2 * I/O functions for libusbx
3 * Copyright © 2007-2009 Daniel Drake <dsd@gentoo.org>
4 * Copyright © 2001 Johannes Erdfelt <johannes@erdfelt.com>
6 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 #ifdef HAVE_SYS_TIME_H
33 #ifdef USBI_TIMERFD_AVAILABLE
34 #include <sys/timerfd.h>
40 * \page io Synchronous and asynchronous device I/O
42 * \section intro Introduction
44 * If you're using libusbx in your application, you're probably wanting to
45 * perform I/O with devices - you want to perform USB data transfers.
47 * libusbx offers two separate interfaces for device I/O. This page aims to
48 * introduce the two in order to help you decide which one is more suitable
49 * for your application. You can also choose to use both interfaces in your
50 * application by considering each transfer on a case-by-case basis.
52 * Once you have read through the following discussion, you should consult the
53 * detailed API documentation pages for the details:
57 * \section theory Transfers at a logical level
59 * At a logical level, USB transfers typically happen in two parts. For
60 * example, when reading data from a endpoint:
61 * -# A request for data is sent to the device
62 * -# Some time later, the incoming data is received by the host
64 * or when writing data to an endpoint:
66 * -# The data is sent to the device
67 * -# Some time later, the host receives acknowledgement from the device that
68 * the data has been transferred.
70 * There may be an indefinite delay between the two steps. Consider a
71 * fictional USB input device with a button that the user can press. In order
72 * to determine when the button is pressed, you would likely submit a request
73 * to read data on a bulk or interrupt endpoint and wait for data to arrive.
74 * Data will arrive when the button is pressed by the user, which is
75 * potentially hours later.
77 * libusbx offers both a synchronous and an asynchronous interface to performing
78 * USB transfers. The main difference is that the synchronous interface
79 * combines both steps indicated above into a single function call, whereas
80 * the asynchronous interface separates them.
82 * \section sync The synchronous interface
84 * The synchronous I/O interface allows you to perform a USB transfer with
85 * a single function call. When the function call returns, the transfer has
86 * completed and you can parse the results.
88 * If you have used the libusb-0.1 before, this I/O style will seem familar to
89 * you. libusb-0.1 only offered a synchronous interface.
91 * In our input device example, to read button presses you might write code
92 * in the following style:
94 unsigned char data[4];
96 int r = libusb_bulk_transfer(handle, LIBUSB_ENDPOINT_IN, data, sizeof(data), &actual_length, 0);
97 if (r == 0 && actual_length == sizeof(data)) {
98 // results of the transaction can now be found in the data buffer
99 // parse them here and report button press
105 * The main advantage of this model is simplicity: you did everything with
106 * a single simple function call.
108 * However, this interface has its limitations. Your application will sleep
109 * inside libusb_bulk_transfer() until the transaction has completed. If it
110 * takes the user 3 hours to press the button, your application will be
111 * sleeping for that long. Execution will be tied up inside the library -
112 * the entire thread will be useless for that duration.
114 * Another issue is that by tieing up the thread with that single transaction
115 * there is no possibility of performing I/O with multiple endpoints and/or
116 * multiple devices simultaneously, unless you resort to creating one thread
119 * Additionally, there is no opportunity to cancel the transfer after the
120 * request has been submitted.
122 * For details on how to use the synchronous API, see the
123 * \ref syncio "synchronous I/O API documentation" pages.
125 * \section async The asynchronous interface
127 * Asynchronous I/O is the most significant new feature in libusb-1.0.
128 * Although it is a more complex interface, it solves all the issues detailed
131 * Instead of providing which functions that block until the I/O has complete,
132 * libusbx's asynchronous interface presents non-blocking functions which
133 * begin a transfer and then return immediately. Your application passes a
134 * callback function pointer to this non-blocking function, which libusbx will
135 * call with the results of the transaction when it has completed.
137 * Transfers which have been submitted through the non-blocking functions
138 * can be cancelled with a separate function call.
