USB3 (wasMoving an HD from one comp to another)

[Lennart Sorensen]

On Fri, Nov 26, 2010 at 10:36:49AM -0800, Tyler Aviss wrote:
> So if you connect a USB2 device to a computer's USB3 bus, does it work
> happily without crapping out the CPU, or only enable DMA/interrupt for
> USB3 enabled devices?

No.  A USB3 port is really two ports in one.  It is a USB2 port with
all the regular pins and an EHCI+OHCI/UHCI controller, and then it is
an XHCI controller as well using seperate data lines for USB3 devices.
Devices start out talking to the USB2 part of the port, and if the device
and the port and any hubs along the way are all able to do USB3, then
they enable the USB3 part of the link for data transfers.  Some control
messages will still use the USB2 part of the link at times apparently.

A USB3 port will not make any change in behaviour to USB1 or 2 devices.
The USB1/2 protocol has no concept of interrupts or DMA.  USB3 has a
lot more in the protocol, but only USB3 devices and ports support the
new features.

> I haven't bought any USB3-enabled boards yet. What's the state of
> linux-friendliness of the controllers. I know i've had a lot of USB2
> devices that would drop-off on various boards in 'nix but not in
> windows for some reason...

Linux has XHCI controller support for a while now.  Probably had it a
while before windows did.  That was the case with USB2's EHCI in the
past as well.  Linux had it first.

Some USB2 controllers were badly made, and there were a few kernel bugs
some years ago too that affected some implementations (SiS chip sets
were especially effected)

-- 
Len Sorensen
--
The Toronto Linux Users Group.      Meetings: http://gtalug.org/
TLUG requests: Linux topics, No HTML, wrap text below 80 columns
How to UNSUBSCRIBE: http://gtalug.org/wiki/Mailing_lists