Re: [PATCH 03/10] b43: Remove uninitialized_var() usage

[Kees Cook]

On Thu, Jun 04, 2020 at 01:08:44PM -0700, Nick Desaulniers wrote:
> On Wed, Jun 3, 2020 at 4:32 PM Kees Cook <keescook@xxxxxxxxxxxx> wrote:
> >
> > Using uninitialized_var() is dangerous as it papers over real bugs[1]
> > (or can in the future), and suppresses unrelated compiler warnings (e.g.
> > "unused variable"). If the compiler thinks it is uninitialized, either
> > simply initialize the variable or make compiler changes. As a precursor
> > to removing[2] this[3] macro[4], just initialize this variable to NULL,
> > and make the (unreachable!) code do a conditional test.
> >
> > [1] https://lore.kernel.org/lkml/20200603174714.192027-1-glider@xxxxxxxxxx/
> > [2] https://lore.kernel.org/lkml/CA+55aFw+Vbj0i=1TGqCR5vQkCzWJ0QxK6CernOU6eedsudAixw@xxxxxxxxxxxxxx/
> > [3] https://lore.kernel.org/lkml/CA+55aFwgbgqhbp1fkxvRKEpzyR5J8n1vKT1VZdz9knmPuXhOeg@xxxxxxxxxxxxxx/
> > [4] https://lore.kernel.org/lkml/CA+55aFz2500WfbKXAx8s67wrm9=yVJu65TpLgN_ybYNv0VEOKA@xxxxxxxxxxxxxx/
> >
> > Signed-off-by: Kees Cook <keescook@xxxxxxxxxxxx>
> > ---
> >  drivers/net/wireless/broadcom/b43/phy_n.c | 10 +++++++---
> >  1 file changed, 7 insertions(+), 3 deletions(-)
> >
> > diff --git a/drivers/net/wireless/broadcom/b43/phy_n.c b/drivers/net/wireless/broadcom/b43/phy_n.c
> > index d3c001fa8eb4..88cdcea10d61 100644
> > --- a/drivers/net/wireless/broadcom/b43/phy_n.c
> > +++ b/drivers/net/wireless/broadcom/b43/phy_n.c
> > @@ -4222,7 +4222,7 @@ static void b43_nphy_tx_gain_table_upload(struct b43_wldev *dev)
> 
> The TODOs and `#if 0` in this function are concerning.  It looks like
> `rf_pwr_offset_table` is only used when `phy->rev` is >=7 && < 19.
> 
> Further, the loop has a case for `phy->rev >= 19` but we would have
> returned earlier if that was the case.

Yeah, that's why I put the "(unreachable!)" note in the commit log. ;)

> 
> >         u32 rfpwr_offset;
> >         u8 pga_gain, pad_gain;
> >         int i;
> > -       const s16 *uninitialized_var(rf_pwr_offset_table);
> > +       const s16 *rf_pwr_offset_table = NULL;
> >
> >         table = b43_nphy_get_tx_gain_table(dev);
> >         if (!table)
> > @@ -4256,9 +4256,13 @@ static void b43_nphy_tx_gain_table_upload(struct b43_wldev *dev)
> >                         pga_gain = (table[i] >> 24) & 0xf;
> >                         pad_gain = (table[i] >> 19) & 0x1f;
> >                         if (b43_current_band(dev->wl) == NL80211_BAND_2GHZ)
> > -                               rfpwr_offset = rf_pwr_offset_table[pad_gain];
> > +                               rfpwr_offset = rf_pwr_offset_table
> > +                                               ? rf_pwr_offset_table[pad_gain]
> > +                                               : 0;
> >                         else
> > -                               rfpwr_offset = rf_pwr_offset_table[pga_gain];
> > +                               rfpwr_offset = rf_pwr_offset_table
> > +                                               ? rf_pwr_offset_table[pga_gain]
> > +                                               : 0;
> 
> 
> The code is trying to check `phy->rev >= 7 && phy->rev < 19` once
> before the loop, then set `rf_pwr_offset_table`, so having another
> conditional on `rf_pwr_offset_table` in the loop is unnecessary. I'm
> ok with initializing it to `NULL`, but I'm not sure the conditional
> check is necessary.  Do you get a compiler warning otherwise?

I mean, sort of the best thing to do is just remove nearly everything
here since it's actually unreachable. But it is commented as "when
supported ..." etc, so I figured I'd leave it. As part of that I didn't
want to leave any chance of a NULL deref, so I added the explicit tests
just for robustness.

*shrug*

-Kees

-- 
Kees Cook