140 * The non-blocking nature of this interface allows you to be simultaneously
141 * performing I/O to multiple endpoints on multiple devices, without having
144 * This added flexibility does come with some complications though:
145 * - In the interest of being a lightweight library, libusbx does not create
146 * threads and can only operate when your application is calling into it. Your
147 * application must call into libusbx from it's main loop when events are ready
148 * to be handled, or you must use some other scheme to allow libusbx to
149 * undertake whatever work needs to be done.
150 * - libusbx also needs to be called into at certain fixed points in time in
151 * order to accurately handle transfer timeouts.
152 * - Memory handling becomes more complex. You cannot use stack memory unless
153 * the function with that stack is guaranteed not to return until the transfer
154 * callback has finished executing.
155 * - You generally lose some linearity from your code flow because submitting
156 * the transfer request is done in a separate function from where the transfer
157 * results are handled. This becomes particularly obvious when you want to
158 * submit a second transfer based on the results of an earlier transfer.
160 * Internally, libusbx's synchronous interface is expressed in terms of function
161 * calls to the asynchronous interface.
163 * For details on how to use the asynchronous API, see the
164 * \ref asyncio "asynchronous I/O API" documentation pages.
169 * \page packetoverflow Packets and overflows
171 * \section packets Packet abstraction
173 * The USB specifications describe how data is transmitted in packets, with
174 * constraints on packet size defined by endpoint descriptors. The host must
175 * not send data payloads larger than the endpoint's maximum packet size.
177 * libusbx and the underlying OS abstract out the packet concept, allowing you
178 * to request transfers of any size. Internally, the request will be divided
179 * up into correctly-sized packets. You do not have to be concerned with
180 * packet sizes, but there is one exception when considering overflows.
182 * \section overflow Bulk/interrupt transfer overflows
184 * When requesting data on a bulk endpoint, libusbx requires you to supply a
185 * buffer and the maximum number of bytes of data that libusbx can put in that
186 * buffer. However, the size of the buffer is not communicated to the device -
187 * the device is just asked to send any amount of data.
189 * There is no problem if the device sends an amount of data that is less than
190 * or equal to the buffer size. libusbx reports this condition to you through
191 * the \ref libusb_transfer::actual_length "libusb_transfer.actual_length"
194 * Problems may occur if the device attempts to send more data than can fit in
195 * the buffer. libusbx reports LIBUSB_TRANSFER_OVERFLOW for this condition but
196 * other behaviour is largely undefined: actual_length may or may not be
197 * accurate, the chunk of data that can fit in the buffer (before overflow)
198 * may or may not have been transferred.
200 * Overflows are nasty, but can be avoided. Even though you were told to
201 * ignore packets above, think about the lower level details: each transfer is
202 * split into packets (typically small, with a maximum size of 512 bytes).
203 * Overflows can only happen if the final packet in an incoming data transfer
204 * is smaller than the actual packet that the device wants to transfer.
205 * Therefore, you will never see an overflow if your transfer buffer size is a
206 * multiple of the endpoint's packet size: the final packet will either
207 * fill up completely or will be only partially filled.
211 * @defgroup asyncio Asynchronous device I/O
213 * This page details libusbx's asynchronous (non-blocking) API for USB device
214 * I/O. This interface is very powerful but is also quite complex - you will
215 * need to read this page carefully to understand the necessary considerations
216 * and issues surrounding use of this interface. Simplistic applications
217 * may wish to consider the \ref syncio "synchronous I/O API" instead.
219 * The asynchronous interface is built around the idea of separating transfer
220 * submission and handling of transfer completion (the synchronous model
221 * combines both of these into one). There may be a long delay between
222 * submission and completion, however the asynchronous submission function
223 * is non-blocking so will return control to your application during that
224 * potentially long delay.
226 * \section asyncabstraction Transfer abstraction
228 * For the asynchronous I/O, libusbx implements the concept of a generic
229 * transfer entity for all types of I/O (control, bulk, interrupt,
230 * isochronous). The generic transfer object must be treated slightly
231 * differently depending on which type of I/O you are performing with it.
233 * This is represented by the public libusb_transfer structure type.
