[RFC] [PATCH] usbatm.[ch]: logic changes + error handling

[Roman Kagan]

On Fri, Apr 15, 2005 at 01:14:25PM +0200, Duncan Sands wrote:
> contiguous?  See usb_sg_request.  Each piece should be contiguous, but
> the whole needn't be.  DMA-able is true.

Hmm, didn't know about this guy.  But it's way too complex to use, and
with data blocks of at most 48 bytes I suspect we'll loose more than we
can gain...

> The less CPU we use, the more there is for everyone else...  Anyway,
> some things could be improved.  For example, if we calculated the crc
> progressively as we copy to the buffer (usbatm_write_cells) then we
> should get better cache utilization.

I'm not sure: usbatm_atm_send(), where it's calculated now (well, in my
patch the crc on the padding is calculated in usbatm_write_cells, but
let's forget about it for the moment), is called by the atm layer
normaly when it has produced the data it gives us in the skb, so it's
cache-hot (unless the cache is too small to fit the whole skb in).

Roman.