On Mon, 21 Dec, at 05:04:11PM, Bryan O'Donoghue wrote:

quoted

+static int efi_capsule_open(struct inode *inode, struct file *file)
+{
+	struct capsule_info *cap_info;
+
+	cap_info = kzalloc(sizeof(*cap_info), GFP_KERNEL);
+	if (!cap_info)
+		return -ENOMEM;
+
+	file->private_data = cap_info;
+
+	return 0;
+}

You have a memory leak here don't you ?

if I do 
for (i = 0; i < N; i++) {
	fd = open(/dev/your_node);
	close(fd);
}

You'll leak that kzalloc...

Nope, it gets freed in efi_capsule_release(). Though the code does
leak cap_info->pages if a capsule is successfully submitted to the
firmware.