Re: [PATCH] can: kvaser_usb: Don't free packets when tight on URBs

From: "Ahmed S. Darwish" <darwish.07@gmail.com>
To: Olivier Sobrie <olivier@sobrie.be>
Cc: Oliver Hartkopp <socketcan@hartkopp.net>,
	Wolfgang Grandegger <wg@grandegger.com>,
	Marc Kleine-Budde <mkl@pengutronix.de>,
	"David S. Miller" <davem@davemloft.net>,
	Paul Gortmaker <paul.gortmaker@windriver.com>,
	Linux-CAN <linux-can@vger.kernel.org>,
	netdev <netdev@vger.kernel.org>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] can: kvaser_usb: Don't free packets when tight on URBs
Date: Wed, 24 Dec 2014 17:52:46 +0200	[thread overview]
Message-ID: <20141224155246.GB27062@vivalin-002> (raw)
In-Reply-To: <CAEM=fMqwt+=1Du74qu3dLJ7-OLnypkhJZ_N8uSWAAtzTLazmOg@mail.gmail.com>

Hi Olivier,

On Wed, Dec 24, 2014 at 01:31:20PM +0100, Olivier Sobrie wrote:
> Hello Ahmed,
> 
> On Tue, Dec 23, 2014 at 05:46:54PM +0200, Ahmed S. Darwish wrote:
> > From: Ahmed S. Darwish <ahmed.darwish@valeo.com>
> > 
> > Flooding the Kvaser CAN to USB dongle with multiple reads and
> > writes in high frequency caused seemingly-random panics in the
> > kernel.
> > 
> > On further inspection, it seems the driver erroneously freed the
> > to-be-transmitted packet upon getting tight on URBs and returning
> > NETDEV_TX_BUSY, leading to invalid memory writes and double frees
> > at a later point in time.
> > 
> > Signed-off-by: Ahmed S. Darwish <ahmed.darwish@valeo.com>
> 
> Thank you for the fix.
> 
> > ---
> >  drivers/net/can/usb/kvaser_usb.c | 8 +++++---
> >  1 file changed, 5 insertions(+), 3 deletions(-)
> > 
> >  (Generated over 3.19.0-rc1)
> > 
> > diff --git a/drivers/net/can/usb/kvaser_usb.c b/drivers/net/can/usb/kvaser_usb.c
> > index 541fb7a..34c35d8 100644
> > --- a/drivers/net/can/usb/kvaser_usb.c
> > +++ b/drivers/net/can/usb/kvaser_usb.c
> > @@ -1286,6 +1286,7 @@ static netdev_tx_t kvaser_usb_start_xmit(struct sk_buff *skb,
> >  	struct kvaser_msg *msg;
> >  	int i, err;
> >  	int ret = NETDEV_TX_OK;
> > +	bool kfree_skb_on_error = true;
> >  
> >  	if (can_dropped_invalid_skb(netdev, skb))
> >  		return NETDEV_TX_OK;
> > @@ -1336,6 +1337,7 @@ static netdev_tx_t kvaser_usb_start_xmit(struct sk_buff *skb,
> >  
> >  	if (!context) {
> >  		netdev_warn(netdev, "cannot find free context\n");
> > +		kfree_skb_on_error = false;
> 
> Instead of using an extra variable, you can also set skb to NULL here.
> Or maybe better, you can move the dev_kfree_skb() in the two previous
> tests (in the check of variables urb and buf).
> 

Nice, I'll move dev_kfree_skb() to the two earlier tests then.

Thanks,

P.S. mailer and patch identation; had to manually fix them
before replying (but thanks for the quick review, ofc ;-))

> Thank you,
> 
> Olivier
> 
> >  		ret =  NETDEV_TX_BUSY;
> >  		goto releasebuf;
> >  	}
> > @@ -1364,8 +1366,7 @@ static netdev_tx_t kvaser_usb_start_xmit(struct sk_buff *skb,
> >  	if (unlikely(err)) {
> >  		can_free_echo_skb(netdev, context->echo_index);
> >  
> > -		skb = NULL; /* set to NULL to avoid double free in
> > -			     * dev_kfree_skb(skb) */
> > +		kfree_skb_on_error = false;
> >  
> >  		atomic_dec(&priv->active_tx_urbs);
> >  		usb_unanchor_urb(urb);
> > @@ -1389,7 +1390,8 @@ releasebuf:
> >  nobufmem:
> >  	usb_free_urb(urb);
> >  nourbmem:
> > -	dev_kfree_skb(skb);
> > +	if (kfree_skb_on_error)
> > +		dev_kfree_skb(skb);
> >  	return ret;
> >  }
> > 