Re: [PATCH wireless-next 2/2] rtlwifi: rtl8821ae: phy: fix an undefined bitwise shift behavior

[Su Hui]

On 2023/11/22 21:02, Dan Carpenter wrote:
> On Wed, Nov 22, 2023 at 05:02:12PM +0800, Su Hui wrote:
> > Clang staic checker warning:
> > drivers/net/wireless/realtek/rtlwifi/rtl8821ae/phy.c:184:49:
> > 	The result of the left shift is undefined due to shifting by '32',
> > 	which is greater or equal to the width of type 'u32'.
> > 	[core.UndefinedBinaryOperatorResult]
> >
> > If the value of the right operand is negative or is greater than or
> > equal to the width of the promoted left operand, the behavior is
> > undefined.[1][2]
> >
> > For example, when using different gcc's compilation optimizaation options
> > (-O0 or -O2), the result of '(u32)data << 32' is different. One is 0, the
> > other is old value of data. Adding an u64 cast to fix this problem.
> >
> > [1]:https://stackoverflow.com/questions/11270492/what-does-the-c-
> > standard-say-about-bitshifting-more-bits-than-the-width-of-type
> > [2]:https://www.open-std.org/jtc1/sc22/wg14/www/docs/n1256.pdf
> >
> > Fixes: 21e4b0726dc6 ("rtlwifi: rtl8821ae: Move driver from staging to regular tree")
> > Signed-off-by: Su Hui <suhui@xxxxxxxxxxxx>
> > ---
> >   drivers/net/wireless/realtek/rtlwifi/rtl8821ae/phy.c | 8 ++++----
> >   1 file changed, 4 insertions(+), 4 deletions(-)
> >
> > diff --git a/drivers/net/wireless/realtek/rtlwifi/rtl8821ae/phy.c b/drivers/net/wireless/realtek/rtlwifi/rtl8821ae/phy.c
> > index 6df270e29e66..89713e0587b5 100644
> > --- a/drivers/net/wireless/realtek/rtlwifi/rtl8821ae/phy.c
> > +++ b/drivers/net/wireless/realtek/rtlwifi/rtl8821ae/phy.c
> > @@ -106,7 +106,7 @@ u32 rtl8821ae_phy_query_bb_reg(struct ieee80211_hw *hw, u32 regaddr,
> >   		regaddr, bitmask);
> >   	originalvalue = rtl_read_dword(rtlpriv, regaddr);
> >   	bitshift = _rtl8821ae_phy_calculate_bit_shift(bitmask);
> > -	returnvalue = (originalvalue & bitmask) >> bitshift;
> > +	returnvalue = (u64)(originalvalue & bitmask) >> bitshift;
> This is a right shift, not a left shift. << vs >>.

Hi,

It's same for right shift and having a really weird result.

The result of '(u32)data >> 32' is different when using different compiler.
Clang: "(unsigned int)41 >> 32" = 2077469672
Gcc: "(unsigned int)41 >> 32" = 0
> >   
> >   	rtl_dbg(rtlpriv, COMP_RF, DBG_TRACE,
> >   		"BBR MASK=0x%x Addr[0x%x]=0x%x\n",
> > @@ -128,7 +128,7 @@ void rtl8821ae_phy_set_bb_reg(struct ieee80211_hw *hw,
> >   		originalvalue = rtl_read_dword(rtlpriv, regaddr);
> >   		bitshift = _rtl8821ae_phy_calculate_bit_shift(bitmask);
> >   		data = ((originalvalue & (~bitmask)) |
> > -			((data << bitshift) & bitmask));
> > +			(((u64)data << bitshift) & bitmask));
> The checker is printing an accurate warning, however, I'm not sure the
> fix is correct.  Obviously, shift wrapping is bad and your patch would
> eliminate that possibility.  However, data is a u32 so we end up
> discarding the high 32 bits.  I can imagine a different static checker
> would complain about that.

Oh, it's my negligence...

Su Hui