blob: 2f25952c86c6cc2e8ee3bc4c4a237798481cba1e [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001Specification and Internals for the New UHCI Driver (Whitepaper...)
2
3 brought to you by
4
5 Georg Acher, acher@in.tum.de (executive slave) (base guitar)
6 Deti Fliegl, deti@fliegl.de (executive slave) (lead voice)
7 Thomas Sailer, sailer@ife.ee.ethz.ch (chief consultant) (cheer leader)
8
9 $Id: README.uhci,v 1.1 1999/12/14 14:03:02 fliegl Exp $
10
11This document and the new uhci sources can be found on
12 http://hotswap.in.tum.de/usb
13
141. General issues
15
161.1 Why a new UHCI driver, we already have one?!?
17
18Correct, but its internal structure got more and more mixed up by the (still
19ongoing) efforts to get isochronous transfers (ISO) to work.
20Since there is an increasing need for reliable ISO-transfers (especially
21for USB-audio needed by TS and for a DAB-USB-Receiver build by GA and DF),
22this state was a bit unsatisfying in our opinion, so we've decided (based
23on knowledge and experiences with the old UHCI driver) to start
24from scratch with a new approach, much simpler but at the same time more
25powerful.
26It is inspired by the way Win98/Win2000 handles USB requests via URBs,
27but it's definitely 100% free of MS-code and doesn't crash while
28unplugging an used ISO-device like Win98 ;-)
29Some code for HW setup and root hub management was taken from the
30original UHCI driver, but heavily modified to fit into the new code.
31The invention of the basic concept, and major coding were completed in two
32days (and nights) on the 16th and 17th of October 1999, now known as the
33great USB-October-Revolution started by GA, DF, and TS ;-)
34
35Since the concept is in no way UHCI dependent, we hope that it will also be
36transferred to the OHCI-driver, so both drivers share a common API.
37
381.2. Advantages and disadvantages
39
40+ All USB transfer types work now!
41+ Asynchronous operation
42+ Simple, but powerful interface (only two calls for start and cancel)
43+ Easy migration to the new API, simplified by a compatibility API
44+ Simple usage of ISO transfers
45+ Automatic linking of requests
46+ ISO transfers allow variable length for each frame and striping
47+ No CPU dependent and non-portable atomic memory access, no asm()-inlines
48+ Tested on x86 and Alpha
49
50- Rewriting for ISO transfers needed
51
521.3. Is there some compatibility to the old API?
53
54Yes, but only for control, bulk and interrupt transfers. We've implemented
55some wrapper calls for these transfer types. The usbcore works fine with
56these wrappers. For ISO there's no compatibility, because the old ISO-API
57and its semantics were unnecessary complicated in our opinion.
58
591.4. What's really working?
60
61As said above, CTRL and BULK already work fine even with the wrappers,
62so legacy code wouldn't notice the change.
63Regarding to Thomas, ISO transfers now run stable with USB audio.
64INT transfers (e.g. mouse driver) work fine, too.
65
661.5. Are there any bugs?
67
68No ;-)
69Hm...
70Well, of course this implementation needs extensive testing on all available
71hardware, but we believe that any fixes shouldn't harm the overall concept.
72
731.6. What should be done next?
74
75A large part of the request handling seems to be identical for UHCI and
76OHCI, so it would be a good idea to extract the common parts and have only
77the HW specific stuff in uhci.c. Furthermore, all other USB device drivers
78should need URBification, if they use isochronous or interrupt transfers.
79One thing missing in the current implementation (and the old UHCI driver)
80is fair queueing for BULK transfers. Since this would need (in principle)
81the alteration of already constructed TD chains (to switch from depth to
82breadth execution), another way has to be found. Maybe some simple
83heuristics work with the same effect.
84
85---------------------------------------------------------------------------
86
872. Internal structure and mechanisms
88
89To get quickly familiar with the internal structures, here's a short
90description how the new UHCI driver works. However, the ultimate source of
91truth is only uhci.c!
92
932.1. Descriptor structure (QHs and TDs)
94
95During initialization, the following skeleton is allocated in init_skel:
96
97 framespecific | common chain
98
99framelist[]
100[ 0 ]-----> TD --> TD -------\
101[ 1 ]-----> TD --> TD --------> TD ----> QH -------> QH -------> QH ---> NULL
102 ... TD --> TD -------/
103[1023]-----> TD --> TD ------/
104
105 ^^ ^^ ^^ ^^ ^^ ^^
106 1024 TDs for 7 TDs for 1 TD for Start of Start of End Chain
107 ISO INT (2-128ms) 1ms-INT CTRL Chain BULK Chain
108
109For each CTRL or BULK transfer a new QH is allocated and the containing data
110transfers are appended as (vertical) TDs. After building the whole QH with its
111dangling TDs, the QH is inserted before the BULK Chain QH (for CTRL) or
112before the End Chain QH (for BULK). Since only the QH->next pointers are
113affected, no atomic memory operation is required. The three QHs in the
114common chain are never equipped with TDs!
115
116For ISO or INT, the TD for each frame is simply inserted into the appropriate
117ISO/INT-TD-chain for the desired frame. The 7 skeleton INT-TDs are scattered
118among the 1024 frames similar to the old UHCI driver.
119
120For CTRL/BULK/ISO, the last TD in the transfer has the IOC-bit set. For INT,
121every TD (there is only one...) has the IOC-bit set.
122
123Besides the data for the UHCI controller (2 or 4 32bit words), the descriptors
124are double-linked through the .vertical and .horizontal elements in the
125SW data of the descriptor (using the double-linked list structures and
126operations), but SW-linking occurs only in closed domains, i.e. for each of
127the 1024 ISO-chains and the 8 INT-chains there is a closed cycle. This
128simplifies all insertions and unlinking operations and avoids costly
129bus_to_virt()-calls.
130
1312.2. URB structure and linking to QH/TDs
132
133During assembly of the QH and TDs of the requested action, these descriptors
134are stored in urb->urb_list, so the allocated QH/TD descriptors are bound to
135this URB.
136If the assembly was successful and the descriptors were added to the HW chain,
137the corresponding URB is inserted into a global URB list for this controller.
138This list stores all pending URBs.
139
1402.3. Interrupt processing
141
142Since UHCI provides no means to directly detect completed transactions, the
143following is done in each UHCI interrupt (uhci_interrupt()):
144
145For each URB in the pending queue (process_urb()), the ACTIVE-flag of the
146associated TDs are processed (depending on the transfer type
147process_{transfer|interrupt|iso}()). If the TDs are not active anymore,
148they indicate the completion of the transaction and the status is calculated.
149Inactive QH/TDs are removed from the HW chain (since the host controller
150already removed the TDs from the QH, no atomic access is needed) and
151eventually the URB is marked as completed (OK or errors) and removed from the
152pending queue. Then the next linked URB is submitted. After (or immediately
153before) that, the completion handler is called.
154
1552.4. Unlinking URBs
156
157First, all QH/TDs stored in the URB are unlinked from the HW chain.
158To ensure that the host controller really left a vertical TD chain, we
159wait for one frame. After that, the TDs are physically destroyed.
160
1612.5. URB linking and the consequences
162
163Since URBs can be linked and the corresponding submit_urb is called in
164the UHCI-interrupt, all work associated with URB/QH/TD assembly has to be
165interrupt save. This forces kmalloc to use GFP_ATOMIC in the interrupt.