 pm)
Re: [GIT PATCH] TTY patches for 2.6.36, Andy Whitcroft, (Tue Aug 10, 12:31 am)