Jeff> If you add "kcalloc" instead of hacking "kmalloc" as I did,
Jeff> drivers wanting zeroed pages make the conscious decision of
Jeff> calling another function, with a potentially slower path.
I still don't see the point of doing a kcalloc() except for making
driver code slightly prettier. You have to remember that full pages
grabbed by the mm system do not come out of kmalloc() but
get_free_pages().
Jeff> Such a split would also make it easier to enumerate
Jeff> architecture-dependent paths. For some architectures may choose
Jeff> the default -- kmalloc+memset -- and other archs can choose a
Jeff> page-zeroing idle task, all without affecting
Jeff> kmalloc/kmem_cache_alloc
It's possible that pre zeroing pages for the mm system might be a win,
and if so it is probably a good thing to explore, however we are
talking two different things here. In the pre-zeroed cage you don't go
through kmalloc and kmem_cache_alloc (SLAB) at all.
Jes
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/