[PATCH v1 1/2] parport: Clean up resources correctly when parport_register_port() fails

[PATCH v1 1/2] parport: Clean up resources correctly when parport_register_port() fails

[Andy Shevchenko]

The smatch warns about uncleaned resources in case the
parport_register_port() fails:

  parport_register_port() warn: '&tmp->full_list' not removed from list

This is indeed an issue introduced when converting code to use
kasprintf(). However, the whole kasprintf() dance in this case
is not needed as dev_set_name() can handle the formatted input
and produces the same result. So, the solution is to delegate
name forming to the dev_set_name() and make device_register()
error path to deal with error handling (via put_device() call).

Fixes: 8d8ae17eb0de ("parport: Use kasprintf() instead of fixed buffer formatting")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Closes: https://lore.kernel.org/r/202310180809.hepZB9k6-lkp@intel.com/
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/parport/share.c | 10 +++-------
 1 file changed, 3 insertions(+), 7 deletions(-)

diff --git a/drivers/parport/share.c b/drivers/parport/share.c
index a9a9cb0477ea..048a459ce4b0 100644
--- a/drivers/parport/share.c
+++ b/drivers/parport/share.c
@@ -361,7 +361,6 @@ static void free_port(struct device *dev)
 		kfree(port->probe_info[d].description);
 	}
 
-	kfree(port->name);
 	kfree(port);
 }
 
@@ -479,16 +478,13 @@ struct parport *parport_register_port(unsigned long base, int irq, int dma,
 	/*
 	 * Now that the portnum is known finish doing the Init.
 	 */
-	tmp->name = kasprintf(GFP_KERNEL, "parport%d", tmp->portnum);
-	if (!tmp->name) {
-		kfree(tmp);
-		return NULL;
-	}
-	dev_set_name(&tmp->bus_dev, tmp->name);
+	dev_set_name(&tmp->bus_dev, "parport%d", tmp->portnum);
 	tmp->bus_dev.bus = &parport_bus_type;
 	tmp->bus_dev.release = free_port;
 	tmp->bus_dev.type = &parport_device_type;
 
+	tmp->name = dev_name(&tmp->bus_dev);
+
 	for (device = 0; device < 5; device++)
 		/* assume the worst */
 		tmp->probe_info[device].class = PARPORT_CLASS_LEGACY;
-- 
2.40.0.1.gaa8946217a0b

[PATCH v1 2/2] parport: Drop even more unneeded NULL or 0 assignments

[Andy Shevchenko]

kzalloc() gives us a zeroed memory, no need to explicitly assing 0 or
NULL or similar to the members of the data structure that has been
allocated with the above mentioned API.

Note, the initializstion of full_list member is not needed anymore
as list_add_tail will rewrite the contents of the prev and next
pointers.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---
 drivers/parport/share.c | 3 ---
 1 file changed, 3 deletions(-)

diff --git a/drivers/parport/share.c b/drivers/parport/share.c
index 048a459ce4b0..e21831d93305 100644
--- a/drivers/parport/share.c
+++ b/drivers/parport/share.c
@@ -459,7 +459,6 @@ struct parport *parport_register_port(unsigned long base, int irq, int dma,
 	sema_init(&tmp->ieee1284.irq, 0);
 	tmp->spintime = parport_default_spintime;
 	atomic_set(&tmp->ref_count, 1);
-	INIT_LIST_HEAD(&tmp->full_list);
 
 	/* Search for the lowest free parport number. */
 
@@ -489,8 +488,6 @@ struct parport *parport_register_port(unsigned long base, int irq, int dma,
 		/* assume the worst */
 		tmp->probe_info[device].class = PARPORT_CLASS_LEGACY;
 
-	tmp->waithead = tmp->waittail = NULL;
-
 	ret = device_register(&tmp->bus_dev);
 	if (ret) {
 		put_device(&tmp->bus_dev);
-- 
2.40.0.1.gaa8946217a0b