[PATCH] ieee802154: mcr20a: read out of bounds in mcr20a_set_channel()

[PATCH] ieee802154: mcr20a: read out of bounds in mcr20a_set_channel()

[Dan Carpenter]

The "channel" variable can be any u8 value.  We need to make sure we
don't read outside of the PLL_INT[] or PLL_FRAC[] arrays.

Fixes: 8c6ad9cc5157 ("ieee802154: Add NXP MCR20A IEEE 802.15.4 transceiver driver")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
---
This patch is obviously harmless, but it's from static analysis.  I'm
pretty sure this is required, but I can't swear.

diff --git a/drivers/net/ieee802154/mcr20a.c b/drivers/net/ieee802154/mcr20a.c
index e428277781ac..4f41d1d3588e 100644
--- a/drivers/net/ieee802154/mcr20a.c
+++ b/drivers/net/ieee802154/mcr20a.c
@@ -512,6 +512,9 @@ mcr20a_set_channel(struct ieee802154_hw *hw, u8 page, u8 channel)
 
 	dev_dbg(printdev(lp), "%s\n", __func__);
 
+	if (channel < 11 || channel - 11 >= ARRAY_SIZE(PLL_INT))
+		return -EINVAL;
+
 	/* freqency = ((PLL_INT+64) + (PLL_FRAC/65536)) * 32 MHz */
 	ret = regmap_write(lp->regmap_dar, DAR_PLL_INT0, PLL_INT[channel - 11]);
 	if (ret)

Re: [PATCH] ieee802154: mcr20a: read out of bounds in mcr20a_set_channel()

[Dan Carpenter]

On Wed, Aug 29, 2018 at 07:50:51PM +0200, Xue Liu wrote:
> Hello Dan,
> 
> 
> On Wed, 29 Aug 2018 at 16:49, Dan Carpenter <dan.carpenter@oracle.com>
> wrote:
> 
> > The "channel" variable can be any u8 value.  We need to make sure we
> > don't read outside of the PLL_INT[] or PLL_FRAC[] arrays.
> >
> I think the “channel” variable can not be any u8 value. This values is
> already checked before set_channel function is called.
> https://github.com/torvalds/linux/blob/master/net/ieee802154/nl802154.c#L978

Oh...  Hm...  I should have reviewed more carefully.  This patch isn't
correct.  But there may still be a bug.  What Smatch is worried about is
the other call tree:

ieee802154_start_req() calls (struct ieee802154_mlme_ops)->start_req
  -> mac802154_mlme_start_req()
     -> mac802154_dev_set_page_channel()
        -> drv_set_channel() calls local->ops->set_channel(&local->hw, page, channel);
           -> mcr20a_set_channel()

So maybe we could move the check from nl802154_set_channel() to
drv_set_channel() so channel is checked on both call trees.

regards,
dan carpenter