From: Pavel Skripkin
> Sent: 26 August 2021 11:55
>
> On 8/26/21 1:22 PM, David Laight wrote:
> > From: Pavel Skripkin
> >> Sent: 26 August 2021 10:28
> >>
> >> On 8/26/21 12:22 PM, Pavel Skripkin wrote:
> >> > On Thu, 26 Aug 2021 08:51:23 +0000
> >> > David Laight <David.Laight@xxxxxxxxxx> wrote:
> > ...
> >> >> ...
> >> >> > -static u32 usb_read32(struct intf_hdl *pintfhdl, u32 addr)
> >> >> > +static int usb_read32(struct intf_hdl *pintfhdl, u32 addr, u32
> >> >> > *data) {
> >> >> > u8 requesttype;
> >> >> > u16 wvalue;
> >> >> > u16 len;
> >> >> > - __le32 data;
> >> >> > + int res;
> >> >> > + __le32 tmp;
> >> >> > +
> >> >> > + if (WARN_ON(unlikely(!data)))
> >> >> > + return -EINVAL;
> >> >> >
> >> >>
> >> >> Kill the NULL check - it is a silly coding error.
> >> >> An OOPS is just as easy to debug.
> >> >>
> >> >
> >> >
> >> > I don't think that one single driver should kill the whole system. It's
> >> > 100% an error, but kernel can recover from it (for example disconnect
> >> > the driver, since it's broken).
> >> >
> >> >
> >> > AFIAK, Greg and Linus do not like BUG_ONs in recoverable state...
> >> > Correct me, if I am wrong
> >> >
> >> Oops, I thought about BUG_ON() instead of WARN_ON(), sorry for
> >> confusion. My point is "we should not let the box boom".
> >
> >
> > There is no point checking for NULL that just can't happen.
> > In this case all the callers will pass the address of a local.
> > There really is no point checking.
> >
>
> We not always read in local variable, there are few places, where we
> read into passed buffer.
It is still a very local coding bug.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
