On 07/22/2010 02:13 PM, ext Felipe Balbi wrote:
Hi,
On 07/22/2010 02:04 PM, Roger Quadros wrote:
We want to get g_mtp upstream, and the only thing blocking it was the
u_char abstraction mechanism.
I need to know how we can proceed with this. I can see the following
options
1) remove, u_char.c and merge it into f_mtp.c so that people who need
character abstraction can use the existing u_serial.c. bugs in TTY/line
discipline should be fixed.
2) change u_char.c to provide packet abstraction. I don't know how this
is done. Does this mean providing a socket interface?
Please let me know if there is any other way we can proceed. Thanks.
another option is make use of the functionfs. Personally, I think
u_char.c is better approach since userland won't have to know about
functionfs-specific details, but on the other hand, using a network-like
interface would be great as it would be a lot easier to later implement
e.g. mtp-over-ip.
I guess it's just a trade-off of performance in this case. If we can
have a generic way to combine multiple ethernet transfers into one big
usb transfer (to play better with dma) and re-use skbuffs, that would be
great IMO and I'm all for using network-like (socket-like) interface.
I agree with you Felipe, but there is work to do to get the socket interface
abstraction which will obviously take more time. So till that is done option (1)
seems the way to go for now.
FYI, the user space part of MTP stack is open sourced
http://wiki.meego.com/Buteo
regards,
-roger
--
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
