Bug#625804: rtc/mc13xxx: don't call rtc_device_register with the lock held
Since commit
f44f7f9 (RTC: Initialize kernel state from RTC)
rtc_device_register reads the programmed alarm. As reading the alarm
needs to take the lock, release it before calling rtc_device_register.
This fixes a deadlock during boot:
INFO: task swapper:1 blocked for more than 120 seconds.
"echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
swapper D c02b175c 0 1 0 0x00000000
[<c02b175c>] (schedule+0x304/0x4f4) from [<c02b25a8>] (__mutex_lock_slowpath+0x7c/0x110)
[<c02b25a8>] (__mutex_lock_slowpath+0x7c/0x110) from [<c020b4cc>] (mc13xxx_rtc_read_time+0x1c/0x118)
[<c020b4cc>] (mc13xxx_rtc_read_time+0x1c/0x118) from [<c0208f04>] (__rtc_read_time+0x58/0x5c)
[<c0208f04>] (__rtc_read_time+0x58/0x5c) from [<c0209508>] (rtc_read_time+0x30/0x48)
[<c0209508>] (rtc_read_time+0x30/0x48) from [<c0209dd4>] (__rtc_read_alarm+0x1c/0x290)
[<c0209dd4>] (__rtc_read_alarm+0x1c/0x290) from [<c0208d58>] (rtc_device_register+0x150/0x27c)
[<c0208d58>] (rtc_device_register+0x150/0x27c) from [<c02b0b74>] (mc13xxx_rtc_probe+0x128/0x17c)
[<c02b0b74>] (mc13xxx_rtc_probe+0x128/0x17c) from [<c01d5280>] (platform_drv_probe+0x1c/0x24)
[<c01d5280>] (platform_drv_probe+0x1c/0x24) from [<c01d3e58>] (driver_probe_device+0x80/0x1a8)
[<c01d3e58>] (driver_probe_device+0x80/0x1a8) from [<c01d400c>] (__driver_attach+0x8c/0x90)
[<c01d400c>] (__driver_attach+0x8c/0x90) from [<c01d3654>] (bus_for_each_dev+0x60/0x8c)
[<c01d3654>] (bus_for_each_dev+0x60/0x8c) from [<c01d2f6c>] (bus_add_driver+0x180/0x248)
[<c01d2f6c>] (bus_add_driver+0x180/0x248) from [<c01d4664>] (driver_register+0x70/0x15c)
[<c01d4664>] (driver_register+0x70/0x15c) from [<c01d5700>] (platform_driver_probe+0x18/0x98)
[<c01d5700>] (platform_driver_probe+0x18/0x98) from [<c00273a8>] (do_one_initcall+0x2c/0x168)
[<c00273a8>] (do_one_initcall+0x2c/0x168) from [<c00083ac>] (kernel_init+0xa0/0x150)
[<c00083ac>] (kernel_init+0xa0/0x150) from [<c0033ff8>] (kernel_thread_exit+0x0/0x8)
Reported-by: Vagrant Cascadian <vagrant@debian.org>
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Closes: http://bugs.debian.org/625804
---
Hello Arnaud,
On Mon, Apr 04, 2011 at 04:52:54AM -0000, Arnaud Patard wrote:
> Index: imx-test/drivers/rtc/rtc-mc13xxx.c
> ===================================================================
> --- imx-test.orig/drivers/rtc/rtc-mc13xxx.c
> +++ imx-test/drivers/rtc/rtc-mc13xxx.c
> @@ -349,8 +349,13 @@ static int __devinit mc13xxx_rtc_probe(s
> if (ret)
> goto err_alarm_irq_request;
>
> + mc13xxx_unlock(mc13xxx);
> +
> priv->rtc = rtc_device_register(pdev->name,
> &pdev->dev, &mc13xxx_rtc_ops, THIS_MODULE);
> +
> + mc13xxx_lock(mc13xxx);
> +
> if (IS_ERR(priv->rtc)) {
> ret = PTR_ERR(priv->rtc);
the patch below is a bit better as it doesn't take the lock in the
likely case that rtc_device_register succeeds.
Best regards
Uwe
diff --git a/drivers/rtc/rtc-mc13xxx.c b/drivers/rtc/rtc-mc13xxx.c
index c5ac037..a1a278b 100644
--- a/drivers/rtc/rtc-mc13xxx.c
+++ b/drivers/rtc/rtc-mc13xxx.c
@@ -349,11 +349,15 @@ static int __devinit mc13xxx_rtc_probe(struct platform_device *pdev)
if (ret)
goto err_alarm_irq_request;
+ mc13xxx_unlock(mc13xxx);
+
priv->rtc = rtc_device_register(pdev->name,
&pdev->dev, &mc13xxx_rtc_ops, THIS_MODULE);
if (IS_ERR(priv->rtc)) {
ret = PTR_ERR(priv->rtc);
+ mc13xxx_lock(mc13xxx);
+
mc13xxx_irq_free(mc13xxx, MC13XXX_IRQ_TODA, priv);
err_alarm_irq_request:
@@ -365,12 +369,12 @@ err_reset_irq_status:
mc13xxx_irq_free(mc13xxx, MC13XXX_IRQ_RTCRST, priv);
err_reset_irq_request:
+ mc13xxx_unlock(mc13xxx);
+
platform_set_drvdata(pdev, NULL);
kfree(priv);
}
- mc13xxx_unlock(mc13xxx);
-
return ret;
}
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Reply to: