aboutsummaryrefslogtreecommitdiffstats
path: root/Documentation/DocBook
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2014-06-03 12:11:20 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2014-06-03 12:11:20 -0400
commite5c4ecdc55b6d824365ba7964bcd3185223f9688 (patch)
tree405556c9fe703094dd0fc07bc89d9a2fdb5ee525 /Documentation/DocBook
parent49eb7b0750d9483c74e9c14ae6ea1e9d62481c3c (diff)
parent4a95b1fce97756d0333f8232eb7ed6974e93b054 (diff)
Merge tag 'usb-3.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb into next
Pull USB driver updates from Greg KH: "Here is the big USB driver pull request for 3.16-rc1. Nothing huge here, but lots of little things in the USB core, and in lots of drivers. Hopefully the USB power management will be work better now that it has been reworked to do per-port power control dynamically. There's also a raft of gadget driver updates and fixes, CONFIG_USB_DEBUG is finally gone now that everything has been converted over to the dynamic debug inteface, the last hold-out drivers were cleaned up and the config option removed. There were also other minor things all through the drivers/usb/ tree, the shortlog shows this pretty well. All have been in linux-next, including the very last patch, which came from linux-next to fix a build issue on some platforms" * tag 'usb-3.16-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (314 commits) usb: hub_handle_remote_wakeup() only exists for CONFIG_PM=y USB: orinoco_usb: remove CONFIG_USB_DEBUG support USB: media: lirc: igorplugusb: remove CONFIG_USB_DEBUG support USB: media: streamzap: remove CONFIG_USB_DEBUG USB: media: redrat3: remove CONFIG_USB_DEBUG usage USB: media: redrat3: remove unneeded tracing macro usb: qcserial: add additional Sierra Wireless QMI devices usb: host: max3421-hcd: Use module_spi_driver usb: host: max3421-hcd: Allow platform-data to specify Vbus polarity usb: host: max3421-hcd: fix "spi_rd8" uses dynamic stack allocation warning usb: host: max3421-hcd: Fix missing unlock in max3421_urb_enqueue() usb: qcserial: add Netgear AirCard 341U Documentation: dt-bindings: update xhci-platform DT binding for R-Car H2 and M2 usb: host: xhci-plat: add xhci_plat_start() usb: host: max3421-hcd: Fix potential NULL urb dereference Revert "usb: gadget: net2280: Add support for PLX USB338X" USB: usbip: remove CONFIG_USB_DEBUG reference USB: remove CONFIG_USB_DEBUG from defconfig files usb: resume child device when port is powered on usb: hub_handle_remote_wakeup() depends on CONFIG_PM_RUNTIME=y ...
Diffstat (limited to 'Documentation/DocBook')
-rw-r--r--Documentation/DocBook/Makefile3
-rw-r--r--Documentation/DocBook/writing_musb_glue_layer.tmpl873
2 files changed, 875 insertions, 1 deletions
diff --git a/Documentation/DocBook/Makefile b/Documentation/DocBook/Makefile
index b444f2e8fe32..bec06659e0eb 100644
--- a/Documentation/DocBook/Makefile
+++ b/Documentation/DocBook/Makefile
@@ -14,7 +14,8 @@ DOCBOOKS := z8530book.xml device-drivers.xml \
14 genericirq.xml s390-drivers.xml uio-howto.xml scsi.xml \ 14 genericirq.xml s390-drivers.xml uio-howto.xml scsi.xml \
15 80211.xml debugobjects.xml sh.xml regulator.xml \ 15 80211.xml debugobjects.xml sh.xml regulator.xml \
16 alsa-driver-api.xml writing-an-alsa-driver.xml \ 16 alsa-driver-api.xml writing-an-alsa-driver.xml \
17 tracepoint.xml drm.xml media_api.xml w1.xml 17 tracepoint.xml drm.xml media_api.xml w1.xml \
18 writing_musb_glue_layer.xml
18 19
19include Documentation/DocBook/media/Makefile 20include Documentation/DocBook/media/Makefile
20 21
diff --git a/Documentation/DocBook/writing_musb_glue_layer.tmpl b/Documentation/DocBook/writing_musb_glue_layer.tmpl
new file mode 100644
index 000000000000..837eca77f274
--- /dev/null
+++ b/Documentation/DocBook/writing_musb_glue_layer.tmpl
@@ -0,0 +1,873 @@
1<?xml version="1.0" encoding="UTF-8"?>
2<!DOCTYPE book PUBLIC "-//OASIS//DTD DocBook XML V4.1.2//EN"
3 "http://www.oasis-open.org/docbook/xml/4.1.2/docbookx.dtd" []>
4
5<book id="Writing-MUSB-Glue-Layer">
6 <bookinfo>
7 <title>Writing an MUSB Glue Layer</title>
8
9 <authorgroup>
10 <author>
11 <firstname>Apelete</firstname>
12 <surname>Seketeli</surname>
13 <affiliation>
14 <address>
15 <email>apelete at seketeli.net</email>
16 </address>
17 </affiliation>
18 </author>
19 </authorgroup>
20
21 <copyright>
22 <year>2014</year>
23 <holder>Apelete Seketeli</holder>
24 </copyright>
25
26 <legalnotice>
27 <para>
28 This documentation is free software; you can redistribute it
29 and/or modify it under the terms of the GNU General Public
30 License as published by the Free Software Foundation; either
31 version 2 of the License, or (at your option) any later version.
