Hello Ping-Ke Shih,
The patch e3ec7017f6a2: "rtw89: add Realtek 802.11ax driver" from Oct
11, 2021, leads to the following potential issue (just from reading
the code):
drivers/net/wireless/realtek/rtw89/core.h
3878 static inline u32
3879 rtw89_read32_mask(struct rtw89_dev *rtwdev, u32 addr, u32 mask)
3880 {
3881 u32 shift = __ffs(mask);
3882 u32 orig;
3883 u32 ret;
3884
3885 orig = rtw89_read32(rtwdev, addr);
--> 3886 ret = (orig & mask) >> shift;
I think this line should be:
ret = (orig & mask) >> (shift - 1);
A typical mask here is 0xff so __ffs() is 1 because the first bit is
set. This code will do: ret = (orig & 0xff) >> 1; I think it should be
ret = (orig & 0xff) >> 0;
If the mask was 0xff00 I would expect >> 8 instead of >> 9 etc.
3887
3888 return ret;
3889 }
regards,
dan carpenter
> -----Original Message-----
> From: Dan Carpenter <[email protected]>
> Sent: Monday, December 12, 2022 11:34 PM
> To: Ping-Ke Shih <[email protected]>
> Cc: [email protected]
> Subject: [bug report] rtw89: add Realtek 802.11ax driver
>
> Hello Ping-Ke Shih,
>
> The patch e3ec7017f6a2: "rtw89: add Realtek 802.11ax driver" from Oct
> 11, 2021, leads to the following potential issue (just from reading
> the code):
>
> drivers/net/wireless/realtek/rtw89/core.h
> 3878 static inline u32
> 3879 rtw89_read32_mask(struct rtw89_dev *rtwdev, u32 addr, u32 mask)
> 3880 {
> 3881 u32 shift = __ffs(mask);
> 3882 u32 orig;
> 3883 u32 ret;
> 3884
> 3885 orig = rtw89_read32(rtwdev, addr);
> --> 3886 ret = (orig & mask) >> shift;
>
> I think this line should be:
>
> ret = (orig & mask) >> (shift - 1);
>
> A typical mask here is 0xff so __ffs() is 1 because the first bit is
> set. This code will do: ret = (orig & 0xff) >> 1; I think it should be
> ret = (orig & 0xff) >> 0;
>
> If the mask was 0xff00 I would expect >> 8 instead of >> 9 etc.
ffs(0xff)=1 and __ffs(0xff)=0, so I think original is correct.
Ping-Ke
On Tue, Dec 13, 2022 at 12:46:22AM +0000, Ping-Ke Shih wrote:
>
>
> > -----Original Message-----
> > From: Dan Carpenter <[email protected]>
> > Sent: Monday, December 12, 2022 11:34 PM
> > To: Ping-Ke Shih <[email protected]>
> > Cc: [email protected]
> > Subject: [bug report] rtw89: add Realtek 802.11ax driver
> >
> > Hello Ping-Ke Shih,
> >
> > The patch e3ec7017f6a2: "rtw89: add Realtek 802.11ax driver" from Oct
> > 11, 2021, leads to the following potential issue (just from reading
> > the code):
> >
> > drivers/net/wireless/realtek/rtw89/core.h
> > 3878 static inline u32
> > 3879 rtw89_read32_mask(struct rtw89_dev *rtwdev, u32 addr, u32 mask)
> > 3880 {
> > 3881 u32 shift = __ffs(mask);
> > 3882 u32 orig;
> > 3883 u32 ret;
> > 3884
> > 3885 orig = rtw89_read32(rtwdev, addr);
> > --> 3886 ret = (orig & mask) >> shift;
> >
> > I think this line should be:
> >
> > ret = (orig & mask) >> (shift - 1);
> >
> > A typical mask here is 0xff so __ffs() is 1 because the first bit is
> > set. This code will do: ret = (orig & 0xff) >> 1; I think it should be
> > ret = (orig & 0xff) >> 0;
> >
> > If the mask was 0xff00 I would expect >> 8 instead of >> 9 etc.
>
> ffs(0xff)=1 and __ffs(0xff)=0, so I think original is correct.
Oh, wow. You're right. I hadn't realized it worked like that.
Sorry!
regards,
dan carpenter