Subversion Repositories or1k

Specification and Internals for the New UHCI Driver (Whitepaper...)

        brought to you by

        Georg Acher, acher@in.tum.de (executive slave) (base guitar)
        Deti Fliegl, deti@fliegl.de (executive slave) (lead voice)
        Thomas Sailer, sailer@ife.ee.ethz.ch (chief consultant) (cheer leader)
 
        $Id: uhci.txt,v 1.1.1.1 2004-04-15 02:32:08 phoenix Exp $

This document and the new uhci sources can be found on
                http://hotswap.in.tum.de/usb

1. General issues

1.1 Why a new UHCI driver, we already have one?!?

Correct, but its internal structure got more and more mixed up by the (still
ongoing) efforts to get isochronous transfers (ISO) to work.
Since there is an increasing need for reliable ISO-transfers (especially 
for USB-audio needed by TS and for a DAB-USB-Receiver build by GA and DF), 
this state was a bit unsatisfying in our opinion, so we've decided (based
on knowledge and experiences with the old UHCI driver) to start 
from scratch with a new approach, much simpler but at the same time more 
powerful.
It is inspired by the way Win98/Win2000 handles USB requests via URBs,
but it's definitely 100% free of MS-code and doesn't crash while 
unplugging an used ISO-device like Win98 ;-)
Some code for HW setup and root hub management was taken from the 
original UHCI driver, but heavily modified to fit into the new code.
The invention of the basic concept, and major coding were completed in two 
days (and nights) on the 16th and 17th of October 1999, now known as the
great USB-October-Revolution started by GA, DF, and TS ;-)

Since the concept is in no way UHCI dependent, we hope that it will also be 
transferred to the OHCI-driver, so both drivers share a common API.

1.2. Advantages and disadvantages

+ All USB transfer types work now!
+ Asynchronous operation
+ Simple, but powerful interface (only two calls for start and cancel)
+ Easy migration to the new API, simplified by a compatibility API
+ Simple usage of ISO transfers
+ Automatic linking of requests
+ ISO transfers allow variable length for each frame and striping
+ No CPU dependent and non-portable atomic memory access, no asm()-inlines
+ Tested on x86 and Alpha

- Rewriting for ISO transfers needed

1.3. Is there some compatibility to the old API?

Yes, but only for control, bulk and interrupt transfers. We've implemented 
some wrapper calls for these transfer types. The usbcore works fine with 
these wrappers. For ISO there's no compatibility, because the old ISO-API 
and its semantics were unnecessary complicated in our opinion.

1.4. What's really working?

As said above, CTRL and BULK already work fine even with the wrappers,
so legacy code wouldn't notice the change.
Regarding to Thomas, ISO transfers now run stable with USB audio.
INT transfers (e.g. mouse driver) work fine, too.

1.5. Are there any bugs?

No ;-)
Hm...
Well, of course this implementation needs extensive testing on all available
hardware, but we believe that any fixes shouldn't harm the overall concept.

1.6. What should be done next?

A large part of the request handling seems to be identical for UHCI and 
OHCI, so it would be a good idea to extract the common parts and have only 
the HW specific stuff in uhci.c. Furthermore, all other USB device drivers
should need URBification, if they use isochronous or interrupt transfers.
One thing missing in the current implementation (and the old UHCI driver) 
is fair queueing for BULK transfers. Since this would need (in principle) 
the alteration of already constructed TD chains (to switch from depth to 
breadth execution), another way has to be found. Maybe some simple 
heuristics work with the same effect.

---------------------------------------------------------------------------

2. Internal structure and mechanisms

To get quickly familiar with the internal structures, here's a short
description how the new UHCI driver works. However, the ultimate source of
truth is only uhci.c!

2.1. Descriptor structure (QHs and TDs)

During initialization, the following skeleton is allocated in init_skel:

         framespecific           |           common chain     

framelist[]
[  0 ]-----> TD --> TD -------\
[  1 ]-----> TD --> TD --------> TD ----> QH -------> QH -------> QH ---> NULL
  ...        TD --> TD -------/
[1023]-----> TD --> TD ------/
             
             ^^     ^^           ^^       ^^          ^^          ^^
   1024 TDs for   7 TDs for    1 TD for   Start of    Start of    End Chain
            ISO  INT (2-128ms) 1ms-INT    CTRL Chain  BULK Chain

For each CTRL or BULK transfer a new QH is allocated and the containing data
transfers are appended as (vertical) TDs. After building the whole QH with its
dangling TDs, the QH is inserted before the BULK Chain QH (for CTRL) or
before the End Chain QH (for BULK). Since only the QH->next pointers are
affected, no atomic memory operation is required. The three QHs in the
common chain are never equipped with TDs!

For ISO or INT, the TD for each frame is simply inserted into the appropriate
ISO/INT-TD-chain for the desired frame. The 7 skeleton INT-TDs are scattered
among the 1024 frames similar to the old UHCI driver.

For CTRL/BULK/ISO, the last TD in the transfer has the IOC-bit set. For INT,
every TD (there is only one...) has the IOC-bit set.

Besides the data for the UHCI controller (2 or 4 32bit words), the descriptors
are double-linked through the .vertical and .horizontal elements in the 
SW data of the descriptor (using the double-linked list structures and 
operations), but SW-linking occurs only in closed domains, i.e. for each of
the 1024 ISO-chains and the 8 INT-chains there is a closed cycle. This 
simplifies all insertions and unlinking operations and avoids costly 
bus_to_virt()-calls.

2.2. URB structure and linking to QH/TDs

During assembly of the QH and TDs of the requested action, these descriptors
are stored in urb->urb_list, so the allocated QH/TD descriptors are bound to
this URB.
If the assembly was successful and the descriptors were added to the HW chain,
the corresponding URB is inserted into a global URB list for this controller.
This list stores all pending URBs.

2.3. Interrupt processing

Since UHCI provides no means to directly detect completed transactions, the
following is done in each UHCI interrupt (uhci_interrupt()):

For each URB in the pending queue (process_urb()), the ACTIVE-flag of the 
associated TDs are processed (depending on the transfer type 
process_{transfer|interrupt|iso}()). If the TDs are not active anymore, 
they indicate the completion of the transaction and the status is calculated. 
Inactive QH/TDs are removed from the HW chain (since the host controller
already removed the TDs from the QH, no atomic access is needed) and 
eventually the URB is marked as completed (OK or errors) and removed from the 
pending queue. Then the next linked URB is submitted. After (or immediately 
before) that, the completion handler is called.

2.4. Unlinking URBs

First, all QH/TDs stored in the URB are unlinked from the HW chain. 
To ensure that the host controller really left a vertical TD chain, we 
wait for one frame. After that, the TDs are physically destroyed.

2.5. URB linking and the consequences

Since URBs can be linked and the corresponding submit_urb is called in
the UHCI-interrupt, all work associated with URB/QH/TD assembly has to be
interrupt save. This forces kmalloc to use GFP_ATOMIC in the interrupt.