32 </para>
33
34 <para>
35 This documentation is distributed in the hope that it will be
36 useful, but WITHOUT ANY WARRANTY; without even the implied
37 warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
38 See the GNU General Public License for more details.
39 </para>
40
41 <para>
42 You should have received a copy of the GNU General Public License
43 along with this documentation; if not, write to the Free Software
44 Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
45 02111-1307 USA
46 </para>
47
48 <para>
49 For more details see the file COPYING in the Linux kernel source
50 tree.
51 </para>
52 </legalnotice>
53 </bookinfo>
54
55<toc></toc>
56
57 <chapter id="introduction">
58 <title>Introduction</title>
59 <para>
60 The Linux MUSB subsystem is part of the larger Linux USB
61 subsystem. It provides support for embedded USB Device Controllers
62 (UDC) that do not use Universal Host Controller Interface (UHCI)
63 or Open Host Controller Interface (OHCI).
64 </para>
65 <para>
66 Instead, these embedded UDC rely on the USB On-the-Go (OTG)
67 specification which they implement at least partially. The silicon
68 reference design used in most cases is the Multipoint USB
69 Highspeed Dual-Role Controller (MUSB HDRC) found in the Mentor
70 Graphics Inventra™ design.
71 </para>
72 <para>
73 As a self-taught exercise I have written an MUSB glue layer for
74 the Ingenic JZ4740 SoC, modelled after the many MUSB glue layers
75 in the kernel source tree. This layer can be found at
76 drivers/usb/musb/jz4740.c. In this documentation I will walk
77 through the basics of the jz4740.c glue layer, explaining the
78 different pieces and what needs to be done in order to write your
79 own device glue layer.
80 </para>
81 </chapter>
82
83 <chapter id="linux-musb-basics">
84 <title>Linux MUSB Basics</title>
85 <para>
86 To get started on the topic, please read USB On-the-Go Basics (see
87 Resources) which provides an introduction of USB OTG operation at
88 the hardware level. A couple of wiki pages by Texas Instruments
89 and Analog Devices also provide an overview of the Linux kernel
90 MUSB configuration, albeit focused on some specific devices
91 provided by these companies. Finally, getting acquainted with the
92 USB specification at USB home page may come in handy, with
93 practical instance provided through the Writing USB Device Drivers
94 documentation (again, see Resources).
95 </para>
96 <para>
97 Linux USB stack is a layered architecture in which the MUSB
98 controller hardware sits at the lowest. The MUSB controller driver
99 abstract the MUSB controller hardware to the Linux USB stack.
100 </para>
101 <programlisting>
102 ------------------------
103 | | &lt;------- drivers/usb/gadget
104 | Linux USB Core Stack | &lt;------- drivers/usb/host
105 | | &lt;------- drivers/usb/core
106 ------------------------
107
108 --------------------------
109 | | &lt;------ drivers/usb/musb/musb_gadget.c
110 | MUSB Controller driver | &lt;------ drivers/usb/musb/musb_host.c
111 | | &lt;------ drivers/usb/musb/musb_core.c
112 --------------------------
113
114 ---------------------------------
115 | MUSB Platform Specific Driver |
116 | | &lt;-- drivers/usb/musb/jz4740.c
117 | aka &quot;Glue Layer&quot; |
118 ---------------------------------
119
120 ---------------------------------
121 | MUSB Controller Hardware |
122 ---------------------------------
123 </programlisting>
124 <para>
125 As outlined above, the glue layer is actually the platform
126 specific code sitting in between the controller driver and the
127 controller hardware.
128 </para>
129 <para>
130 Just like a Linux USB driver needs to register itself with the
131 Linux USB subsystem, the MUSB glue layer needs first to register
132 itself with the MUSB controller driver. This will allow the
133 controller driver to know about which device the glue layer
134 supports and which functions to call when a supported device is
135 detected or released; remember we are talking about an embedded
136 controller chip here, so no insertion or removal at run-time.
