Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Sun, 20 Jan 2019 16:20:00 -0500
From:   Scott Bauer <sbauer@plzdonthack.me>
To:     David Kozub <zub@linux.fjfi.cvut.cz>
Cc:     linux-kernel@vger.kernel.org, axboe@kernel.dk, hch@infradead.org,
        jonathan.derrick@intel.com
Message-ID: <20190120212000.GA4387@hacktheplanet>
References: <1547760716-7304-12-git-send-email-zub@linux.fjfi.cvut.cz>
 <20190119171550.GB12171@hacktheplanet>
 <alpine.LRH.2.21.1901201045280.4278@linux.fjfi.cvut.cz>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <alpine.LRH.2.21.1901201045280.4278@linux.fjfi.cvut.cz>
User-Agent: Mutt/1.9.4 (2018-02-28)
Subject: Re: [PATCH v2 11/16] block: sed-opal: ioctl for writing to shadow mbr
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Sun, Jan 20, 2019 at 11:27:30AM +0100, David Kozub wrote:
> On Sat, 19 Jan 2019, Scott Bauer wrote:
> 
> > On Thu, Jan 17, 2019 at 09:31:51PM +0000, David Kozub wrote:
> > 
> > > +static int write_shadow_mbr(struct opal_dev *dev, void *data)
> > > +{
> > > +	struct opal_shadow_mbr *shadow = data;
> > > +	const u8 __user *src;
> > > +	u8 *dst;
> > > +	size_t off = 0;
> > > +	u64 len;
> > > +	int err = 0;
> > > +
> > > +	/* do the actual transmission(s) */
> > > +	src = (u8 *) shadow->data;
> > > +	while (off < shadow->size) {
> > > +		err = cmd_start(dev, opaluid[OPAL_MBR], opalmethod[OPAL_SET]);
> > > +		add_token_u8(&err, dev, OPAL_STARTNAME);
> > > +		add_token_u8(&err, dev, OPAL_WHERE);
> > > +		add_token_u64(&err, dev, shadow->offset + off);
> > > +		add_token_u8(&err, dev, OPAL_ENDNAME);
> > > +
> > > +		add_token_u8(&err, dev, OPAL_STARTNAME);
> > > +		add_token_u8(&err, dev, OPAL_VALUES);
> > > +
> > > +		/*
> > > +		 * The bytestring header is either 1 or 2 bytes, so assume 2.
> > > +		 * There also needs to be enough space to accommodate the
> > > +		 * trailing OPAL_ENDNAME (1 byte) and tokens added by
> > > +		 * cmd_finalize.
> > > +		 */
> > > +		len = min(remaining_size(dev) - (2+1+CMD_FINALIZE_BYTES_NEEDED),
> > > +			  (size_t)(shadow->size - off));
> > 
> > What if remaining_size(dev) <  2 + 1 + CMD_FINALIZE_BYTES_NEEDED? If that's possible we
> > get min(UINT_MAX(ish) , some size larger than our remaining buffer) and that's not good.
> 
> This is only possible for uselessly small values of IO_BUFFER_LENGTH, which
> is a compile-time value. Originally I thought it's OK as nobody would set
> the value so low. But on a second thought, after reading your comment, I
> think that even if IO_BUFFER_LENGTH was set to such a value, the code should
> fail gracefully.

Naw, just leave it the way it is. I didnt follow cmd_start down deep enough to realize we reset
the position on cmd_start, so there is no way for my scenario to happen. We'll never change IO_BUFFER_LENGTH
to something small. If someone wants to write a bug in their custom kernel, it's not our problem.