Alan Stern wrote:
> On Fri, 30 Jan 2009, David Vrabel wrote:
>
>> All,
>>
>> The epic USB 3.0 thread was getting a bit long so here's some actual
>> (untested!) code to discuss.
>>
>> If an HCD supports sg lists then we simply submit one urb with the
>> mapped sg list (urb->sg) instead of many urbs each with an
>> urb->transfer_buffer.
>>
>> I think this is a good first step. Later on we can look at improving the
>> API to allow urbs with sg lists to be submitted asynchronously which may
>> be more useful to drivers other than mass storage.
>
> It's a decent start.
>
> Instead of adding new fields to struct urb, we could overload the
> existing fields. For instance, the scatterlist address could be stored
> in urb->transfer_buffer and the number of entries could be stored in
> urb->number_of_packets. If that value is 0 then it's a normal URB.
> (That's because only Bulk will use sg; Iso won't.)
I think this is too confusing. I could add some inline functions for
retrieving the sg related values if you still want the overloaded fields.
> The code that takes care of setting up the transfer buffer addresses
> should be split out and EXPORTed, because any HCD that supports sg will
> want to use it.
I'm not sure what code you're referring to?
The sg list is already mapped and the HCD need only do something like this:
for_each_sg(sg, sg, std->urb->sg_nents, i) {
std->sgl_virt[i].buf_ptr = cpu_to_le64(sg_dma_address(sg));
std->sgl_virt[i].len = cpu_to_le32(sg_dma_len(sg));
}
to fill in the hardware's sg list.
> Perhaps it would be nice to see more code sharing between the two
> sg_init routines, but that's a minor point.
Heh, yeah. It was a Friday afternoon hack just to get something out
there to discuss.
> Lastly, usbmon will of course have to be modified.
Thanks for the heads up on this. I hadn't considered usbmon at all.
David
--
David Vrabel, Senior Software Engineer, Drivers
CSR, Churchill House, Cambridge Business Park, Tel: +44 (0)1223 692562
Cowley Road, Cambridge, CB4 0WZ http://www.csr.com/
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