137 </para>
138 <para>
139 All of this information is passed to the MUSB controller driver
140 through a platform_driver structure defined in the glue layer as:
141 </para>
142 <programlisting linenumbering="numbered">
143static struct platform_driver jz4740_driver = {
144 .probe = jz4740_probe,
145 .remove = jz4740_remove,
146 .driver = {
147 .name = "musb-jz4740",
148 },
149};
150 </programlisting>
151 <para>
152 The probe and remove function pointers are called when a matching
153 device is detected and, respectively, released. The name string
154 describes the device supported by this glue layer. In the current
155 case it matches a platform_device structure declared in
156 arch/mips/jz4740/platform.c. Note that we are not using device
157 tree bindings here.
158 </para>
159 <para>
160 In order to register itself to the controller driver, the glue
161 layer goes through a few steps, basically allocating the
162 controller hardware resources and initialising a couple of
163 circuits. To do so, it needs to keep track of the information used
164 throughout these steps. This is done by defining a private
165 jz4740_glue structure:
166 </para>
167 <programlisting linenumbering="numbered">
168struct jz4740_glue {
169 struct device *dev;
170 struct platform_device *musb;
171 struct clk *clk;
172};
173 </programlisting>
174 <para>
175 The dev and musb members are both device structure variables. The
176 first one holds generic information about the device, since it's
177 the basic device structure, and the latter holds information more
178 closely related to the subsystem the device is registered to. The
179 clk variable keeps information related to the device clock
180 operation.
181 </para>
182 <para>
183 Let's go through the steps of the probe function that leads the
184 glue layer to register itself to the controller driver.
185 </para>
186 <para>
187 N.B.: For the sake of readability each function will be split in
188 logical parts, each part being shown as if it was independent from
189 the others.
190 </para>
191 <programlisting linenumbering="numbered">
192static int jz4740_probe(struct platform_device *pdev)
193{
194 struct platform_device *musb;
195 struct jz4740_glue *glue;
196 struct clk *clk;
197 int ret;
198
199 glue = devm_kzalloc(&amp;pdev->dev, sizeof(*glue), GFP_KERNEL);
200 if (!glue)
201 return -ENOMEM;
202
203 musb = platform_device_alloc("musb-hdrc", PLATFORM_DEVID_AUTO);
204 if (!musb) {
205 dev_err(&amp;pdev->dev, "failed to allocate musb device\n");
206 return -ENOMEM;
207 }
208
209 clk = devm_clk_get(&amp;pdev->dev, "udc");
210 if (IS_ERR(clk)) {
211 dev_err(&amp;pdev->dev, "failed to get clock\n");
212 ret = PTR_ERR(clk);
213 goto err_platform_device_put;
214 }
215
216 ret = clk_prepare_enable(clk);
217 if (ret) {
218 dev_err(&amp;pdev->dev, "failed to enable clock\n");
219 goto err_platform_device_put;
220 }
221
222 musb->dev.parent = &amp;pdev->dev;
223
224 glue->dev = &amp;pdev->dev;
225 glue->musb = musb;
226 glue->clk = clk;
227
228 return 0;
229
230err_platform_device_put:
231 platform_device_put(musb);
232 return ret;
233}
234 </programlisting>
235 <para>
236 The first few lines of the probe function allocate and assign the
237 glue, musb and clk variables. The GFP_KERNEL flag (line 8) allows
238 the allocation process to sleep and wait for memory, thus being
239 usable in a blocking situation. The PLATFORM_DEVID_AUTO flag (line
240 12) allows automatic allocation and management of device IDs in
241 order to avoid device namespace collisions with explicit IDs. With
242 devm_clk_get() (line 18) the glue layer allocates the clock -- the
243 <literal>devm_</literal> prefix indicates that clk_get() is
244 managed: it automatically frees the allocated clock resource data
245 when the device is released -- and enable it.
246 </para>
247 <para>
248 Then comes the registration steps:
249 </para>
250 <programlisting linenumbering="numbered">
251static int jz4740_probe(struct platform_device *pdev)
252{
253 struct musb_hdrc_platform_data *pdata = &amp;jz4740_musb_platform_data;
254
255 pdata->platform_ops = &amp;jz4740_musb_ops;
256
257 platform_set_drvdata(pdev, glue);
258
259 ret = platform_device_add_resources(musb, pdev->resource,
260 pdev->num_resources);
261 if (ret) {
262 dev_err(&amp;pdev->dev, "failed to add resources\n");
263 goto err_clk_disable;
264 }
265
266 ret = platform_device_add_data(musb, pdata, sizeof(*pdata));
267 if (ret) {
268 dev_err(&amp;pdev->dev, "failed to add platform_data\n");
269 goto err_clk_disable;
270 }
271
272 return 0;
273
274err_clk_disable:
275 clk_disable_unprepare(clk);
276err_platform_device_put:
277 platform_device_put(musb);
278 return ret;
279}
280 </programlisting>
281 <para>
282 The first step is to pass the device data privately held by the
283 glue layer on to the controller driver through
284 platform_set_drvdata() (line 7). Next is passing on the device
285 resources information, also privately held at that point, through
286 platform_device_add_resources() (line 9).
