Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Sun, 20 Jun 2021 22:25:08 -0700
From:   Dmitry Torokhov <dmitry.torokhov@gmail.com>
To:     Linus Torvalds <torvalds@linux-foundation.org>
Cc:     Alexander Larkin <avlarkin82@gmail.com>,
        Dan Carpenter <dan.carpenter@oracle.com>,
        linux-input@vger.kernel.org, lkml <linux-kernel@vger.kernel.org>,
        Security Officers <security@kernel.org>,
        Murray McAllister <murray.mcallister@gmail.com>
Subject: Re: [PATCH] Input: joydev - prevent potential write out of bounds in
 ioctl
Message-ID: <YNAitJfOpoBkFitU@google.com>
References: <20210620120030.1513655-1-avlarkin82@gmail.com>
 <CAADWXX-fpcPh+jGX7=Hbkqr7yhwzbUT915NBBzqHGecFVbxmzw@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAADWXX-fpcPh+jGX7=Hbkqr7yhwzbUT915NBBzqHGecFVbxmzw@mail.gmail.com>
Precedence: bulk

On Sun, Jun 20, 2021 at 09:37:47AM -0700, Linus Torvalds wrote:
> On Sun, Jun 20, 2021 at 5:01 AM Alexander Larkin <avlarkin82@gmail.com> wrote:
> >
> >     The problem is that the check of user input values that is just
> >     before the fixed line of code is for the part of first values
> >     (before len or before len/2), but then the usage of all the values
> >     including i >= len (or i >= len/2) could be.
> 
> No, I think the problem is simpler than that.
> 
> > -       for (i = 0; i < joydev->nabs; i++)
> > +       for (i = 0; i < len && i < joydev->nabs; i++)
> >                 joydev->absmap[joydev->abspam[i]] = i;
> 
> This part is unnecessary - all values of "joydev->abspam[i]" have been
> validated (either they are the old ones, or the new ones that we just
> validated).
> 
> >         memcpy(joydev->keypam, keypam, len);
> >
> > -       for (i = 0; i < joydev->nkey; i++)
> > +       for (i = 0; i < (len / 2) && i < joydev->nkey; i++)
> >                 joydev->keymap[keypam[i] - BTN_MISC] = i;
> 
> The problem here is not that we walk past "len/2", but that the code
> *should* have used
> 
>         joydev->keymap[joydev->keypam[i] - BTN_MISC] = i;
> 
> (note the "keypam[1]" vs "joydev->keypam[i]").
> 
> And the reason it *should* walk the whole "joydev->nkey" is that if
> there are later cases with the same keypam value, the later ones
> should override the previous ones (well, that "should" is more a
> "traditionally have").

Yes, we can discuss whether "short" ioctl should clear out the part of
map that is not supplied by the call, but given that I consider joydev
legacy my preference would be to leave this as it was.

> 
> So I think the right patch is this one-liner
> 
>   diff --git a/drivers/input/joydev.c b/drivers/input/joydev.c
>   index da8963a9f044..947d440a3be6 100644
>   --- a/drivers/input/joydev.c
>   +++ b/drivers/input/joydev.c
>   @@ -499,7 +499,7 @@ static int joydev_handle_JSIOCSBTNMAP(struct
> joydev *joydev,
>         memcpy(joydev->keypam, keypam, len);
> 
>         for (i = 0; i < joydev->nkey; i++)
>   -             joydev->keymap[keypam[i] - BTN_MISC] = i;
>   +             joydev->keymap[joydev->keypam[i] - BTN_MISC] = i;
> 
>     out:
>         kfree(keypam);
> 
> (whitespace-damaged, I would like Dmitry to think about it rather than
> apply this mindlessly.
> 
> Dmitry?

Yes, this makes sense to me and it is safe as joydev->keypam is
guaranteed to be the right size.

Are you going to reformat this and resend or should I?


Thanks.

-- 
Dmitry