[PATCH] serial: core: fix -EPROBE_DEFER handling in init

[PATCH] serial: core: fix -EPROBE_DEFER handling in init

[Dan Carpenter]

The -EPROBE_DEFER error path in serial_base_device_init() is a bit
awkward.  Before the call to device_initialize(dev) then we need to
manually release all the device resources.  And after the call then we
need to call put_device() to release the resources.  Doing either one
wrong will result in a leak or a use after free.

So let's wait to return -EPROBE_DEFER until after the call to
device_initialize(dev) so that way callers do not have to handle
-EPROBE_DEFER as a special case.  Now callers can just use put_device()
for clean up.

The second issue with the -EPROBE_DEFER path is that deferring is not
supposed to be a fatal error, but instead it's normal part of the
init process and the kernel recovers from it automatically.  That means
we should not print an error message but just a debug message on this
path.

Fixes: 539914240a01 ("serial: core: Fix probing serial_base_bus devices")
Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
---
 drivers/tty/serial/serial_base_bus.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/tty/serial/serial_base_bus.c b/drivers/tty/serial/serial_base_bus.c
index 9de9f6fa6481..6ff59c89d867 100644
--- a/drivers/tty/serial/serial_base_bus.c
+++ b/drivers/tty/serial/serial_base_bus.c
@@ -50,17 +50,17 @@ static int serial_base_device_init(struct uart_port *port,
 				   void (*release)(struct device *dev),
 				   int id)
 {
-	if (!serial_base_initialized) {
-		dev_err(port->dev, "uart_add_one_port() called before arch_initcall()?\n");
-		return -EPROBE_DEFER;
-	}
-
 	device_initialize(dev);
 	dev->type = type;
 	dev->parent = parent_dev;
 	dev->bus = &serial_base_bus_type;
 	dev->release = release;
 
+	if (!serial_base_initialized) {
+		dev_dbg(port->dev, "uart_add_one_port() called before arch_initcall()?\n");
+		return -EPROBE_DEFER;
+	}
+
 	return dev_set_name(dev, "%s.%s.%d", type->name, dev_name(port->dev), id);
 }
 
-- 
2.39.2

Re: [PATCH] serial: core: fix -EPROBE_DEFER handling in init

[Tony Lindgren]

* Dan Carpenter <dan.carpenter@linaro.org> [230614 06:36]:
> The -EPROBE_DEFER error path in serial_base_device_init() is a bit
> awkward.  Before the call to device_initialize(dev) then we need to
> manually release all the device resources.  And after the call then we
> need to call put_device() to release the resources.  Doing either one
> wrong will result in a leak or a use after free.
> 
> So let's wait to return -EPROBE_DEFER until after the call to
> device_initialize(dev) so that way callers do not have to handle
> -EPROBE_DEFER as a special case.  Now callers can just use put_device()
> for clean up.
> 
> The second issue with the -EPROBE_DEFER path is that deferring is not
> supposed to be a fatal error, but instead it's normal part of the
> init process and the kernel recovers from it automatically.  That means
> we should not print an error message but just a debug message on this
> path.
> 
> Fixes: 539914240a01 ("serial: core: Fix probing serial_base_bus devices")
> Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>

Thanks for fixing these issues:

Reviewed-by: Tony Lindgren <tony@atomide.com>