287 </para>
288 <para>
289 Finally comes passing on the platform specific data to the
290 controller driver (line 16). Platform data will be discussed in
291 <link linkend="device-platform-data">Chapter 4</link>, but here
292 we are looking at the platform_ops function pointer (line 5) in
293 musb_hdrc_platform_data structure (line 3). This function
294 pointer allows the MUSB controller driver to know which function
295 to call for device operation:
296 </para>
297 <programlisting linenumbering="numbered">
298static const struct musb_platform_ops jz4740_musb_ops = {
299 .init = jz4740_musb_init,
300 .exit = jz4740_musb_exit,
301};
302 </programlisting>
303 <para>
304 Here we have the minimal case where only init and exit functions
305 are called by the controller driver when needed. Fact is the
306 JZ4740 MUSB controller is a basic controller, lacking some
307 features found in other controllers, otherwise we may also have
308 pointers to a few other functions like a power management function
309 or a function to switch between OTG and non-OTG modes, for
310 instance.
311 </para>
312 <para>
313 At that point of the registration process, the controller driver
314 actually calls the init function:
315 </para>
316 <programlisting linenumbering="numbered">
317static int jz4740_musb_init(struct musb *musb)
318{
319 musb->xceiv = usb_get_phy(USB_PHY_TYPE_USB2);
320 if (!musb->xceiv) {
321 pr_err("HS UDC: no transceiver configured\n");
322 return -ENODEV;
323 }
324
325 /* Silicon does not implement ConfigData register.
326 * Set dyn_fifo to avoid reading EP config from hardware.
327 */
328 musb->dyn_fifo = true;
329
330 musb->isr = jz4740_musb_interrupt;
331
332 return 0;
333}
334 </programlisting>
335 <para>
336 The goal of jz4740_musb_init() is to get hold of the transceiver
337 driver data of the MUSB controller hardware and pass it on to the
338 MUSB controller driver, as usual. The transceiver is the circuitry
339 inside the controller hardware responsible for sending/receiving
340 the USB data. Since it is an implementation of the physical layer
341 of the OSI model, the transceiver is also referred to as PHY.
342 </para>
343 <para>
344 Getting hold of the MUSB PHY driver data is done with
345 usb_get_phy() which returns a pointer to the structure
346 containing the driver instance data. The next couple of
347 instructions (line 12 and 14) are used as a quirk and to setup
348 IRQ handling respectively. Quirks and IRQ handling will be
349 discussed later in <link linkend="device-quirks">Chapter
350 5</link> and <link linkend="handling-irqs">Chapter 3</link>.
351 </para>
352 <programlisting linenumbering="numbered">
353static int jz4740_musb_exit(struct musb *musb)
354{
355 usb_put_phy(musb->xceiv);
356
357 return 0;
358}
359 </programlisting>
360 <para>
361 Acting as the counterpart of init, the exit function releases the
362 MUSB PHY driver when the controller hardware itself is about to be
363 released.
364 </para>
365 <para>
366 Again, note that init and exit are fairly simple in this case due
367 to the basic set of features of the JZ4740 controller hardware.
368 When writing an musb glue layer for a more complex controller
369 hardware, you might need to take care of more processing in those
370 two functions.
371 </para>
372 <para>
373 Returning from the init function, the MUSB controller driver jumps
374 back into the probe function:
375 </para>
376 <programlisting linenumbering="numbered">
377static int jz4740_probe(struct platform_device *pdev)
378{
379 ret = platform_device_add(musb);
380 if (ret) {
381 dev_err(&amp;pdev->dev, "failed to register musb device\n");
382 goto err_clk_disable;
383 }
384
385 return 0;
386
387err_clk_disable:
388 clk_disable_unprepare(clk);
389err_platform_device_put:
390 platform_device_put(musb);
391 return ret;
392}
393 </programlisting>
394 <para>
395 This is the last part of the device registration process where the
396 glue layer adds the controller hardware device to Linux kernel
397 device hierarchy: at this stage, all known information about the
398 device is passed on to the Linux USB core stack.
399 </para>
400 <programlisting linenumbering="numbered">
401static int jz4740_remove(struct platform_device *pdev)
402{
403 struct jz4740_glue *glue = platform_get_drvdata(pdev);
404
405 platform_device_unregister(glue->musb);
406 clk_disable_unprepare(glue->clk);
407
408 return 0;
409}
410 </programlisting>
411 <para>
412 Acting as the counterpart of probe, the remove function unregister
413 the MUSB controller hardware (line 5) and disable the clock (line
414 6), allowing it to be gated.
