On Tue, Oct 23, 2007 at 02:08:19PM -0700, Roland Dreier wrote:

quoted text
>  > +void pps_unregister_source(int source)
>  > ...
>  > +	wait_event(pps->usage_queue, atomic_read(&pps->usage) == 0);
>  > +
>  > +	pps_sysfs_remove_source_entry(pps);
>  > +	pps_unregister_cdev(pps);
>  > +	kfree(pps);
> 
> This reference counting looks dubious to me... later on in the code
> you have:
> 
>  > +static int pps_cdev_open(struct inode *inode, struct file *file)
>  > +{
>  > +	struct pps_device *pps = container_of(inode->i_cdev,
>  > +						struct pps_device, cdev);
>  > +
>  > +	/* Lock the PPS source against (possible) deregistration */
>  > +	atomic_inc(&pps->usage);
> 
> with no locking, so I see no reason why the atomic_inc() couldn't
> happen right after the wait_event() sees a count of 0 and lets the
> deregistration continue.  Which would lead to use-after-free.

Since pps_unregister_source() is defined as:

void pps_unregister_source(int source)
{
        struct pps_device *pps;

        spin_lock_irq(&idr_lock);
        pps = idr_find(&pps_idr, source);

        if (!pps) {
                spin_unlock_irq(&idr_lock);
                return;
        }

        /* This should be done first in order to deny IRQ handlers
         * to access PPS structs
         */

        idr_remove(&pps_idr, pps->id);
        spin_unlock_irq(&idr_lock);

        wait_event(pps->usage_queue, atomic_read(&pps->usage) == 0);

        pps_sysfs_remove_source_entry(pps);
        pps_unregister_cdev(pps);
        kfree(pps);
}

changing pps_cdev_open() as follow should resolve this problem:

static int pps_cdev_open(struct inode *inode, struct file *file)
{
        struct pps_device *pps = container_of(inode->i_cdev,
                                                struct pps_device, cdev);
        int found;

        spin_lock_irq(&idr_lock);
        found = idr_find(&pps_idr, pps->id) != NULL;

        /* Lock the PPS source against (possible) deregistration */
        if (found)
                atomic_inc(&pps->usage);

        spin_unlock_irq(&idr_lock);

        if (!found)
                return -ENODEV;

        file->private_data = pps;

        return 0;
}

Is that right?

Rodolfo

-- 

GNU/Linux Solutions                  e-mail:    giometti@enneenne.com
Linux Device Driver                             giometti@gnudd.com
Embedded Systems                     		giometti@linux.it
UNIX programming                     phone:     +39 349 2432127
-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/