Re: [patch-2.3.40-pre6] kzalloc() (ala kmem_zalloc() of SVR4)

[Jamie Lokier]

Jeff Garzik wrote:
> >From the last discussion, I believe Jes if not others pointed out that
> adding GFP_ZERO meant adding something new to a speed-critical path,
> which might not be a good idea.

The test can be done at compile time in almost all cases.

static inline void kmalloc(size_t size, int gfp)
{
	if (!builtin_constant_p (gfp) || (gfp & GFP_ZERO))
		__kzalloc (size, gfp);
	else
		__kmalloc (size, gfp);
}

I don't advocate that particular approach but it's there if you want it.

-- Jamie