415 </para>
416 </chapter>
417
418 <chapter id="handling-irqs">
419 <title>Handling IRQs</title>
420 <para>
421 Additionally to the MUSB controller hardware basic setup and
422 registration, the glue layer is also responsible for handling the
423 IRQs:
424 </para>
425 <programlisting linenumbering="numbered">
426static irqreturn_t jz4740_musb_interrupt(int irq, void *__hci)
427{
428 unsigned long flags;
429 irqreturn_t retval = IRQ_NONE;
430 struct musb *musb = __hci;
431
432 spin_lock_irqsave(&amp;musb->lock, flags);
433
434 musb->int_usb = musb_readb(musb->mregs, MUSB_INTRUSB);
435 musb->int_tx = musb_readw(musb->mregs, MUSB_INTRTX);
436 musb->int_rx = musb_readw(musb->mregs, MUSB_INTRRX);
437
438 /*
439 * The controller is gadget only, the state of the host mode IRQ bits is
440 * undefined. Mask them to make sure that the musb driver core will
441 * never see them set
442 */
443 musb->int_usb &amp;= MUSB_INTR_SUSPEND | MUSB_INTR_RESUME |
444 MUSB_INTR_RESET | MUSB_INTR_SOF;
445
446 if (musb->int_usb || musb->int_tx || musb->int_rx)
447 retval = musb_interrupt(musb);
448
449 spin_unlock_irqrestore(&amp;musb->lock, flags);
450
451 return retval;
452}
453 </programlisting>
454 <para>
455 Here the glue layer mostly has to read the relevant hardware
456 registers and pass their values on to the controller driver which
457 will handle the actual event that triggered the IRQ.
458 </para>
459 <para>
460 The interrupt handler critical section is protected by the
461 spin_lock_irqsave() and counterpart spin_unlock_irqrestore()
462 functions (line 7 and 24 respectively), which prevent the
463 interrupt handler code to be run by two different threads at the
464 same time.
465 </para>
466 <para>
467 Then the relevant interrupt registers are read (line 9 to 11):
468 </para>
469 <itemizedlist>
470 <listitem>
471 <para>
472 MUSB_INTRUSB: indicates which USB interrupts are currently
473 active,
474 </para>
475 </listitem>
476 <listitem>
477 <para>
478 MUSB_INTRTX: indicates which of the interrupts for TX
479 endpoints are currently active,
480 </para>
481 </listitem>
482 <listitem>
483 <para>
484 MUSB_INTRRX: indicates which of the interrupts for TX
485 endpoints are currently active.
486 </para>
487 </listitem>
488 </itemizedlist>
489 <para>
490 Note that musb_readb() is used to read 8-bit registers at most,
491 while musb_readw() allows us to read at most 16-bit registers.
492 There are other functions that can be used depending on the size
493 of your device registers. See musb_io.h for more information.
494 </para>
495 <para>
496 Instruction on line 18 is another quirk specific to the JZ4740
497 USB device controller, which will be discussed later in <link
498 linkend="device-quirks">Chapter 5</link>.
499 </para>
500 <para>
501 The glue layer still needs to register the IRQ handler though.
502 Remember the instruction on line 14 of the init function:
503 </para>
504 <programlisting linenumbering="numbered">
505static int jz4740_musb_init(struct musb *musb)
506{
507 musb->isr = jz4740_musb_interrupt;
508
509 return 0;
510}
511 </programlisting>
512 <para>
513 This instruction sets a pointer to the glue layer IRQ handler
514 function, in order for the controller hardware to call the handler
515 back when an IRQ comes from the controller hardware. The interrupt
516 handler is now implemented and registered.
517 </para>
518 </chapter>
519
520 <chapter id="device-platform-data">
521 <title>Device Platform Data</title>
522 <para>
523 In order to write an MUSB glue layer, you need to have some data
524 describing the hardware capabilities of your controller hardware,
525 which is called the platform data.
526 </para>
527 <para>
528 Platform data is specific to your hardware, though it may cover a
529 broad range of devices, and is generally found somewhere in the
530 arch/ directory, depending on your device architecture.
531 </para>
532 <para>
533 For instance, platform data for the JZ4740 SoC is found in
534 arch/mips/jz4740/platform.c. In the platform.c file each device of
535 the JZ4740 SoC is described through a set of structures.
