Re: pool_cache interface and memory constraints

[David Young <dyoung%pobox.com@localhost>]

On Sat, Nov 26, 2011 at 11:04:56AM -0700, Sverre Froyen wrote:
> Ideally, we would need separate pool caches for PCI, ISA, etc., each with its 
> own set of constraints. These constraints are part of the DMA tag structures. 
> It would therefore be nice to be able to access the pool cache by the DMA tag 
> and the size rather than a pool cache name string. Should pool cache (and 
> pool) be extended to support this?

It may help to have a look at the way that I handle this in ixgbe(4).
Start at ixgbe_jcl_reinit() and then have a look at ixgbe_getjcl().

Many improvements to ixgbe(4)'s scheme can be made.  For example,
instead of a freelist protected by a mutex (ixgbe_extmem_head_t), a
pool_cache of ixgbe_extmem_t's could be used.  Also, the system should
arbitrate between drivers for access to DMA buffers---that would be a
big improvement over the current mbuf-cluster scheme where a few driver
instances can starve all others of clusters.

Dave

-- 
David Young
dyoung%pobox.com@localhost    Urbana, IL    (217) 721-9981