Skip to contentSkip to conversation

spi: fix refcount-related spidev oops-on-rmmod

PATCH In review David Brownell 2 participants 3 messages May 22, 2008 – May 24, 2008 2 messages not in the archive SPI

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 spi: fix refcount-related spidev oops-on-rmmod

from: David Brownell <[email protected]>

This addresses other oopsing paths in "spidev" by changing how it
manages refcounting. It decouples the lifecycle of the per-device
data from the class device (not just the spi device):

- Use class_{create,destroy} not class_{register,unregister}.
- Use device_{create,destroy} not device_{register,unregister}.
- Free the per-device data only when TWO conditions are true:
* Driver is unbound from underlying SPI device, and
* Device is no longer open (new)

Also, spi_{get,set}_drvdata not dev_{get,set}_drvdata for simpler code.

---
 drivers/spi/spidev.c |   64 +++++++++++++++++++++++++++------------------------
 1 file changed, 34 insertions(+), 30 deletions(-)
drivers/spi/spidev.c +34 -30
@@ -25,6 +25,7 @@
 #include <linux/ioctl.h>
 #include <linux/fs.h>
 #include <linux/device.h>
+#include <linux/err.h>
 #include <linux/list.h>
 #include <linux/errno.h>
 #include <linux/mutex.h>
@@ -67,11 +68,12 @@ static unsigned long	minors[N_SPI_MINORS
 				| SPI_LSB_FIRST | SPI_3WIRE | SPI_LOOP)
 
 struct spidev_data {
-	struct device		dev;
+	dev_t			devt;
 	spinlock_t		spi_lock;
 	struct spi_device	*spi;
 	struct list_head	device_entry;
 
+	/* buffer is NULL unless this device is open (users > 0) */
 	struct mutex		buf_lock;
 	unsigned		users;
 	u8			*buffer;
@@ -467,7 +469,7 @@ static int spidev_open(struct inode *ino
 	mutex_lock(&device_list_lock);
 
 	list_for_each_entry(spidev, &device_list, device_entry) {
-		if (spidev->dev.devt == inode->i_rdev) {
+		if (spidev->devt == inode->i_rdev) {
 			status = 0;
 			break;
 		}
@@ -500,10 +502,22 @@ static int spidev_release(struct inode *
 	mutex_lock(&device_list_lock);
 	spidev = filp->private_data;
 	filp->private_data = NULL;
+
+	/* last close? */
 	spidev->users--;
 	if (!spidev->users) {
+		int		dofree;
+
 		kfree(spidev->buffer);
 		spidev->buffer = NULL;
+
+		/* ... after we unbound from the underlying device? */
+		spin_lock_irq(&spidev->spi_lock);
+		dofree = (spidev->spi == NULL);
+		spin_unlock_irq(&spidev->spi_lock);
+
+		if (dofree)
+			kfree(spidev);
 	}
 	mutex_unlock(&device_list_lock);
 
@@ -530,19 +544,7 @@ static struct file_operations spidev_fop
  * It also simplifies memory management.
  */
 
-static void spidev_classdev_release(struct device *dev)
-{
-	struct spidev_data	*spidev;
-
-	spidev = container_of(dev, struct spidev_data, dev);
-	kfree(spidev);
-}
-
-static struct class spidev_class = {
-	.name		= "spidev",
-	.owner		= THIS_MODULE,
-	.dev_release	= spidev_classdev_release,
-};
+static struct class *spidev_class;
 
 /*-------------------------------------------------------------------------*/
 
@@ -570,20 +572,20 @@ static int spidev_probe(struct spi_devic
 	mutex_lock(&device_list_lock);
 	minor = find_first_zero_bit(minors, N_SPI_MINORS);
 	if (minor < N_SPI_MINORS) {
-		spidev->dev.parent = &spi->dev;
-		spidev->dev.class = &spidev_class;
-		spidev->dev.devt = MKDEV(SPIDEV_MAJOR, minor);
-		snprintf(spidev->dev.bus_id, sizeof spidev->dev.bus_id,
+		struct device *dev;
+
+		spidev->devt = MKDEV(SPIDEV_MAJOR, minor);
+		dev = device_create(spidev_class, &spi->dev, spidev->devt,
 				"spidev%d.%d",
 				spi->master->bus_num, spi->chip_select);
-		status = device_register(&spidev->dev);
+		status = IS_ERR(dev) ? PTR_ERR(dev) : 0;
 	} else {
 		dev_dbg(&spi->dev, "no minor number available!\n");
 		status = -ENODEV;
 	}
 	if (status == 0) {
 		set_bit(minor, minors);
-		dev_set_drvdata(&spi->dev, spidev);
+		spi_set_drvdata(spi, spidev);
 		list_add(&spidev->device_entry, &device_list);
 	}
 	mutex_unlock(&device_list_lock);
@@ -596,19 +598,21 @@ static int spidev_probe(struct spi_devic
 
