On Fri, 2008-07-25 at 09:39 +0800, Herbert Xu wrote:

On Thu, Jul 24, 2008 at 12:47:19PM -0500, Matt Mackall wrote:

quoted

Let's try this again: did you know that ksize could fail depending on
kernel configuration? Most of us would answer no. That suggests the API
is bad. This ranks 12 on Rusty's spectrum of user-friendly APIs:

I think you misunderstood my argument.  I never suggested changing
the existing ksize interface to return an error onto unsuspecting
users.  I suggested creating a new interface that is explicitly
designed to return an error if the underlying implementation
is unable to support this.

I think that could probably be made to work. Perhaps something like:

size_t kmalloc_extra(void *); /* how many extra bytes in this kmalloc?
*/

Which, if it didn't work, could return a nice safe 0. We could argue
about signedness a bit, but I think this would always be safe.

This will also work with all our current kmalloc implementations. The
trouble was calling ksize() on kmem_cache_alloc objects, which happens
to work with SLAB and SLOB.

-- 
Mathematics is the supreme nostalgia of our time.