536 </para>
537 <para>
538 Here is the part of arch/mips/jz4740/platform.c that covers the
539 USB Device Controller (UDC):
540 </para>
541 <programlisting linenumbering="numbered">
542/* USB Device Controller */
543struct platform_device jz4740_udc_xceiv_device = {
544 .name = "usb_phy_gen_xceiv",
545 .id = 0,
546};
547
548static struct resource jz4740_udc_resources[] = {
549 [0] = {
550 .start = JZ4740_UDC_BASE_ADDR,
551 .end = JZ4740_UDC_BASE_ADDR + 0x10000 - 1,
552 .flags = IORESOURCE_MEM,
553 },
554 [1] = {
555 .start = JZ4740_IRQ_UDC,
556 .end = JZ4740_IRQ_UDC,
557 .flags = IORESOURCE_IRQ,
558 .name = "mc",
559 },
560};
561
562struct platform_device jz4740_udc_device = {
563 .name = "musb-jz4740",
564 .id = -1,
565 .dev = {
566 .dma_mask = &amp;jz4740_udc_device.dev.coherent_dma_mask,
567 .coherent_dma_mask = DMA_BIT_MASK(32),
568 },
569 .num_resources = ARRAY_SIZE(jz4740_udc_resources),
570 .resource = jz4740_udc_resources,
571};
572 </programlisting>
573 <para>
574 The jz4740_udc_xceiv_device platform device structure (line 2)
575 describes the UDC transceiver with a name and id number.
576 </para>
577 <para>
578 At the time of this writing, note that
579 &quot;usb_phy_gen_xceiv&quot; is the specific name to be used for
580 all transceivers that are either built-in with reference USB IP or
581 autonomous and doesn't require any PHY programming. You will need
582 to set CONFIG_NOP_USB_XCEIV=y in the kernel configuration to make
583 use of the corresponding transceiver driver. The id field could be
584 set to -1 (equivalent to PLATFORM_DEVID_NONE), -2 (equivalent to
585 PLATFORM_DEVID_AUTO) or start with 0 for the first device of this
586 kind if we want a specific id number.
587 </para>
588 <para>
589 The jz4740_udc_resources resource structure (line 7) defines the
590 UDC registers base addresses.
591 </para>
592 <para>
593 The first array (line 9 to 11) defines the UDC registers base
594 memory addresses: start points to the first register memory
595 address, end points to the last register memory address and the
596 flags member defines the type of resource we are dealing with. So
597 IORESOURCE_MEM is used to define the registers memory addresses.
598 The second array (line 14 to 17) defines the UDC IRQ registers
599 addresses. Since there is only one IRQ register available for the
600 JZ4740 UDC, start and end point at the same address. The
601 IORESOURCE_IRQ flag tells that we are dealing with IRQ resources,
602 and the name &quot;mc&quot; is in fact hard-coded in the MUSB core
603 in order for the controller driver to retrieve this IRQ resource
604 by querying it by its name.
605 </para>
606 <para>
607 Finally, the jz4740_udc_device platform device structure (line 21)
608 describes the UDC itself.
609 </para>
610 <para>
611 The &quot;musb-jz4740&quot; name (line 22) defines the MUSB
612 driver that is used for this device; remember this is in fact
613 the name that we used in the jz4740_driver platform driver
614 structure in <link linkend="linux-musb-basics">Chapter
615 2</link>. The id field (line 23) is set to -1 (equivalent to
616 PLATFORM_DEVID_NONE) since we do not need an id for the device:
617 the MUSB controller driver was already set to allocate an
618 automatic id in <link linkend="linux-musb-basics">Chapter
619 2</link>. In the dev field we care for DMA related information
620 here. The dma_mask field (line 25) defines the width of the DMA
621 mask that is going to be used, and coherent_dma_mask (line 26)
622 has the same purpose but for the alloc_coherent DMA mappings: in
623 both cases we are using a 32 bits mask. Then the resource field
624 (line 29) is simply a pointer to the resource structure defined
625 before, while the num_resources field (line 28) keeps track of
626 the number of arrays defined in the resource structure (in this
627 case there were two resource arrays defined before).
628 </para>
629 <para>
630 With this quick overview of the UDC platform data at the arch/
631 level now done, let's get back to the MUSB glue layer specific
632 platform data in drivers/usb/musb/jz4740.c:
633 </para>
634 <programlisting linenumbering="numbered">
635static struct musb_hdrc_config jz4740_musb_config = {
636 /* Silicon does not implement USB OTG. */
637 .multipoint = 0,
638 /* Max EPs scanned, driver will decide which EP can be used. */
639 .num_eps = 4,
640 /* RAMbits needed to configure EPs from table */
641 .ram_bits = 9,
642 .fifo_cfg = jz4740_musb_fifo_cfg,
643 .fifo_cfg_size = ARRAY_SIZE(jz4740_musb_fifo_cfg),
644};
645
646static struct musb_hdrc_platform_data jz4740_musb_platform_data = {
647 .mode = MUSB_PERIPHERAL,
648 .config = &amp;jz4740_musb_config,
649};
650 </programlisting>
651 <para>
652 First the glue layer configures some aspects of the controller
653 driver operation related to the controller hardware specifics.