235 * \section asynctrf Asynchronous transfers
237 * We can view asynchronous I/O as a 5 step process:
238 * -# <b>Allocation</b>: allocate a libusb_transfer
239 * -# <b>Filling</b>: populate the libusb_transfer instance with information
240 * about the transfer you wish to perform
241 * -# <b>Submission</b>: ask libusbx to submit the transfer
242 * -# <b>Completion handling</b>: examine transfer results in the
243 * libusb_transfer structure
244 * -# <b>Deallocation</b>: clean up resources
247 * \subsection asyncalloc Allocation
249 * This step involves allocating memory for a USB transfer. This is the
250 * generic transfer object mentioned above. At this stage, the transfer
251 * is "blank" with no details about what type of I/O it will be used for.
253 * Allocation is done with the libusb_alloc_transfer() function. You must use
254 * this function rather than allocating your own transfers.
256 * \subsection asyncfill Filling
258 * This step is where you take a previously allocated transfer and fill it
259 * with information to determine the message type and direction, data buffer,
260 * callback function, etc.
262 * You can either fill the required fields yourself or you can use the
263 * helper functions: libusb_fill_control_transfer(), libusb_fill_bulk_transfer()
264 * and libusb_fill_interrupt_transfer().
266 * \subsection asyncsubmit Submission
268 * When you have allocated a transfer and filled it, you can submit it using
269 * libusb_submit_transfer(). This function returns immediately but can be
270 * regarded as firing off the I/O request in the background.
272 * \subsection asynccomplete Completion handling
274 * After a transfer has been submitted, one of four things can happen to it:
276 * - The transfer completes (i.e. some data was transferred)
277 * - The transfer has a timeout and the timeout expires before all data is
279 * - The transfer fails due to an error
280 * - The transfer is cancelled
282 * Each of these will cause the user-specified transfer callback function to
283 * be invoked. It is up to the callback function to determine which of the
284 * above actually happened and to act accordingly.
286 * The user-specified callback is passed a pointer to the libusb_transfer
287 * structure which was used to setup and submit the transfer. At completion
288 * time, libusbx has populated this structure with results of the transfer:
289 * success or failure reason, number of bytes of data transferred, etc. See
290 * the libusb_transfer structure documentation for more information.
292 * \subsection Deallocation
294 * When a transfer has completed (i.e. the callback function has been invoked),
295 * you are advised to free the transfer (unless you wish to resubmit it, see
296 * below). Transfers are deallocated with libusb_free_transfer().
298 * It is undefined behaviour to free a transfer which has not completed.
300 * \section asyncresubmit Resubmission
302 * You may be wondering why allocation, filling, and submission are all
303 * separated above where they could reasonably be combined into a single
306 * The reason for separation is to allow you to resubmit transfers without
307 * having to allocate new ones every time. This is especially useful for
308 * common situations dealing with interrupt endpoints - you allocate one
309 * transfer, fill and submit it, and when it returns with results you just
310 * resubmit it for the next interrupt.
312 * \section asynccancel Cancellation
314 * Another advantage of using the asynchronous interface is that you have
315 * the ability to cancel transfers which have not yet completed. This is
316 * done by calling the libusb_cancel_transfer() function.
318 * libusb_cancel_transfer() is asynchronous/non-blocking in itself. When the
319 * cancellation actually completes, the transfer's callback function will
320 * be invoked, and the callback function should check the transfer status to
321 * determine that it was cancelled.
323 * Freeing the transfer after it has been cancelled but before cancellation
324 * has completed will result in undefined behaviour.
326 * When a transfer is cancelled, some of the data may have been transferred.
327 * libusbx will communicate this to you in the transfer callback. Do not assume
328 * that no data was transferred.
330 * \section bulk_overflows Overflows on device-to-host bulk/interrupt endpoints
332 * If your device does not have predictable transfer sizes (or it misbehaves),
333 * your application may submit a request for data on an IN endpoint which is
334 * smaller than the data that the device wishes to send. In some circumstances
335 * this will cause an overflow, which is a nasty condition to deal with. See
336 * the \ref packetoverflow page for discussion.
338 * \section asyncctrl Considerations for control transfers
340 * The <tt>libusb_transfer</tt> structure is generic and hence does not
341 * include specific fields for the control-specific setup packet structure.
