> ANother question: how much work is there to be done on the userspace
> side of things?
>
> > > */
> > >
> > > -extern void *kmem_alloc_io(size_t size, int align_mask, gfp_t flags);
> > > -extern void *kmem_alloc_large(size_t size, gfp_t);
> > > +extern void *xfs_kmem_alloc(size_t, gfp_t, bool, int);
> > > static inline void kmem_free(const void *ptr)
> > > {
> > > kvfree(ptr);
> > > }
>
> Didn't an earlier patch get rid of kmem_free(), or am I just
> imagining this? Seems silly to leave this behind, now that the
> only place that needs kvfree() is the callers to kmem_alloc_io and
> kmem_alloc_large...
Yes, it's on my roadmap, I just wanted to work on the other wrappers first and
send it later when I had kmem_alloc_{io, large}() worked out. I'll make sure to
convert this one on the V2.
>
> Cheers,
>
> Dave.
> --
> Dave Chinner
> david@xxxxxxxxxxxxx
>
--
Carlos
