Re: [PATCH 14/14] powerpc: remove address space overrides using set_fs()

[Christophe Leroy]

Le 03/09/2020 à 17:56, Christoph Hellwig a écrit :
> On Thu, Sep 03, 2020 at 05:49:09PM +0200, Christoph Hellwig wrote:
> > On Thu, Sep 03, 2020 at 05:43:25PM +0200, Christophe Leroy wrote:
> > > Le 03/09/2020 à 16:22, Christoph Hellwig a écrit :
> > > > Stop providing the possibility to override the address space using
> > > > set_fs() now that there is no need for that any more.
> > > >
> > > > Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> > > > ---
> > >
> > >
> > > >    -static inline int __access_ok(unsigned long addr, unsigned long size,
> > > > -			mm_segment_t seg)
> > > > +static inline bool __access_ok(unsigned long addr, unsigned long size)
> > > >    {
> > > > -	if (addr > seg.seg)
> > > > -		return 0;
> > > > -	return (size == 0 || size - 1 <= seg.seg - addr);
> > > > +	if (addr >= TASK_SIZE_MAX)
> > > > +		return false;
> > > > +	return size == 0 || size <= TASK_SIZE_MAX - addr;
> > > >    }
> > >
> > > You don't need to test size == 0 anymore. It used to be necessary because
> > > of the 'size - 1', as size is unsigned.
> > >
> > > Now you can directly do
> > >
> > > 	return size <= TASK_SIZE_MAX - addr;
> > >
> > > If size is 0, this will always be true (because you already know that addr
> > > is not >= TASK_SIZE_MAX
> >
> > True.  What do you think of Linus' comment about always using the
> > ppc32 version on ppc64 as well with this?

I have nothing against it. That's only adding a substract, all args are 
already in registers so that will be in the noise for a modern CPU.

> i.e. something like this folded in:
>
> diff --git a/arch/powerpc/include/asm/uaccess.h b/arch/powerpc/include/asm/uaccess.h
> index 5363f7fc6dd06c..be070254e50943 100644
> --- a/arch/powerpc/include/asm/uaccess.h
> +++ b/arch/powerpc/include/asm/uaccess.h
> @@ -11,26 +11,14 @@
>   #ifdef __powerpc64__
>   /* We use TASK_SIZE_USER64 as TASK_SIZE is not constant */
>   #define TASK_SIZE_MAX		TASK_SIZE_USER64
> -
> -/*
> - * This check is sufficient because there is a large enough gap between user
> - * addresses and the kernel addresses.
> - */
> -static inline bool __access_ok(unsigned long addr, unsigned long size)
> -{
> -	return addr < TASK_SIZE_MAX && size < TASK_SIZE_MAX;
> -}
> -
>   #else
>   #define TASK_SIZE_MAX		TASK_SIZE
> +#endif
>   
>   static inline bool __access_ok(unsigned long addr, unsigned long size)
>   {
> -	if (addr >= TASK_SIZE_MAX)
> -		return false;
> -	return size == 0 || size <= TASK_SIZE_MAX - addr;
> +	return addr < TASK_SIZE_MAX && size <= TASK_SIZE_MAX - addr;
>   }
> -#endif /* __powerpc64__ */
>   
>   #define access_ok(addr, size)		\
>   	(__chk_user_ptr(addr),		\


Christophe