linux-kernel - Re: [PATCH v2] usb:host: fix divide-by-zero in function fhci_queue

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-ID: <2e3a7b4df8744b508b866db40bb8cf28@huawei.com>
Date:   Thu, 18 Apr 2019 09:31:58 +0000
From:   "zhuyan (M)" <zhuyan34@...wei.com>
To:     Alan Stern <stern@...land.harvard.edu>
CC:     Greg KH <gregkh@...uxfoundation.org>,
        "anton@...msg.org" <anton@...msg.org>,
        "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v2] usb:host: fix divide-by-zero in function
 fhci_queue_urb

On Wed, 17 Apr 2019 14:59:27 -0400, Alan Stern wrote:
> 
> On Wed, 17 Apr 2019, zhuyan (M) wrote:
> 
> > On Wed, 17 Apr 2019, Alan Stern wrote:
> > 
> > > On Wed, 17 Apr 2019, zhuyan (M) wrote:
> > > 
> > > > On Tue, 16 Apr 2019 11:07:56 -0400, Alan Stern wrote:
> > > > 
> > > > > On Tue, 16 Apr 2019, zhuyan (M) wrote:
> > > > > > On Tue, 16 Apr 2019 at 11:45:45 +0200, Greg KH wrote:
> > > > > > > On Tue, Apr 09, 2019 at 10:37:12PM +0800, zhuyan wrote:
> > > > > > > > In function fhci_queue_urb, the divisor of expression 
> > > > > > > > (urb->transfer_buffer_length % usb_maxpacket(urb->dev,
> > > > > > > > urb->pipe,
> > > > > > > > usb_pipeout(urb->pipe))) may be zero.
> > > > > > > 
> > > > > > > How can you hit that?
> > > > > > > 
> > > > > > > > When it is zero, unexpected results may occur, so it is 
> > > > > > > > necessary to ensure that the divisor is not zero.
> > > > > > > > 
> > > > > > > > Signed-off-by: zhuyan <zhuyan34@...wei.com>
> > > > > > > 
> > > > > > > I need a "Full" name here, not just a single name.  Whatever you use to sign documents is good.
> > > > > > > 
> > > > > > > thanks,
> > > > > > > 
> > > > > > > greg k-h
> > > > > > 
> > > > > > In function usb_maxpacket, when ep is NULL, its return value is 0.  
> > > > > 
> > > > > fhci_queue_urb() shouldn't use urb->pipe to compute the 
> > > > > maxpacket size anyway.  It should use usb_endpoint_maxp(&urb->ep->desc).
> > > > 
> > > > Currently, fhci_queue_urb(), call usb_maxpacket() multiple times 
> > > > to calculate  the maxpacket size. The usb_maxpacket() will call
> > > > usb_endpoint_maxp() to compute the maxpacket size.
> > > 
> > > I know that.  What fhci_queue_urb() is doing is wrong.  You should change it: 
> > > Make it call usb_endpoint_maxp directly instead of calling usb_maxpacket.
> > > 
> > 
> > From 1996456d0cc17b5ff7746a598ff355b25d13db3e Mon Sep 17 00:00:00 2001
> > From: zhuyan <zhuyan34@...wei.com>
> > Date: Thu, 18 Apr 2019 00:53:03 +0800
> > Subject: [PATCH] usb: host: fix divide-by-zero in function 
> > fhci_queue_urb
> > 
> > fhci_queue_urb() shouldn't use urb->pipe to compute the maxpacket size 
> > anyway.It should use usb_endpoint_maxp(&urb->ep->desc).
> > 
> > In function fhci_queue_urb, the divisor of expression 
> > (urb->transfer_buffer_length % usb_maxpacket(urb->dev, urb->pipe,
> > usb_pipeout(urb->pipe))) may be zero. When it is zero, unexpected 
> > results may occur, so it is necessary to ensure that the divisor is not zero.
> > 
> > Signed-off-by: zhuyan <zhuyan34@...wei.com>
> > ---
> >  drivers/usb/host/fhci-sched.c | 13 +++++++------
> >  1 file changed, 7 insertions(+), 6 deletions(-)
> > 
> > diff --git a/drivers/usb/host/fhci-sched.c 
> > b/drivers/usb/host/fhci-sched.c index 3d12cdd..7dcfe22 100644
> > --- a/drivers/usb/host/fhci-sched.c
> > +++ b/drivers/usb/host/fhci-sched.c
> > @@ -704,6 +704,7 @@ void fhci_queue_urb(struct fhci_hcd *fhci, struct urb *urb)
> >  	struct td *td;
> >  	u8 *data;
> >  	u16 cnt = 0;
> > +	u16 max_pkt_size = 0;
> >  
> >  	if (ed == NULL) {
> >  		ed = fhci_get_empty_ed(fhci);
> > @@ -727,8 +728,7 @@ void fhci_queue_urb(struct fhci_hcd *fhci, struct urb *urb)
> >  		}
> >  		ed->speed = (urb->dev->speed == USB_SPEED_LOW) ?
> >  			FHCI_LOW_SPEED : FHCI_FULL_SPEED;
> > -		ed->max_pkt_size = usb_maxpacket(urb->dev,
> > -			urb->pipe, usb_pipeout(urb->pipe));
> > +		ed->max_pkt_size = usb_endpoint_maxp(&urb->ep->desc);
> >  		urb->ep->hcpriv = ed;
> >  		fhci_dbg(fhci, "new ep speed=%d max_pkt_size=%d\n",
> >  			 ed->speed, ed->max_pkt_size);
> > @@ -765,11 +765,12 @@ void fhci_queue_urb(struct fhci_hcd *fhci, 
> > struct urb *urb)
> >  
> >  	switch (ed->mode) {
> >  	case FHCI_TF_BULK:
> > +		max_pkt_size = usb_endpoint_maxp(&urb->ep->desc);
> >  		if (urb->transfer_flags & URB_ZERO_PACKET &&
> >  				urb->transfer_buffer_length > 0 &&
> > +				(max_pkt_size != 0) &&
> 
> Now you shouldn't need to add this extra test.
> 
> Alan Stern
> 
> >  				((urb->transfer_buffer_length %
> > -				usb_maxpacket(urb->dev, urb->pipe,
> > -				usb_pipeout(urb->pipe))) == 0))
> > +				max_pkt_size) == 0))
> >  			urb_state = US_BULK0;
> >  		while (data_len > 4096) {
> >  			td = fhci_td_fill(fhci, urb, urb_priv, ed, cnt, @@ -807,8 +808,8 
> > @@ void fhci_queue_urb(struct fhci_hcd *fhci, struct urb *urb)
> >  		break;
> >  	case FHCI_TF_CTRL:
> >  		ed->dev_addr = usb_pipedevice(urb->pipe);
> > -		ed->max_pkt_size = usb_maxpacket(urb->dev, urb->pipe,
> > -			usb_pipeout(urb->pipe));
> > +		ed->max_pkt_size = usb_endpoint_maxp(&urb->ep->desc);
> > +
> >  		/* setup stage */
> >  		td = fhci_td_fill(fhci, urb, urb_priv, ed, cnt++, FHCI_TA_SETUP,
> >  			USB_TD_TOGGLE_DATA0, urb->setup_packet, 8, 0, 0, true);

But the return value of usb_endpoint_maxp() may be 0. 
Do we need to take protective measures?

Yan Zhu