Re: [PATCH 7/13] i386 sg: add support for chaining scatterlists

[Jens Axboe <jens.axboe@oracle.com>]

On Thu, May 10 2007, Andrew Morton wrote:
> On Thu, 10 May 2007 12:46:53 +0200 Jens Axboe <jens.axboe@oracle.com> wrote:
> 
> > On Thu, May 10 2007, Jens Axboe wrote:
> > > Yeah, I'm inclined to agree. Perhaps it would be better to put this
> > > stuff in lib/scatterlist.c or something like that instead?
> > > 
> > > > > +/*
> > > > > + * Chain previous sglist to this one
> > > > > + */
> > > > > +static inline void sg_chain(struct scatterlist *prv, unsigned int prv_nents,
> > > > > +			    struct scatterlist *sgl)
> > > > > +{
> > > > > +#ifndef ARCH_HAS_SG_CHAIN
> > > > > +	BUG();
> > > > > +#endif
> > > > 
> > > > Can use BUILD_BUG_ON here.  Or just #error.
> > > 
> > > Good idea, thanks!
> > 
> > No wait a second, that wont work. The code is always being built in sg
> > scsi_lib.c, it should just not be called unless we can do chaining. We
> > will never have a large number of segments that require chaining without
> > ARCH_HAS_SG_CHAIN, so it'll never be called in that case. So it has to
> > remain as it is, a BUG().
> 
> Confused.  If it should never be called, why does it even get compiled in?

I can hide it behind ARCH_HAS_SG_CHAIN and provide something ala

#define sg_chain(prv, x, sgl)   BUG()

for when that is not defined. It still needs to be visible.

-- 
Jens Axboe