 static int spidev_remove(struct spi_device *spi)
 {
-	struct spidev_data	*spidev = dev_get_drvdata(&spi->dev);
+	struct spidev_data	*spidev = spi_get_drvdata(spi);
 
 	/* make sure ops on existing fds can abort cleanly */
 	spin_lock_irq(&spidev->spi_lock);
 	spidev->spi = NULL;
+	spi_set_drvdata(spi, NULL);
 	spin_unlock_irq(&spidev->spi_lock);
 
 	/* prevent new opens */
 	mutex_lock(&device_list_lock);
 	list_del(&spidev->device_entry);
-	dev_set_drvdata(&spi->dev, NULL);
-	clear_bit(MINOR(spidev->dev.devt), minors);
-	device_unregister(&spidev->dev);
+	device_destroy(spidev_class, spidev->devt);
+	clear_bit(MINOR(spidev->devt), minors);
+	if (spidev->users == 0)
+		kfree(spidev);
 	mutex_unlock(&device_list_lock);
 
 	return 0;
@@ -644,15 +648,15 @@ static int __init spidev_init(void)
 	if (status < 0)
 		return status;
 
-	status = class_register(&spidev_class);
-	if (status < 0) {
+	spidev_class = class_create(THIS_MODULE, "spidev");
+	if (IS_ERR(spidev_class)) {
 		unregister_chrdev(SPIDEV_MAJOR, spidev_spi.driver.name);
-		return status;
+		return PTR_ERR(spidev_class);
 	}
 
 	status = spi_register_driver(&spidev_spi);
 	if (status < 0) {
-		class_unregister(&spidev_class);
+		class_destroy(spidev_class);
 		unregister_chrdev(SPIDEV_MAJOR, spidev_spi.driver.name);
 	}
 	return status;
@@ -662,7 +666,7 @@ module_init(spidev_init);
 static void __exit spidev_exit(void)
 {
 	spi_unregister_driver(&spidev_spi);
-	class_unregister(&spidev_class);
+	class_destroy(spidev_class);
 	unregister_chrdev(SPIDEV_MAJOR, spidev_spi.driver.name);
 }
 module_exit(spidev_exit);

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

A message that is not in the archive (sent off-list or not yet received): 200805221348.38163.david-b-yBeKhBN/[email protected]

#2 lore raw

drivers/spi/spidev.c static int spidev_release(struct inode * · line 520

Quoted 7 lines · On Thu, 22 May 2008 13:48:38 -0700 David Brownell…
drivers/spi/spidev.c +7 -0
+		/* ... after we unbound from the underlying device? */
+		spin_lock_irq(&spidev->spi_lock);
+		dofree = (spidev->spi == NULL);
+		spin_unlock_irq(&spidev->spi_lock);
+
+		if (dofree)
+			kfree(spidev);

that's very weird-looking code.

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/

A message that is not in the archive (sent off-list or not yet received): 20080523174814.cd2952c3.akpm-de/[email protected]

#4 lore raw
Quoted 11 lines · Andrew Morton

drivers/spi/spidev.c static int spidev_release(struct inode * · line 520

Quoted 7 lines · On Thu, 22 May 2008 13:48:38 -0700 David Brownell…
drivers/spi/spidev.c +7 -0
+		/* ... after we unbound from the underlying device? */
+		spin_lock_irq(&spidev->spi_lock);
+		dofree = (spidev->spi == NULL);
+		spin_unlock_irq(&spidev->spi_lock);
+
+		if (dofree)
+			kfree(spidev);

that's very weird-looking code.

Just sticking to the locking policy that access to that field is always protected by that lock ... "clearly OK", even if it might be a bit on the overkill side.

I'd rather be consistent in that than have to worry about whether, in this one path, it's OK to switch to a different policy since something else ensures that didn't race with spidev_remove().

- Dave

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/