linux-kernel - Re: [PATCH] platform/chrome: cros_ec: fix read overflow in cros_ec_lpc

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-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 15 Dec 2021 11:19:35 +0300
From:   Dan Carpenter <dan.carpenter@...cle.com>
To:     Guenter Roeck <groeck@...gle.com>
Cc:     Benson Leung <bleung@...omium.org>,
        Bill Richardson <wfrichar@...omium.org>,
        Guenter Roeck <groeck@...omium.org>,
        Javier Martinez Canillas <javier@....samsung.com>,
        Olof Johansson <olof@...om.net>,
        Gwendal Grignou <gwendal@...omium.org>,
        linux-kernel@...r.kernel.org, kernel-janitors@...r.kernel.org
Subject: Re: [PATCH] platform/chrome: cros_ec: fix read overflow in
 cros_ec_lpc_readmem()

On Tue, Dec 14, 2021 at 03:02:41PM -0800, Guenter Roeck wrote:
>  On Thu, Dec 9, 2021 at 6:35 AM Dan Carpenter <dan.carpenter@...cle.com> wrote:
> >
> > If bytes is larger than EC_MEMMAP_SIZE (255) then "EC_MEMMAP_SIZE -
> > bytes" is a very high unsigned value and basically offset is
> > accepted.  The second problem is that it uses >= instead of > so this
> > means that we are not able to read the very last byte.
> >
> > Fixes: ec2f33ab582b ("platform/chrome: Add cros_ec_lpc driver for x86 devices")
> > Signed-off-by: Dan Carpenter <dan.carpenter@...cle.com>
> > ---
> >  drivers/platform/chrome/cros_ec_lpc.c | 3 ++-
> >  1 file changed, 2 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/platform/chrome/cros_ec_lpc.c b/drivers/platform/chrome/cros_ec_lpc.c
> > index d6306d2a096f..7e1d175def9f 100644
> > --- a/drivers/platform/chrome/cros_ec_lpc.c
> > +++ b/drivers/platform/chrome/cros_ec_lpc.c
> > @@ -290,7 +290,8 @@ static int cros_ec_lpc_readmem(struct cros_ec_device *ec, unsigned int offset,
> >         char *s = dest;
> >         int cnt = 0;
> >
> > -       if (offset >= EC_MEMMAP_SIZE - bytes)
> > +       if (offset > EC_MEMMAP_SIZE ||
> > +           bytes > EC_MEMMAP_SIZE - offset)
> 
> I think that means we have the same problem if offset >
> EC_MEMMAP_SIZE, only now that condition isn't detected anymore because
> EC_MEMMAP_SIZE - offset is a very large number.

That's the bug which my patch addresses.  (My patch is option 1).

> I think what we really want is
>         if (offset + bytes > EC_MEMMAP_SIZE)
> only without the overflow. Not sure how we can get there without
> checking each part.
>         if (offset > EC_MEMMAP_SIZE || bytes > EC_MEMMAP_SIZE || bytes
> + offset > EC_MEMMAP_SIZE)

That is another solution which works.

>                 return -EINVAL;
> Maybe that ?
>         if ((u64) offset + bytes > EC_MEMMAP_SIZE)
>                 return -EINVAL;

A third viable solution.

I generally prefer option 2 to option 3.  I generally use that in code
that I'm writing.  There was one time Linus said he liked option 1
which I used here because it works regardless of the types or the valu
of EC_MEMMAP_SIZE.  This code already used the bytes > size - offset
idiom so I kept it as similar as possible.

regards,
dan carpenter