From: Niklas Schnelle
> Sent: 21 April 2021 12:18
>
> When PCI_IOBASE is not defined, it is set to 0 such that it is ignored
> in calls to the readX/writeX primitives. This triggers clang's
> -Wnull-pointer-arithmetic warning and will result in illegal accesses on
> platforms that do not support I/O ports if drivers do still attempt to
> access them.
>
> Make things explicit and silence the warning by letting inb() and
> friends fail with WARN_ONCE() and a 0xff... return in case PCI_IOBASE is
> not defined.
...
>
> diff --git a/include/asm-generic/io.h b/include/asm-generic/io.h
> index c6af40ce03be..aabb0a8186ee 100644
> --- a/include/asm-generic/io.h
> +++ b/include/asm-generic/io.h
...
> @@ -458,12 +454,17 @@ static inline void writesq(volatile void __iomem *addr, const void *buffer,
> #define _inb _inb
> static inline u8 _inb(unsigned long addr)
> {
> +#ifdef PCI_IOBASE
> u8 val;
>
> __io_pbr();
> val = __raw_readb(PCI_IOBASE + addr);
> __io_par(val);
> return val;
> +#else
> + WARN_ONCE(1, "No I/O port support\n");
> + return ~0;
> +#endif
> }
> #endif
I suspect that this might be better not inlined
when PCI_IOBASE is undefined.
Otherwise you get quite a lot of bloat from all the
WARN_ONCE() calls.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
Registration No: 1397386 (Wales)
On Wed, 2021-04-21 at 11:24 +0000, David Laight wrote:
> From: Niklas Schnelle
> > Sent: 21 April 2021 12:18
> >
> > When PCI_IOBASE is not defined, it is set to 0 such that it is ignored
> > in calls to the readX/writeX primitives. This triggers clang's
> > -Wnull-pointer-arithmetic warning and will result in illegal accesses on
> > platforms that do not support I/O ports if drivers do still attempt to
> > access them.
> >
> > Make things explicit and silence the warning by letting inb() and
> > friends fail with WARN_ONCE() and a 0xff... return in case PCI_IOBASE is
> > not defined.
> ...
> > diff --git a/include/asm-generic/io.h b/include/asm-generic/io.h
> > index c6af40ce03be..aabb0a8186ee 100644
> > --- a/include/asm-generic/io.h
> > +++ b/include/asm-generic/io.h
> ...
> > @@ -458,12 +454,17 @@ static inline void writesq(volatile void __iomem *addr, const void *buffer,
> > #define _inb _inb
> > static inline u8 _inb(unsigned long addr)
> > {
> > +#ifdef PCI_IOBASE
> > u8 val;
> >
> > __io_pbr();
> > val = __raw_readb(PCI_IOBASE + addr);
> > __io_par(val);
> > return val;
> > +#else
> > + WARN_ONCE(1, "No I/O port support\n");
> > + return ~0;
> > +#endif
> > }
> > #endif
>
> I suspect that this might be better not inlined
> when PCI_IOBASE is undefined.
>
> Otherwise you get quite a lot of bloat from all the
> WARN_ONCE() calls.
>
> David
Hmm, I was wondering if we should rather have a large ifdef block of
all these functions stubbed to WARN_ONCE rather than in each function.
As I understand it this would be necessary if we want the inline gone.
They would still be static though so we still get a copy per
compilation unit that uses it or am I misunderstanding?
>
> -
> Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK
> Registration No: 1397386 (Wales)
>
On Wed, Apr 21, 2021 at 1:50 PM Niklas Schnelle <[email protected]> wrote:
> On Wed, 2021-04-21 at 11:24 +0000, David Laight wrote:
> >
> > I suspect that this might be better not inlined
> > when PCI_IOBASE is undefined.
> >
> > Otherwise you get quite a lot of bloat from all the
> > WARN_ONCE() calls.
>
> Hmm, I was wondering if we should rather have a large ifdef block of
> all these functions stubbed to WARN_ONCE rather than in each function.
> As I understand it this would be necessary if we want the inline gone.
> They would still be static though so we still get a copy per
> compilation unit that uses it or am I misunderstanding?
I wouldn't worry too much about the size of known broken drivers during
compile testing. Also, since the functions are marked 'inline' and not
'__always_inline', the compiler is free to decide not to inline them if
the contents are excessively big, and since the strings are all identical,
they should be constant-folded.
If you want to make this a little smaller, using pr_warn_once()
would be a little smaller, but also give less information.
Arnd