654 This is done through the jz4740_musb_config musb_hdrc_config
655 structure.
656 </para>
657 <para>
658 Defining the OTG capability of the controller hardware, the
659 multipoint member (line 3) is set to 0 (equivalent to false)
660 since the JZ4740 UDC is not OTG compatible. Then num_eps (line
661 5) defines the number of USB endpoints of the controller
662 hardware, including endpoint 0: here we have 3 endpoints +
663 endpoint 0. Next is ram_bits (line 7) which is the width of the
664 RAM address bus for the MUSB controller hardware. This
665 information is needed when the controller driver cannot
666 automatically configure endpoints by reading the relevant
667 controller hardware registers. This issue will be discussed when
668 we get to device quirks in <link linkend="device-quirks">Chapter
669 5</link>. Last two fields (line 8 and 9) are also about device
670 quirks: fifo_cfg points to the USB endpoints configuration table
671 and fifo_cfg_size keeps track of the size of the number of
672 entries in that configuration table. More on that later in <link
673 linkend="device-quirks">Chapter 5</link>.
674 </para>
675 <para>
676 Then this configuration is embedded inside
677 jz4740_musb_platform_data musb_hdrc_platform_data structure (line
678 11): config is a pointer to the configuration structure itself,
679 and mode tells the controller driver if the controller hardware
680 may be used as MUSB_HOST only, MUSB_PERIPHERAL only or MUSB_OTG
681 which is a dual mode.
682 </para>
683 <para>
684 Remember that jz4740_musb_platform_data is then used to convey
685 platform data information as we have seen in the probe function
686 in <link linkend="linux-musb-basics">Chapter 2</link>
687 </para>
688 </chapter>
689
690 <chapter id="device-quirks">
691 <title>Device Quirks</title>
692 <para>
693 Completing the platform data specific to your device, you may also
694 need to write some code in the glue layer to work around some
695 device specific limitations. These quirks may be due to some
696 hardware bugs, or simply be the result of an incomplete
697 implementation of the USB On-the-Go specification.
698 </para>
699 <para>
700 The JZ4740 UDC exhibits such quirks, some of which we will discuss
701 here for the sake of insight even though these might not be found
702 in the controller hardware you are working on.
703 </para>
704 <para>
705 Let's get back to the init function first:
706 </para>
707 <programlisting linenumbering="numbered">
708static int jz4740_musb_init(struct musb *musb)
709{
710 musb->xceiv = usb_get_phy(USB_PHY_TYPE_USB2);
711 if (!musb->xceiv) {
712 pr_err("HS UDC: no transceiver configured\n");
713 return -ENODEV;
714 }
715
716 /* Silicon does not implement ConfigData register.
717 * Set dyn_fifo to avoid reading EP config from hardware.
718 */
719 musb->dyn_fifo = true;
720
721 musb->isr = jz4740_musb_interrupt;
722
723 return 0;
724}
725 </programlisting>
726 <para>
727 Instruction on line 12 helps the MUSB controller driver to work
728 around the fact that the controller hardware is missing registers
729 that are used for USB endpoints configuration.
730 </para>
731 <para>
732 Without these registers, the controller driver is unable to read
733 the endpoints configuration from the hardware, so we use line 12
734 instruction to bypass reading the configuration from silicon, and
735 rely on a hard-coded table that describes the endpoints
736 configuration instead:
737 </para>
738 <programlisting linenumbering="numbered">
739static struct musb_fifo_cfg jz4740_musb_fifo_cfg[] = {
740{ .hw_ep_num = 1, .style = FIFO_TX, .maxpacket = 512, },
741{ .hw_ep_num = 1, .style = FIFO_RX, .maxpacket = 512, },
742{ .hw_ep_num = 2, .style = FIFO_TX, .maxpacket = 64, },
743};
744 </programlisting>
745 <para>
746 Looking at the configuration table above, we see that each
747 endpoints is described by three fields: hw_ep_num is the endpoint
748 number, style is its direction (either FIFO_TX for the controller
749 driver to send packets in the controller hardware, or FIFO_RX to
750 receive packets from hardware), and maxpacket defines the maximum
751 size of each data packet that can be transmitted over that
752 endpoint. Reading from the table, the controller driver knows that
753 endpoint 1 can be used to send and receive USB data packets of 512
754 bytes at once (this is in fact a bulk in/out endpoint), and
755 endpoint 2 can be used to send data packets of 64 bytes at once
756 (this is in fact an interrupt endpoint).
757 </para>
758 <para>
759 Note that there is no information about endpoint 0 here: that one
760 is implemented by default in every silicon design, with a
761 predefined configuration according to the USB specification. For
762 more examples of endpoint configuration tables, see musb_core.c.
