Skip to contentSkip to conversation

USB: gadget: fix mm lifetime use-after-free in async read paths

PATCH In review Gabriel Prostitis 2 participants 5 messages Jun 1 – Jul 1 USB

lore
Share & tools
lore
#0 lore raw

v2 cover USB: gadget: fix mm lifetime use-after-free in async read paths

There is a use-after-free in the USB gadget FunctionFS and legacy GadgetFS asynchronous read paths. Both implementations store current->mm in per-request state without taking a reference, and later use it in a workqueue context via kthread_use_mm().

If the submitting task exits before the USB request completes, the stored mm_struct may be freed while a pending request still references it, leading to a use-after-free.

The issue affects:

- FunctionFS: drivers/usb/gadget/function/f_fs.c
- GadgetFS legacy: drivers/usb/gadget/legacy/inode.c

Fix this by taking a reference to mm_struct with mmgrab() when queueing the request and releasing it with mmdrop() after completion. Before using the saved mm_struct, acquire a temporary reference with mmget_not_zero() to ensure it is still alive.

The issue can be triggered by submitting asynchronous reads on OUT endpoints (e.g. via io_uring for FunctionFS) and exiting the submitting task before completion. This may result in memory corruption in the address space of another process if the freed mm_struct is reclaimed during the race window.

KASAN reports confirm use-after-free in the workqueue completion path when accessing the stale mm_struct.

---
Changes in v2:
- Fix swapped lines in drivers/usb/gadget/legacy/inode.c
- Link to v1: https://patch.msgid.link/[email protected]

To: Greg Kroah-Hartman <[email protected]>
Cc: [email protected]
Cc: [email protected]

---
Gabriel Prostitis (2):
USB: gadget: ffs: fix mm lifetime handling
USB: gadget: inode: fix mm lifetime handling

 drivers/usb/gadget/function/f_fs.c | 27 ++++++++++++++++++++-------
 drivers/usb/gadget/legacy/inode.c  | 17 +++++++++++++----
 2 files changed, 33 insertions(+), 11 deletions(-)

---
base-commit: 22d91cef94b5b86cff0d68ebfce7741740672704
change-id: 20260531-mm-uaf-fix-91d9dacac692

Best regards,

Signature · 1 line
Gabriel Prostitis <[email protected]>
#1 lore raw

PATCH v2 1/2 USB: gadget: ffs: fix mm lifetime handling

io_data stores a pointer to the submitting task's mm_struct, but does not currently hold a reference to it while async requests are pending.

This can result in a use-after-free if the task exits before completion handling finishes.

Take a reference with mmgrab() when queuing the read request and release it with mmdrop() on request completion.

---
 drivers/usb/gadget/function/f_fs.c | 27 ++++++++++++++++++++-------
 1 file changed, 20 insertions(+), 7 deletions(-)
drivers/usb/gadget/function/f_fs.c +20 -7
index 002c3441bea3..674f2fd5450f 100644
@@ -864,9 +864,15 @@ static void ffs_user_copy_worker(struct work_struct *work)
 	bool kiocb_has_eventfd = io_data->kiocb->ki_flags & IOCB_EVENTFD;
 
 	if (io_data->read && ret > 0) {
-		kthread_use_mm(io_data->mm);
-		ret = ffs_copy_to_iter(io_data->buf, ret, &io_data->data);
-		kthread_unuse_mm(io_data->mm);
+		if (mmget_not_zero(io_data->mm)) {
+			kthread_use_mm(io_data->mm);
+			ret = ffs_copy_to_iter(io_data->buf, ret, &io_data->data);
+			kthread_unuse_mm(io_data->mm);
+			mmput(io_data->mm);
+		} else {
+			ret = -EFAULT;
+		}
+		mmdrop(io_data->mm);
 	}
 
 	io_data->kiocb->ki_complete(io_data->kiocb, ret);
@@ -1261,16 +1267,20 @@ static ssize_t ffs_epfile_write_iter(struct kiocb *kiocb, struct iov_iter *from)
 
 	kiocb->private = p;
 
