Re: [PATCH] Altix I/O code cleanup

[Jes Sorensen <jes@trained-monkey.org>]

>>>>> "Christoph" == Christoph Hellwig <hch@infradead.org> writes:

Christoph> On Fri, Oct 10, 2003 at 04:49:57PM -0500, Patrick Gefre
Christoph> wrote:

Christoph> Why do you remove the per-nod wrappers?  Unlike the other
Christoph> these actually had some use as preparation for a node-aware
Christoph> kmalloc..

I believe it was me who nuked it, until we start using it properly I
don't see any reason for keeping it as a placeholder.

>>  - intr_hdl = snia_kmem_alloc_node(sizeof(struct hub_intr_s),
>> KM_NOSLEEP, cnode); + intr_hdl = kmalloc(sizeof(struct hub_intr_s),
>> GFP_KERNEL); ASSERT_ALWAYS(intr_hdl);

Christoph> NULL return not handled again (and the assert is totally
Christoph> useless)

ASSERT_ALWAYS checks it, it may not be pretty but it does check it.

>> -#define NEWAf(ptr,n,f) (ptr = snia_kmem_zalloc((n)*sizeof
>> (*(ptr)), (f&PCIIO_NOSLEEP)?KM_NOSLEEP:KM_SLEEP)) -#define
>> NEWA(ptr,n) (ptr = snia_kmem_zalloc((n)*sizeof (*(ptr)), KM_SLEEP))
>> +#define NEWAf(ptr,n,f) (ptr = snia_kmem_zalloc((n)*sizeof
>> (*(ptr)))) +#define NEWA(ptr,n) (ptr = snia_kmem_zalloc((n)*sizeof
>> (*(ptr)))) #define DELA(ptr,n) (kfree(ptr))

Christoph> What about killing this stupid wrappers while you're at it?
Christoph> Also PCIIO_NOSLEEP is never set.

All the NEWA stuff is going, I sent Pat a patch for it already, it's
going away.

Cheers,
Jes