Re: Re: [PATCH] usb: cdnsp: Fix potential dereference of NULL pointer

[刘永志]

> -----Original Messages-----
> From: "Greg KH" <gregkh@xxxxxxxxxxxxxxxxxxx>
> Sent Time: 2022-05-19 23:58:35 (Thursday)
> To: "Yongzhi Liu" <lyz_cs@xxxxxxxxxx>
> Cc: pawell@xxxxxxxxxxx, peter.chen@xxxxxxx, linux-usb@xxxxxxxxxxxxxxx, linux-kernel@xxxxxxxxxxxxxxx, fuyq@xxxxxxxxxxxxxx
> Subject: Re: [PATCH] usb: cdnsp:  Fix potential dereference of NULL pointer
> 
> On Wed, May 18, 2022 at 09:35:43AM -0700, Yongzhi Liu wrote:
> > The return value of cdnsp_get_transfer_ring()
> > needs to be checked to avoid use of NULL pointer
> > in case of an acquisition failure.
> 
> Please use the full 72 columns
> 
> > 
> > Fixes: 3d8290455 ("usb: cdnsp: cdns3 Add main part of Cadence USBSSP DRD Driver")
> > 
> > Signed-off-by: Yongzhi Liu <lyz_cs@xxxxxxxxxx>
> 
> Please do not put a blank line between "Fixes:" and your signed off by
> line.
> 
> > ---
> >  drivers/usb/cdns3/cdnsp-ring.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/drivers/usb/cdns3/cdnsp-ring.c b/drivers/usb/cdns3/cdnsp-ring.c
> > index 1b14384..9f206b9 100644
> > --- a/drivers/usb/cdns3/cdnsp-ring.c
> > +++ b/drivers/usb/cdns3/cdnsp-ring.c
> > @@ -655,6 +655,8 @@ static int cdnsp_cmd_set_deq(struct cdnsp_device *pdev,
> >  	 * to reflect the new position.
> >  	 */
> >  	ep_ring = cdnsp_get_transfer_ring(pdev, pep, deq_state->stream_id);
> > +	if (!ep_ring)
> > +		return -EINVAL;
> 
> How did you test this?
> 
> Don't you need to properly clean up and handle the hardware issues if
> you exit early here?

I find this by a static analyzer based on frequency and similarity, which report many null ptr deref bugs.
In cdns3/cdnsp-ring.c, I find that we usually check the return value when call function 'usb_get_intfdata'.
If 'deq_state->stream_id' is special, the 'ep_ring' is also null. Therefore, i think we should add null checks here.
I will resubmit a new patch if you think the bug is real.
Thanks for your reply and advice.

> 
> Without good testing, I am loath to take this.
> 
> thanks,
> 
> greg k-h