343 * In order to perform a control transfer, you must place the 8-byte setup
344 * packet at the start of the data buffer. To simplify this, you could
345 * cast the buffer pointer to type struct libusb_control_setup, or you can
346 * use the helper function libusb_fill_control_setup().
348 * The wLength field placed in the setup packet must be the length you would
349 * expect to be sent in the setup packet: the length of the payload that
350 * follows (or the expected maximum number of bytes to receive). However,
351 * the length field of the libusb_transfer object must be the length of
352 * the data buffer - i.e. it should be wLength <em>plus</em> the size of
353 * the setup packet (LIBUSB_CONTROL_SETUP_SIZE).
355 * If you use the helper functions, this is simplified for you:
356 * -# Allocate a buffer of size LIBUSB_CONTROL_SETUP_SIZE plus the size of the
357 * data you are sending/requesting.
358 * -# Call libusb_fill_control_setup() on the data buffer, using the transfer
359 * request size as the wLength value (i.e. do not include the extra space you
360 * allocated for the control setup).
361 * -# If this is a host-to-device transfer, place the data to be transferred
362 * in the data buffer, starting at offset LIBUSB_CONTROL_SETUP_SIZE.
363 * -# Call libusb_fill_control_transfer() to associate the data buffer with
364 * the transfer (and to set the remaining details such as callback and timeout).
365 * - Note that there is no parameter to set the length field of the transfer.
366 * The length is automatically inferred from the wLength field of the setup
368 * -# Submit the transfer.
370 * The multi-byte control setup fields (wValue, wIndex and wLength) must
371 * be given in little-endian byte order (the endianness of the USB bus).
372 * Endianness conversion is transparently handled by
373 * libusb_fill_control_setup() which is documented to accept host-endian
376 * Further considerations are needed when handling transfer completion in
377 * your callback function:
378 * - As you might expect, the setup packet will still be sitting at the start
379 * of the data buffer.
380 * - If this was a device-to-host transfer, the received data will be sitting
381 * at offset LIBUSB_CONTROL_SETUP_SIZE into the buffer.
382 * - The actual_length field of the transfer structure is relative to the
383 * wLength of the setup packet, rather than the size of the data buffer. So,
384 * if your wLength was 4, your transfer's <tt>length</tt> was 12, then you
385 * should expect an <tt>actual_length</tt> of 4 to indicate that the data was
386 * transferred in entirity.
388 * To simplify parsing of setup packets and obtaining the data from the
389 * correct offset, you may wish to use the libusb_control_transfer_get_data()
390 * and libusb_control_transfer_get_setup() functions within your transfer
393 * Even though control endpoints do not halt, a completed control transfer
394 * may have a LIBUSB_TRANSFER_STALL status code. This indicates the control
395 * request was not supported.
397 * \section asyncintr Considerations for interrupt transfers
399 * All interrupt transfers are performed using the polling interval presented
400 * by the bInterval value of the endpoint descriptor.
402 * \section asynciso Considerations for isochronous transfers
404 * Isochronous transfers are more complicated than transfers to
405 * non-isochronous endpoints.
407 * To perform I/O to an isochronous endpoint, allocate the transfer by calling
408 * libusb_alloc_transfer() with an appropriate number of isochronous packets.
410 * During filling, set \ref libusb_transfer::type "type" to
411 * \ref libusb_transfer_type::LIBUSB_TRANSFER_TYPE_ISOCHRONOUS
412 * "LIBUSB_TRANSFER_TYPE_ISOCHRONOUS", and set
413 * \ref libusb_transfer::num_iso_packets "num_iso_packets" to a value less than
414 * or equal to the number of packets you requested during allocation.
415 * libusb_alloc_transfer() does not set either of these fields for you, given
416 * that you might not even use the transfer on an isochronous endpoint.
