Re: [PATCH] cciss: Fix warnings during compilation under 32bitenvironment

[James Bottomley <James.Bottomley@SteelEye.com>]

On Fri, 2007-04-20 at 12:30 -0700, Andrew Morton wrote:
> On Fri, 20 Apr 2007 14:50:06 -0400
> James Bottomley <James.Bottomley@SteelEye.com> wrote:
> 
> > > CONFIG_LBD=y gives us an additional 3kb of instructions on i386
> > > allnoconfig.  Other architectures might do less well.  It's not a huge
> > > difference, but that's the way in which creeping bloatiness happens.
> > 
> > OK, sure, but if we really care about this saving, then unconditionally
> > casting to u64 is therefore wrong as well ... this is starting to open
> > quite a large can of worms ...
> > 
> > For the record, if we have to do this, I fancy sector_upper_32() ... we
> > should already have some similar accessor for dma_addr_t as well.
> 
> hm.  How about this?
> 
> --- a/include/linux/kernel.h~upper-32-bits
> +++ a/include/linux/kernel.h
> @@ -40,6 +40,17 @@ extern const char linux_proc_banner[];
>  #define DIV_ROUND_UP(n,d) (((n) + (d) - 1) / (d))
>  #define roundup(x, y) ((((x) + ((y) - 1)) / (y)) * (y))
>  
> +/**
> + * upper_32_bits - return bits 32-63 of a number
> + * @n: the number we're accessing
> + *
> + * A basic shift-right of a 64- or 32-bit quantity.  Use this to suppress
> + * the "right shift count >= width of type" warning when that quantity is
> + * 32-bits.
> + */
> +#define upper_32_bits(n) (((u64)(n)) >> 32)

Won't this have the unwanted side effect of promoting everything in a
calculation to long long on 32 bit platforms, even if n was only 32
bits?

> +
> +
>  #define	KERN_EMERG	"<0>"	/* system is unusable			*/
>  #define	KERN_ALERT	"<1>"	/* action must be taken immediately	*/
>  #define	KERN_CRIT	"<2>"	/* critical conditions			*/
> _
> 
> It seems to generate the desired code.  I avoided Alan's ((n >> 31) >> 1)
> trick because it'll generate peculiar results with signed 64-bit
> quantities.

I've seen the trick done similarly with ((n >> 16) >> 16) which
shouldn't have the issue.

James