-	if (p->aio)
+	if (p->aio) {
+		mmgrab(p->mm);
 		kiocb_set_cancel_fn(kiocb, ffs_aio_cancel);
+	}
 
 	res = ffs_epfile_io(kiocb->ki_filp, p);
 	if (res == -EIOCBQUEUED)
 		return res;
-	if (p->aio)
+	if (p->aio) {
+		mmdrop(p->mm);
 		kfree(p);
-	else
+	} else {
 		*from = p->data;
+	}
 	return res;
 }
 
@@ -1305,14 +1315,17 @@ static ssize_t ffs_epfile_read_iter(struct kiocb *kiocb, struct iov_iter *to)
 
 	kiocb->private = p;
 
-	if (p->aio)
+	if (p->aio) {
+		mmgrab(p->mm);
 		kiocb_set_cancel_fn(kiocb, ffs_aio_cancel);
+	}
 
 	res = ffs_epfile_io(kiocb->ki_filp, p);
 	if (res == -EIOCBQUEUED)
 		return res;
 
 	if (p->aio) {
+		mmdrop(p->mm);
 		kfree(p->to_free);
 		kfree(p);
 	} else {
Signature · 1 line
2.54.0
#2 lore raw

PATCH v2 2/2 USB: gadget: inode: fix mm lifetime handling

In reply to Gabriel Prostitis

priv stores a pointer to the submitting task's mm_struct, but does not currently hold a reference to it while async requests are pending.

This can result in a use-after-free if the task exits before completion handling finishes.

Take a reference with mmgrab() when queuing the read request and release it with mmdrop() on request completion.

---
 drivers/usb/gadget/legacy/inode.c | 17 +++++++++++++----
 1 file changed, 13 insertions(+), 4 deletions(-)
drivers/usb/gadget/legacy/inode.c +13 -4
index d87a8ab51510..db961aaa3740 100644
@@ -471,11 +471,17 @@ static void ep_user_copy_worker(struct work_struct *work)
 	struct kiocb *iocb = priv->iocb;
 	size_t ret;
 
-	kthread_use_mm(mm);
-	ret = copy_to_iter(priv->buf, priv->actual, &priv->to);
-	kthread_unuse_mm(mm);
-	if (!ret)
+	if (mmget_not_zero(mm)) {
+		kthread_use_mm(mm);
+		ret = copy_to_iter(priv->buf, priv->actual, &priv->to);
+		kthread_unuse_mm(mm);
+		mmput(mm);
+		if (!ret)
+			ret = -EFAULT;
+	} else {
 		ret = -EFAULT;
+	}
+	mmdrop(mm);
 
 	/* completing the iocb can drop the ctx and mm, don't touch mm after */
 	iocb->ki_complete(iocb, ret);
@@ -501,6 +507,7 @@ static void ep_aio_complete(struct usb_ep *ep, struct usb_request *req)
 	 * complete the aio request immediately.
 	 */
 	if (priv->to_free == NULL || unlikely(req->actual == 0)) {
+		mmdrop(priv->mm);
 		kfree(req->buf);
 		kfree(priv->to_free);
 		kfree(priv);
@@ -541,6 +548,7 @@ static ssize_t ep_aio(struct kiocb *iocb,
 	priv->epdata = epdata;
 	priv->actual = 0;
 	priv->mm = current->mm; /* mm teardown waits for iocbs in exit_aio() */
+	mmgrab(priv->mm);
 
 	/* each kiocb is coupled to one usb_request, but we can't
 	 * allocate or submit those if the host disconnected.
@@ -570,6 +578,7 @@ static ssize_t ep_aio(struct kiocb *iocb,
 
 fail:
 	spin_unlock_irq(&epdata->dev->lock);
+	mmdrop(priv->mm);
 	kfree(priv->to_free);
 	kfree(priv);
 	put_ep(epdata);
Signature · 1 line
2.54.0
maintainer
#4 lore raw
Quoted 3 lines · Gabriel Prostitis

Hi,

Just a gentle ping on this series.

I have no context here :(

Quoted 2 lines

I noticed that it has not shown up in -next yet, so I wanted to check whether there is anything else I should address.

The merge window happened which kept things from being able to be merged for 2 weeks. Now I have 3000+ patches to dig through and review. Please help out by reviewing other patches sent to the list so that yours can move up faster.

thanks,

greg k-h