418 * Next, populate the length field for the first num_iso_packets entries in
419 * the \ref libusb_transfer::iso_packet_desc "iso_packet_desc" array. Section
420 * 5.6.3 of the USB2 specifications describe how the maximum isochronous
421 * packet length is determined by the wMaxPacketSize field in the endpoint
423 * Two functions can help you here:
425 * - libusb_get_max_iso_packet_size() is an easy way to determine the max
426 * packet size for an isochronous endpoint. Note that the maximum packet
427 * size is actually the maximum number of bytes that can be transmitted in
428 * a single microframe, therefore this function multiplies the maximum number
429 * of bytes per transaction by the number of transaction opportunities per
431 * - libusb_set_iso_packet_lengths() assigns the same length to all packets
432 * within a transfer, which is usually what you want.
434 * For outgoing transfers, you'll obviously fill the buffer and populate the
435 * packet descriptors in hope that all the data gets transferred. For incoming
436 * transfers, you must ensure the buffer has sufficient capacity for
437 * the situation where all packets transfer the full amount of requested data.
439 * Completion handling requires some extra consideration. The
440 * \ref libusb_transfer::actual_length "actual_length" field of the transfer
441 * is meaningless and should not be examined; instead you must refer to the
442 * \ref libusb_iso_packet_descriptor::actual_length "actual_length" field of
443 * each individual packet.
445 * The \ref libusb_transfer::status "status" field of the transfer is also a
447 * - If the packets were submitted and the isochronous data microframes
448 * completed normally, status will have value
449 * \ref libusb_transfer_status::LIBUSB_TRANSFER_COMPLETED
450 * "LIBUSB_TRANSFER_COMPLETED". Note that bus errors and software-incurred
451 * delays are not counted as transfer errors; the transfer.status field may
452 * indicate COMPLETED even if some or all of the packets failed. Refer to
453 * the \ref libusb_iso_packet_descriptor::status "status" field of each
454 * individual packet to determine packet failures.
455 * - The status field will have value
456 * \ref libusb_transfer_status::LIBUSB_TRANSFER_ERROR
457 * "LIBUSB_TRANSFER_ERROR" only when serious errors were encountered.
458 * - Other transfer status codes occur with normal behaviour.
460 * The data for each packet will be found at an offset into the buffer that
461 * can be calculated as if each prior packet completed in full. The
462 * libusb_get_iso_packet_buffer() and libusb_get_iso_packet_buffer_simple()
463 * functions may help you here.
465 * \section asyncmem Memory caveats
467 * In most circumstances, it is not safe to use stack memory for transfer
468 * buffers. This is because the function that fired off the asynchronous
469 * transfer may return before libusbx has finished using the buffer, and when
470 * the function returns it's stack gets destroyed. This is true for both
471 * host-to-device and device-to-host transfers.
473 * The only case in which it is safe to use stack memory is where you can
474 * guarantee that the function owning the stack space for the buffer does not
475 * return until after the transfer's callback function has completed. In every
476 * other case, you need to use heap memory instead.
478 * \section asyncflags Fine control
480 * Through using this asynchronous interface, you may find yourself repeating
481 * a few simple operations many times. You can apply a bitwise OR of certain
482 * flags to a transfer to simplify certain things:
483 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_SHORT_NOT_OK
484 * "LIBUSB_TRANSFER_SHORT_NOT_OK" results in transfers which transferred
485 * less than the requested amount of data being marked with status
486 * \ref libusb_transfer_status::LIBUSB_TRANSFER_ERROR "LIBUSB_TRANSFER_ERROR"
487 * (they would normally be regarded as COMPLETED)
488 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_BUFFER
489 * "LIBUSB_TRANSFER_FREE_BUFFER" allows you to ask libusbx to free the transfer
490 * buffer when freeing the transfer.
491 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_TRANSFER
492 * "LIBUSB_TRANSFER_FREE_TRANSFER" causes libusbx to automatically free the
493 * transfer after the transfer callback returns.
495 * \section asyncevent Event handling
497 * In accordance of the aim of being a lightweight library, libusbx does not
498 * create threads internally. This means that libusbx code does not execute
499 * at any time other than when your application is calling a libusbx function.
500 * However, an asynchronous model requires that libusbx perform work at various
501 * points in time - namely processing the results of previously-submitted
502 * transfers and invoking the user-supplied callback function.
