Am Mittwoch, 14. April 2010 16:47:18 schrieb Yauheni Kaliuta:
> > unless you provide your own hard_xmit() to do buffereing.
>
> I have some ideas and your proposals are very important (it's RFC).
> Thank you!
You need to have at least two URBs submitted under load for
optimal performance.
So I suggest you implemet start_hard_xmit() with the following algorithm:
if (fits(skb, ncm->current_buffer) {
copy_into_buffer(skb, ncm->current_buffer);
if (!maximum_skb_fits(ncm->current_buffer)
if (empty(ncm->free_buffers)
stop_queue();
} else {
add(ncm->current_buffer, ncm->used_buffers);
submit(ncm->current_buffer);
ncm->current_buffer = get_buffer(ncm->free_buffers);
copy_into_buffer(skb, ncm->current_buffer);
}
tx_callback():
add(current_buffer, ncm->free_buffers);
if (empty(ncm->used_buffers)) {
if (has_content(ncm->current_buffer)) {
add(ncm->current_buffer, ncm->used_buffers);
submit(ncm->current_buffer);
}
}
start_queue();
I suggest you use buffers of 64Ks, alloc them coherently
and use usb_anchors to queue them associated with preallocated URBs.
Regards
Oliver
--
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