763 </para>
764 <para>
765 Let's now get back to the interrupt handler function:
766 </para>
767 <programlisting linenumbering="numbered">
768static irqreturn_t jz4740_musb_interrupt(int irq, void *__hci)
769{
770 unsigned long flags;
771 irqreturn_t retval = IRQ_NONE;
772 struct musb *musb = __hci;
773
774 spin_lock_irqsave(&amp;musb->lock, flags);
775
776 musb->int_usb = musb_readb(musb->mregs, MUSB_INTRUSB);
777 musb->int_tx = musb_readw(musb->mregs, MUSB_INTRTX);
778 musb->int_rx = musb_readw(musb->mregs, MUSB_INTRRX);
779
780 /*
781 * The controller is gadget only, the state of the host mode IRQ bits is
782 * undefined. Mask them to make sure that the musb driver core will
783 * never see them set
784 */
785 musb->int_usb &amp;= MUSB_INTR_SUSPEND | MUSB_INTR_RESUME |
786 MUSB_INTR_RESET | MUSB_INTR_SOF;
787
788 if (musb->int_usb || musb->int_tx || musb->int_rx)
789 retval = musb_interrupt(musb);
790
791 spin_unlock_irqrestore(&amp;musb->lock, flags);
792
793 return retval;
794}
795 </programlisting>
796 <para>
797 Instruction on line 18 above is a way for the controller driver to
798 work around the fact that some interrupt bits used for USB host
799 mode operation are missing in the MUSB_INTRUSB register, thus left
800 in an undefined hardware state, since this MUSB controller
801 hardware is used in peripheral mode only. As a consequence, the
802 glue layer masks these missing bits out to avoid parasite
803 interrupts by doing a logical AND operation between the value read
804 from MUSB_INTRUSB and the bits that are actually implemented in
805 the register.
806 </para>
807 <para>
808 These are only a couple of the quirks found in the JZ4740 USB
809 device controller. Some others were directly addressed in the MUSB
810 core since the fixes were generic enough to provide a better
811 handling of the issues for others controller hardware eventually.
812 </para>
813 </chapter>
814
815 <chapter id="conclusion">
816 <title>Conclusion</title>
817 <para>
818 Writing a Linux MUSB glue layer should be a more accessible task,
819 as this documentation tries to show the ins and outs of this
820 exercise.
821 </para>
822 <para>
823 The JZ4740 USB device controller being fairly simple, I hope its
824 glue layer serves as a good example for the curious mind. Used
825 with the current MUSB glue layers, this documentation should
826 provide enough guidance to get started; should anything gets out
827 of hand, the linux-usb mailing list archive is another helpful
828 resource to browse through.
829 </para>
830 </chapter>
831
832 <chapter id="acknowledgements">
833 <title>Acknowledgements</title>
834 <para>
835 Many thanks to Lars-Peter Clausen and Maarten ter Huurne for
836 answering my questions while I was writing the JZ4740 glue layer
837 and for helping me out getting the code in good shape.
838 </para>
839 <para>
840 I would also like to thank the Qi-Hardware community at large for
841 its cheerful guidance and support.
842 </para>
843 </chapter>
844
845 <chapter id="resources">
846 <title>Resources</title>
847 <para>
848 USB Home Page:
849 <ulink url="http://www.usb.org">http://www.usb.org</ulink>
850 </para>
851 <para>
852 linux-usb Mailing List Archives:
853 <ulink url="http://marc.info/?l=linux-usb">http://marc.info/?l=linux-usb</ulink>
854 </para>
855 <para>
856 USB On-the-Go Basics:
857 <ulink url="http://www.maximintegrated.com/app-notes/index.mvp/id/1822">http://www.maximintegrated.com/app-notes/index.mvp/id/1822</ulink>
858 </para>
859 <para>
860 Writing USB Device Drivers:
861 <ulink url="https://www.kernel.org/doc/htmldocs/writing_usb_driver/index.html">https://www.kernel.org/doc/htmldocs/writing_usb_driver/index.html</ulink>
862 </para>
863 <para>
864 Texas Instruments USB Configuration Wiki Page:
865 <ulink url="http://processors.wiki.ti.com/index.php/Usbgeneralpage">http://processors.wiki.ti.com/index.php/Usbgeneralpage</ulink>
866 </para>
867 <para>
868 Analog Devices Blackfin MUSB Configuration:
869 <ulink url="http://docs.blackfin.uclinux.org/doku.php?id=linux-kernel:drivers:musb">http://docs.blackfin.uclinux.org/doku.php?id=linux-kernel:drivers:musb</ulink>
870 </para>
871 </chapter>
872
873</book>