504 * This gives rise to the libusb_handle_events() function which your
505 * application must call into when libusbx has work do to. This gives libusbx
506 * the opportunity to reap pending transfers, invoke callbacks, etc.
508 * The first issue to discuss here is how your application can figure out
509 * when libusbx has work to do. In fact, there are two naive options which
510 * do not actually require your application to know this:
511 * -# Periodically call libusb_handle_events() in non-blocking mode at fixed
512 * short intervals from your main loop
513 * -# Repeatedly call libusb_handle_events() in blocking mode from a dedicated
516 * The first option is plainly not very nice, and will cause unnecessary
517 * CPU wakeups leading to increased power usage and decreased battery life.
518 * The second option is not very nice either, but may be the nicest option
519 * available to you if the "proper" approach can not be applied to your
520 * application (read on...).
522 * The recommended option is to integrate libusbx with your application main
523 * event loop. libusbx exposes a set of file descriptors which allow you to do
524 * this. Your main loop is probably already calling poll() or select() or a
525 * variant on a set of file descriptors for other event sources (e.g. keyboard
526 * button presses, mouse movements, network sockets, etc). You then add
527 * libusbx's file descriptors to your poll()/select() calls, and when activity
528 * is detected on such descriptors you know it is time to call
529 * libusb_handle_events().
531 * There is one final event handling complication. libusbx supports
532 * asynchronous transfers which time out after a specified time period, and
533 * this requires that libusbx is called into at or after the timeout so that
534 * the timeout can be handled. So, in addition to considering libusbx's file
535 * descriptors in your main event loop, you must also consider that libusbx
536 * sometimes needs to be called into at fixed points in time even when there
537 * is no file descriptor activity.
539 * For the details on retrieving the set of file descriptors and determining
540 * the next timeout, see the \ref poll "polling and timing" API documentation.
544 * @defgroup poll Polling and timing
546 * This page documents libusbx's functions for polling events and timing.
547 * These functions are only necessary for users of the
548 * \ref asyncio "asynchronous API". If you are only using the simpler
549 * \ref syncio "synchronous API" then you do not need to ever call these
552 * The justification for the functionality described here has already been
553 * discussed in the \ref asyncevent "event handling" section of the
554 * asynchronous API documentation. In summary, libusbx does not create internal
555 * threads for event processing and hence relies on your application calling
556 * into libusbx at certain points in time so that pending events can be handled.
557 * In order to know precisely when libusbx needs to be called into, libusbx
558 * offers you a set of pollable file descriptors and information about when
559 * the next timeout expires.
561 * If you are using the asynchronous I/O API, you must take one of the two
562 * following options, otherwise your I/O will not complete.
564 * \section pollsimple The simple option
566 * If your application revolves solely around libusbx and does not need to
567 * handle other event sources, you can have a program structure as follows:
569 // initialize libusbx
570 // find and open device
571 // maybe fire off some initial async I/O
573 while (user_has_not_requested_exit)
574 libusb_handle_events(ctx);
579 * With such a simple main loop, you do not have to worry about managing
580 * sets of file descriptors or handling timeouts. libusb_handle_events() will
581 * handle those details internally.
583 * \section pollmain The more advanced option
585 * \note This functionality is currently only available on Unix-like platforms.
586 * On Windows, libusb_get_pollfds() simply returns NULL. Exposing event sources
587 * on Windows will require some further thought and design.
589 * In more advanced applications, you will already have a main loop which
590 * is monitoring other event sources: network sockets, X11 events, mouse
591 * movements, etc. Through exposing a set of file descriptors, libusbx is
592 * designed to cleanly integrate into such main loops.
594 * In addition to polling file descriptors for the other event sources, you
595 * take a set of file descriptors from libusbx and monitor those too. When you
596 * detect activity on libusbx's file descriptors, you call
597 * libusb_handle_events_timeout() in non-blocking mode.
599 * What's more, libusbx may also need to handle events at specific moments in
600 * time. No file descriptor activity is generated at these times, so your
601 * own application needs to be continually aware of when the next one of these
602 * moments occurs (through calling libusb_get_next_timeout()), and then it
603 * needs to call libusb_handle_events_timeout() in non-blocking mode when
604 * these moments occur. This means that you need to adjust your
605 * poll()/select() timeout accordingly.
