Re: SLAB vs. pci_alloc_xxx in usb-uhci patch [RFC: API]

[David Brownell <david-b@pacbell.net>]

> >     The reverse mapping
> > code hast to be less than 0.1KB.
> 
> If reverse mapping means bus_to_virt(), then I would suggest not to
> provide it since it is a confusing interface. OTOH, only a few drivers
> need or want to retrieve the virtual address that lead to some bus dma

Your SCSI code went the other way; the logic is about the same.
That's easy enough ... I'm not going to argue that point any longer.

The driver might even have Real Intelligence to apply.  But I wonder
how many assumptions drivers will end up making about those dma
mappings.  It may be important to expose the "logical" page size to
the driver ("don't cross 4k boundaries"); currently it's hidden.

Other than that L1_CACHE goof, it seems this was the main thing
needing to change in the I API sent by.  Sound right?  Implementation
would be a different question.  I'm not in the least attached to what
I sent by, but some implementation is needed.  Slab-like, or buddy? :)


> Does 'usable' apply to Java applications ? :-)

Servers and other non-gui tools?  I don't see why not.  You can make
good systems software in many languages.  There are advantages to
not having those classes of memory-related bugs.  I'm looking forward
to GCC 3.0 with GCJ, compiling Java just like C.  But that's OT.

- Dave