Skip to contentSkip to conversation

usb: gadget: use after free in dev_config

PATCH Superseded Hangyu Hua 2 participants 3 messages Dec 28, 2021 – Dec 30, 2021 USB

v1 of 4 Superseded Latest is v4 →

lore
Share & tools
lore

Hangyu Hua

author
#0 lore raw

PATCH 1/1 usb: gadget: use after free in dev_config

There are two bugs:
dev->buf does not need to be released if it already exists before executing dev_config.
dev->config and dev->hs_config and dev->dev need to be cleaned if dev_config fails to avoid UAF.

---
 drivers/usb/gadget/legacy/inode.c | 23 ++++++++++++++++++-----
 1 file changed, 18 insertions(+), 5 deletions(-)
drivers/usb/gadget/legacy/inode.c +18 -5
index 3b58f4fc0a80..2ea02887025b 100644
@@ -1826,8 +1826,9 @@ dev_config (struct file *fd, const char __user *buf, size_t len, loff_t *ptr)
 	spin_lock_irq (&dev->lock);
 	value = -EINVAL;
 	if (dev->buf) {
+		spin_unlock_irq(&dev->lock);
 		kfree(kbuf);
-		goto fail;
+		return value;
 	}
 	dev->buf = kbuf;
 
@@ -1835,8 +1836,10 @@ dev_config (struct file *fd, const char __user *buf, size_t len, loff_t *ptr)
 	dev->config = (void *) kbuf;
 	total = le16_to_cpu(dev->config->wTotalLength);
 	if (!is_valid_config(dev->config, total) ||
-			total > length - USB_DT_DEVICE_SIZE)
+			total > length - USB_DT_DEVICE_SIZE) {
+		dev->config = NULL;
 		goto fail;
+	}
 	kbuf += total;
 	length -= total;
 
@@ -1845,8 +1848,11 @@ dev_config (struct file *fd, const char __user *buf, size_t len, loff_t *ptr)
 		dev->hs_config = (void *) kbuf;
 		total = le16_to_cpu(dev->hs_config->wTotalLength);
 		if (!is_valid_config(dev->hs_config, total) ||
-				total > length - USB_DT_DEVICE_SIZE)
+				total > length - USB_DT_DEVICE_SIZE) {
+			dev->config = NULL;
+			dev->hs_config = NULL;
 			goto fail;
+		}
 		kbuf += total;
 		length -= total;
 	} else {
@@ -1856,13 +1862,20 @@ dev_config (struct file *fd, const char __user *buf, size_t len, loff_t *ptr)
 	/* could support multiple configs, using another encoding! */
 
 	/* device descriptor (tweaked for paranoia) */
-	if (length != USB_DT_DEVICE_SIZE)
+	if (length != USB_DT_DEVICE_SIZE) {
+		dev->config = NULL;
+		dev->hs_config = NULL;
 		goto fail;
+	}
 	dev->dev = (void *)kbuf;
 	if (dev->dev->bLength != USB_DT_DEVICE_SIZE
 			|| dev->dev->bDescriptorType != USB_DT_DEVICE
-			|| dev->dev->bNumConfigurations != 1)
+			|| dev->dev->bNumConfigurations != 1) {
+		dev->config = NULL;
+		dev->hs_config = NULL;
+		dev->dev = NULL;
 		goto fail;
+	}
 	dev->dev->bcdUSB = cpu_to_le16 (0x0200);
 
 	/* triggers gadgetfs_bind(); then we can enumerate. */
Signature · 1 line
2.25.1
#1 lore raw
Quoted 1 line · Hangyu Hua

There are two bugs:

You should break this up into two separate patches, one for each bug.

Quoted 2 lines

dev->buf does not need to be released if it already exists before executing dev_config.

That's right. The call to dev_config should fail without changing any of the stored values.

Quoted 2 lines

dev->config and dev->hs_config and dev->dev need to be cleaned if dev_config fails to avoid UAF.

Do they really need to be cleared? I think if dev_config fails then those pointers never get used, so it doesn't matter what they contain.

Of course, clearing them doesn't hurt, but it would be best to clear all of them in the "fail:" part of the routine. And then you would want to change the pathway where usb_gadget_probe_driver fails, to make it go to "fail:" also.

Alan Stern

author
#2 lore raw

Thank for your suggesions. I will break this up into two separate patches later. And i think it's necessary to clear up dev->config and dev->hs_config and dev->dev. For example, dev->hs_config is used in
ep0_read() which may lead a UAF.

Thanks.

Quoted 24 lines · Alan Stern
Quoted 1 line · Hangyu Hua

There are two bugs:

You should break this up into two separate patches, one for each bug.

Quoted 2 lines

dev->buf does not need to be released if it already exists before executing dev_config.

That's right. The call to dev_config should fail without changing any of the stored values.

Quoted 2 lines

dev->config and dev->hs_config and dev->dev need to be cleaned if dev_config fails to avoid UAF.

Do they really need to be cleared? I think if dev_config fails then those pointers never get used, so it doesn't matter what they contain.

Of course, clearing them doesn't hurt, but it would be best to clear all of them in the "fail:" part of the routine. And then you would want to change the pathway where usb_gadget_probe_driver fails, to make it go to "fail:" also.

Alan Stern