Re: [Intel-IOMMU 02/10] Library routine for pre-allocat pool handling

[Andrew Morton <akpm@linux-foundation.org>]

On Mon, 11 Jun 2007 13:44:42 -0700
"Keshavamurthy, Anil S" <anil.s.keshavamurthy@intel.com> wrote:

> In the first implementation of ours, we had used mempools api's to 
> allocate memory and we were told that mempools with GFP_ATOMIC is
> useless and hence in the second implementation we came up with
> resource pools ( which is preallocate pools) and again as I understand
> the argument is why create another when we have slab allocation which
> is similar to this resource pools.

Odd.  mempool with GFP_ATOMIC is basically equivalent to your
resource-pools, isn't it?: we'll try the slab allocator and if that failed,
fall back to the reserves.

It's missing the recharge-from-a-kernel-thread functionality but that can be
added easily enough if it's useful.  It's slightly abusive of the mempool
philosophy, but it's probably better to do that than to create a new and
very-similar thing.

> Hence, can I assume that the conclusion of this 
> discussion is to use kmem_cache_alloc() functions 
> to allocate memory in dma_map_{single|sg} API's?
> 
> Again, if dma_map_{single|sg} API's fails due to 
> failure to allocate memory, the only thing that can
> be done is to panic as this is what few of the other 
> IOMMU implementation is doing today. 

If the only option is to panic then something's busted.  If it's network IO
then there should be a way of dropping the frame.  If it's disk IO then we
should report the failure and cause an IO error.