Waiting for the device release method to be called when
going through i2c_del_adapter is wrong and leads to deadlock
when removing an i2c mux device.
For instance when using the OF i2c mux unitest removal we get this.
[<c055dfdc>] (__schedule) from [<c0561b74>] (schedule_timeout+0x18/0x198)
[<c0561b74>] (schedule_timeout) from [<c055ee74>] (wait_for_common+0xf8/0x138)
[<c055ee74>] (wait_for_common) from [<c03f724c>] (i2c_del_adapter+0x174/0x1cc)
[<c03f724c>] (i2c_del_adapter) from [<c03f8480>] (i2c_del_mux_adapter+0x48/0x60)
[<c03f8480>] (i2c_del_mux_adapter) from [<c046cf00>] (selftest_i2c_mux_remove+0x28/0x34)
[<c046cf00>] (selftest_i2c_mux_remove) from [<c03f5234>] (i2c_device_remove+0x34/0x70)
[<c03f5234>] (i2c_device_remove) from [<c03322ec>] (__device_release_driver+0x7c/0xc0)
[<c03322ec>] (__device_release_driver) from [<c0332968>] (driver_detach+0x8c/0xb4)
[<c0332968>] (driver_detach) from [<c0332088>] (bus_remove_driver+0x64/0x8c)
[<c0332088>] (bus_remove_driver) from [<c07f1858>] (of_selftest+0x1f84/0x20f0)
[<c07f1858>] (of_selftest) from [<c0008b04>] (do_one_initcall+0x104/0x1b4)
[<c0008b04>] (do_one_initcall) from [<c07c4d9c>] (kernel_init_freeable+0x110/0x1d8)
[<c07c4d9c>] (kernel_init_freeable) from [<c05591ec>] (kernel_init+0x8/0xe4)
[<c05591ec>] (kernel_init) from [<c000deb8>] (ret_from_fork+0x14/0x3c)
2 locks held by swapper/0/1:
#0: (&dev->mutex){......}, at: [<c0332944>] driver_detach+0x68/0xb4
#1: (&dev->mutex){......}, at: [<c0332954>] driver_detach+0x78/0xb4
Kernel panic - not syncing: hung_task: blocked tasks
CPU: 0 PID: 16 Comm: khungtaskd Not tainted 3.19.0-rc4-00022-g261647c #443
Hardware name: Generic AM33XX (Flattened Device Tree)
[<c00140a8>] (unwind_backtrace) from [<c00110dc>] (show_stack+0x10/0x14)
[<c00110dc>] (show_stack) from [<c055c3ac>] (dump_stack+0x70/0x8c)
[<c055c3ac>] (dump_stack) from [<c055ad10>] (panic+0x88/0x1f0)
[<c055ad10>] (panic) from [<c00adef8>] (watchdog+0x2d4/0x350)
[<c00adef8>] (watchdog) from [<c004e984>] (kthread+0xd8/0xec)
[<c004e984>] (kthread) from [<c000deb8>] (ret_from_fork+0x14/0x3c)
The device release method is never called and we hang waiting for it.
This patch is marked as an [RFC] since the original code seems to
really want to protect against a race from sysfs accessors, which
I don't see how it could be a problem.
Signed-off-by: Pantelis Antoniou <pantelis.antoniou@xxxxxxxxxxxx>
---
drivers/i2c/i2c-core.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/i2c/i2c-core.c b/drivers/i2c/i2c-core.c
index 39d25a8..e020a16 100644
--- a/drivers/i2c/i2c-core.c
+++ b/drivers/i2c/i2c-core.c
@@ -1185,7 +1185,7 @@ EXPORT_SYMBOL_GPL(i2c_new_dummy);
static void i2c_adapter_dev_release(struct device *dev)
{
struct i2c_adapter *adap = to_i2c_adapter(dev);
- complete(&adap->dev_released);
+ /* XXX complete(&adap->dev_released); */
}
/*
@@ -1797,11 +1797,11 @@ void i2c_del_adapter(struct i2c_adapter *adap)
dev_dbg(&adap->dev, "adapter [%s] unregistered\n", adap->name);
/* clean up the sysfs representation */
- init_completion(&adap->dev_released);
+ /* XXX init_completion(&adap->dev_released); */
device_unregister(&adap->dev);
/* wait for sysfs to drop all references */
- wait_for_completion(&adap->dev_released);
+ /* XXX wait_for_completion(&adap->dev_released); */
/* free bus id */
mutex_lock(&core_lock);
--
1.7.12
--
To unsubscribe from this list: send the line "unsubscribe linux-i2c" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
