[v3,2/2] usb: gadget: f_fs: change ep->ep safe in ffs_epfile_io()

In ffs_epfile_io(), when read/write data in blocking mode, it will wait
the completion in interruptible mode, if task receive a signal, it will
terminate the wait, at same time, if function unbind occurs,
ffs_func_unbind() will kfree all eps, ffs_epfile_io() still try to
dequeue request by dereferencing ep which may become invalid.

Fix it by add ep spinlock and will not dereference ep if it is not valid.

Reported-by: Michael Wu <michael@allwinnertech.com>
Reviewed-by: John Keeping <john@metanate.com>
Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com>
---
v2: add Reviewed-by from John keeping
v3: add Reported-by from Michael Wu

 drivers/usb/gadget/function/f_fs.c | 6 ++++++
 1 file changed, 6 insertions(+)

On 6/1/2022 12:15 PM, Linyu Yuan wrote:
> In ffs_epfile_io(), when read/write data in blocking mode, it will wait
> the completion in interruptible mode, if task receive a signal, it will
> terminate the wait, at same time, if function unbind occurs,
> ffs_func_unbind() will kfree all eps, ffs_epfile_io() still try to
> dequeue request by dereferencing ep which may become invalid.
> 
> Fix it by add ep spinlock and will not dereference ep if it is not valid.
> 
> Reported-by: Michael Wu <michael@allwinnertech.com>
> Reviewed-by: John Keeping <john@metanate.com>
> Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com>
> ---
> v2: add Reviewed-by from John keeping
> v3: add Reported-by from Michael Wu
> 
>   drivers/usb/gadget/function/f_fs.c | 6 ++++++
>   1 file changed, 6 insertions(+)
> 
> diff --git a/drivers/usb/gadget/function/f_fs.c b/drivers/usb/gadget/function/f_fs.c
> index d4d8940..9bf9287 100644
> --- a/drivers/usb/gadget/function/f_fs.c
> +++ b/drivers/usb/gadget/function/f_fs.c
> @@ -1077,6 +1077,11 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
>   		spin_unlock_irq(&epfile->ffs->eps_lock);
>   
>   		if (wait_for_completion_interruptible(&io_data->done)) {
> +			spin_lock_irq(&epfile->ffs->eps_lock);
> +			if (epfile->ep != ep) {
> +				ret = -ESHUTDOWN;
> +				goto error_lock;
> +			}
>   			/*
>   			 * To avoid race condition with ffs_epfile_io_complete,
>   			 * dequeue the request first then check
> @@ -1084,6 +1089,7 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
>   			 * condition with req->complete callback.
>   			 */
>   			usb_ep_dequeue(ep->ep, req);
> +			spin_unlock_irq(&epfile->ffs->eps_lock);
>   			wait_for_completion(&io_data->done);
>   			interrupted = io_data->status < 0;
>   		}

Tested-by: Michael Wu <michael@allwinnertech.com>

I've tested Linyu's patches [PATCH v3 1/2] [PATCH v3 2/2]. I believe it 
fixes the bug I reported.

What's more, John's solution [1] also works in my tests. It looks 
simpler. I'm not sure if it's as complete as Linyu's solution.


[1] https://lore.kernel.org/linux-usb/YpUJkxWBNuZiW7Xk@donbot/

On Thu, Jun 02, 2022 at 06:39:30PM +0800, Michael Wu wrote:
> On 6/1/2022 12:15 PM, Linyu Yuan wrote:
> > In ffs_epfile_io(), when read/write data in blocking mode, it will wait
> > the completion in interruptible mode, if task receive a signal, it will
> > terminate the wait, at same time, if function unbind occurs,
> > ffs_func_unbind() will kfree all eps, ffs_epfile_io() still try to
> > dequeue request by dereferencing ep which may become invalid.
> > 
> > Fix it by add ep spinlock and will not dereference ep if it is not valid.
> > 
> > Reported-by: Michael Wu <michael@allwinnertech.com>
> > Reviewed-by: John Keeping <john@metanate.com>
> > Signed-off-by: Linyu Yuan <quic_linyyuan@quicinc.com>
> > ---
> > v2: add Reviewed-by from John keeping
> > v3: add Reported-by from Michael Wu
> > 
> >   drivers/usb/gadget/function/f_fs.c | 6 ++++++
> >   1 file changed, 6 insertions(+)
> > 
> > diff --git a/drivers/usb/gadget/function/f_fs.c b/drivers/usb/gadget/function/f_fs.c
> > index d4d8940..9bf9287 100644
> > --- a/drivers/usb/gadget/function/f_fs.c
> > +++ b/drivers/usb/gadget/function/f_fs.c
> > @@ -1077,6 +1077,11 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
> >   		spin_unlock_irq(&epfile->ffs->eps_lock);
> >   		if (wait_for_completion_interruptible(&io_data->done)) {
> > +			spin_lock_irq(&epfile->ffs->eps_lock);
> > +			if (epfile->ep != ep) {
> > +				ret = -ESHUTDOWN;
> > +				goto error_lock;
> > +			}
> >   			/*
> >   			 * To avoid race condition with ffs_epfile_io_complete,
> >   			 * dequeue the request first then check
> > @@ -1084,6 +1089,7 @@ static ssize_t ffs_epfile_io(struct file *file, struct ffs_io_data *io_data)
> >   			 * condition with req->complete callback.
> >   			 */
> >   			usb_ep_dequeue(ep->ep, req);
> > +			spin_unlock_irq(&epfile->ffs->eps_lock);
> >   			wait_for_completion(&io_data->done);
> >   			interrupted = io_data->status < 0;
> >   		}
> 
> Tested-by: Michael Wu <michael@allwinnertech.com>
> 
> I've tested Linyu's patches [PATCH v3 1/2] [PATCH v3 2/2]. I believe it
> fixes the bug I reported.
> 
> What's more, John's solution [1] also works in my tests. It looks simpler.
> I'm not sure if it's as complete as Linyu's solution.

It's not as comprehensive, let's focus on this thread.

> [1] https://lore.kernel.org/linux-usb/YpUJkxWBNuZiW7Xk@donbot/