From: YoungJun Cho <yj44....@samsung.com>
This patch fixes wrong pointer access issue to filp->f_op and
filp->private_data.
The exynos_drm_gem_mmap_ioctl() changes filp->f_op and
filp->private_data temporarily and restore them to use
original ones in exynos_drm_gem_mmap_buffer() but there
was no lock between the changing and the restoring so
wrong pointer access to filp->f_op and filp->private_data
was induced by vm close callback.
So this patch uses mutex lock properly to resolve this issue.
Signed-off-by: YoungJun Cho <yj44.cho at samsung.com>
Signed-off-by: Inki Dae <inki.dae at samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park at samsung.com>
---
drivers/gpu/drm/exynos/exynos_drm_gem.c | 33 ++++++++++++++++++++-----------
1 file changed, 21 insertions(+), 12 deletions(-)
diff --git a/drivers/gpu/drm/exynos/exynos_drm_gem.c
b/drivers/gpu/drm/exynos/exynos_drm_gem.c
index 4731807..67e17ce 100644
--- a/drivers/gpu/drm/exynos/exynos_drm_gem.c
+++ b/drivers/gpu/drm/exynos/exynos_drm_gem.c
@@ -329,17 +329,11 @@ static struct drm_file *exynos_drm_find_drm_file(struct
drm_device *drm_dev,
{
struct drm_file *file_priv;
- mutex_lock(&drm_dev->struct_mutex);
-
/* find current process's drm_file from filelist. */
- list_for_each_entry(file_priv, &drm_dev->filelist, lhead) {
- if (file_priv->filp == filp) {
- mutex_unlock(&drm_dev->struct_mutex);
+ list_for_each_entry(file_priv, &drm_dev->filelist, lhead)
+ if (file_priv->filp == filp)
return file_priv;
- }
- }
- mutex_unlock(&drm_dev->struct_mutex);
WARN_ON(1);
return ERR_PTR(-EFAULT);
@@ -400,9 +394,7 @@ static int exynos_drm_gem_mmap_buffer(struct file *filp,
*/
drm_gem_object_reference(obj);
- mutex_lock(&drm_dev->struct_mutex);
drm_vm_open_locked(drm_dev, vma);
- mutex_unlock(&drm_dev->struct_mutex);
return 0;
}
@@ -432,6 +424,16 @@ int exynos_drm_gem_mmap_ioctl(struct drm_device *dev, void
*data,
}
/*
+ * We have to use gem object and its fops for specific mmaper,
+ * but vm_mmap() can deliver only filp. So we have to change
+ * filp->f_op and filp->private_data temporarily, then restore
+ * again. So it is important to keep lock until restoration the
+ * settings to prevent others from misuse of filp->f_op or
+ * filp->private_data.
+ */
+ mutex_lock(&dev->struct_mutex);
+
+ /*
* Set specific mmper's fops. And it will be restored by
* exynos_drm_gem_mmap_buffer to dev->driver->fops.
* This is used to call specific mapper temporarily.
@@ -448,13 +450,20 @@ int exynos_drm_gem_mmap_ioctl(struct drm_device *dev,
void *data,
addr = vm_mmap(file_priv->filp, 0, args->size,
PROT_READ | PROT_WRITE, MAP_SHARED, 0);
- drm_gem_object_unreference_unlocked(obj);
+ drm_gem_object_unreference(obj);
if (IS_ERR((void *)addr)) {
- file_priv->filp->private_data = file_priv;
+ /* check filp->f_op, filp->private_data are restored */
+ if (file_priv->filp->f_op == &exynos_drm_gem_fops) {
+ file_priv->filp->f_op = fops_get(dev->driver->fops);
+ file_priv->filp->private_data = file_priv;
+ }
+ mutex_unlock(&dev->struct_mutex);
return PTR_ERR((void *)addr);
}
+ mutex_unlock(&dev->struct_mutex);
+
args->mapped = addr;
DRM_DEBUG_KMS("mapped = 0x%lx\n", (unsigned long)args->mapped);
--
1.7.9.5
