Re: [PATCH] Prevent out-of-bounds read/write in bcmasp_netfilt_rd and bcmasp_netfilt_wr

From: Greg KH <gregkh@linuxfoundation.org>
To: Justin Chen <justin.chen@broadcom.com>
Cc: pabeni@redhat.com, florian.fainelli@broadcom.com,
	netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
	edumazet@google.com, bcm-kernel-feedback-list@broadcom.com,
	kuba@kernel.org, Yuran Pereira <yuran.pereira@hotmail.com>,
	linux-kernel-mentees@lists.linuxfoundation.org,
	davem@davemloft.net
Subject: Re: [PATCH] Prevent out-of-bounds read/write in bcmasp_netfilt_rd and bcmasp_netfilt_wr
Date: Fri, 3 Nov 2023 19:33:34 +0100	[thread overview]
Message-ID: <2023110318-utensil-figure-eb80@gregkh> (raw)
In-Reply-To: <e3ac58b2-bb78-4364-94c0-f18c376ac132@broadcom.com>

On Fri, Nov 03, 2023 at 11:23:16AM -0700, Justin Chen wrote:
> 
> 
> On 11/3/23 5:57 AM, Greg KH wrote:
> > On Fri, Nov 03, 2023 at 05:57:48PM +0530, Yuran Pereira wrote:
> > > The functions `bcmasp_netfilt_rd` and `bcmasp_netfilt_wr` both call
> > > `bcmasp_netfilt_get_reg_offset` which, when it fails, returns `-EINVAL`.
> > > This could lead to an out-of-bounds read or write when `rx_filter_core_rl`
> > > or `rx_filter_core_wl` is called.
> > > 
> > > This patch adds a check in both functions to return immediately if
> > > `bcmasp_netfilt_get_reg_offset` fails. This prevents potential out-of-bounds read
> > > or writes, and ensures that no undefined or buggy behavior would originate from
> > > the failure of `bcmasp_netfilt_get_reg_offset`.
> > > 
> > > Addresses-Coverity-IDs: 1544536 ("Out-of-bounds access")
> > > Signed-off-by: Yuran Pereira <yuran.pereira@hotmail.com>
> > > ---
> > >   drivers/net/ethernet/broadcom/asp2/bcmasp.c | 4 ++++
> > >   1 file changed, 4 insertions(+)
> > > 
> > > diff --git a/drivers/net/ethernet/broadcom/asp2/bcmasp.c b/drivers/net/ethernet/broadcom/asp2/bcmasp.c
> > > index 29b04a274d07..8b90b761bdec 100644
> > > --- a/drivers/net/ethernet/broadcom/asp2/bcmasp.c
> > > +++ b/drivers/net/ethernet/broadcom/asp2/bcmasp.c
> > > @@ -227,6 +227,8 @@ static void bcmasp_netfilt_wr(struct bcmasp_priv *priv,
> > >   	reg_offset = bcmasp_netfilt_get_reg_offset(priv, nfilt, reg_type,
> > >   						   offset);
> > > +	if (reg_offset < 0)
> > > +		return;
> > >   	rx_filter_core_wl(priv, val, reg_offset);
> > >   }
> > > @@ -244,6 +246,8 @@ static u32 bcmasp_netfilt_rd(struct bcmasp_priv *priv,
> > >   	reg_offset = bcmasp_netfilt_get_reg_offset(priv, nfilt, reg_type,
> > >   						   offset);
> > > +	if (reg_offset < 0)
> > > +		return 0;
> > 
> > Shouldn't you return an error here?
> > 
> > thanks
> > 
> > greg k-h
> 
> As long as offset is less than MAX_WAKE_FILTER_SIZE we don't need to worry
> about error checking. This is already checked before we call
> netfilt_get_reg_offset() in both cases. Instead of returning -EINVAL in
> neffilt_get_reg_offset() lets return 0. This will silence the coverity
> check. In practice we will never hit this logic.

Then don't change it, coverity is incorrect here.

greg k-h
_______________________________________________
Linux-kernel-mentees mailing list
Linux-kernel-mentees@lists.linuxfoundation.org
https://lists.linuxfoundation.org/mailman/listinfo/linux-kernel-mentees