607 * libusbx provides you with a set of file descriptors to poll and expects you
608 * to poll all of them, treating them as a single entity. The meaning of each
609 * file descriptor in the set is an internal implementation detail,
610 * platform-dependent and may vary from release to release. Don't try and
611 * interpret the meaning of the file descriptors, just do as libusbx indicates,
612 * polling all of them at once.
614 * In pseudo-code, you want something that looks like:
616 // initialise libusbx
618 libusb_get_pollfds(ctx)
619 while (user has not requested application exit) {
620 libusb_get_next_timeout(ctx);
621 poll(on libusbx file descriptors plus any other event sources of interest,
622 using a timeout no larger than the value libusbx just suggested)
623 if (poll() indicated activity on libusbx file descriptors)
624 libusb_handle_events_timeout(ctx, &zero_tv);
625 if (time has elapsed to or beyond the libusbx timeout)
626 libusb_handle_events_timeout(ctx, &zero_tv);
627 // handle events from other sources here
633 * \subsection polltime Notes on time-based events
635 * The above complication with having to track time and call into libusbx at
636 * specific moments is a bit of a headache. For maximum compatibility, you do
637 * need to write your main loop as above, but you may decide that you can
638 * restrict the supported platforms of your application and get away with
639 * a more simplistic scheme.
641 * These time-based event complications are \b not required on the following
644 * - Linux, provided that the following version requirements are satisfied:
645 * - Linux v2.6.27 or newer, compiled with timerfd support
646 * - glibc v2.9 or newer
647 * - libusbx v1.0.5 or newer
649 * Under these configurations, libusb_get_next_timeout() will \em always return
650 * 0, so your main loop can be simplified to:
652 // initialise libusbx
654 libusb_get_pollfds(ctx)
655 while (user has not requested application exit) {
656 poll(on libusbx file descriptors plus any other event sources of interest,
657 using any timeout that you like)
658 if (poll() indicated activity on libusbx file descriptors)
659 libusb_handle_events_timeout(ctx, &zero_tv);
660 // handle events from other sources here
666 * Do remember that if you simplify your main loop to the above, you will
667 * lose compatibility with some platforms (including legacy Linux platforms,
668 * and <em>any future platforms supported by libusbx which may have time-based
669 * event requirements</em>). The resultant problems will likely appear as
670 * strange bugs in your application.
672 * You can use the libusb_pollfds_handle_timeouts() function to do a runtime
673 * check to see if it is safe to ignore the time-based event complications.
674 * If your application has taken the shortcut of ignoring libusbx's next timeout
675 * in your main loop, then you are advised to check the return value of
676 * libusb_pollfds_handle_timeouts() during application startup, and to abort
677 * if the platform does suffer from these timing complications.
679 * \subsection fdsetchange Changes in the file descriptor set
681 * The set of file descriptors that libusbx uses as event sources may change
682 * during the life of your application. Rather than having to repeatedly
683 * call libusb_get_pollfds(), you can set up notification functions for when
684 * the file descriptor set changes using libusb_set_pollfd_notifiers().
686 * \subsection mtissues Multi-threaded considerations
688 * Unfortunately, the situation is complicated further when multiple threads
689 * come into play. If two threads are monitoring the same file descriptors,
690 * the fact that only one thread will be woken up when an event occurs causes
693 * The events lock, event waiters lock, and libusb_handle_events_locked()
694 * entities are added to solve these problems. You do not need to be concerned
695 * with these entities otherwise.
697 * See the extra documentation: \ref mtasync
700 /** \page mtasync Multi-threaded applications and asynchronous I/O
702 * libusbx is a thread-safe library, but extra considerations must be applied
703 * to applications which interact with libusbx from multiple threads.
705 * The underlying issue that must be addressed is that all libusbx I/O
706 * revolves around monitoring file descriptors through the poll()/select()
707 * system calls. This is directly exposed at the
708 * \ref asyncio "asynchronous interface" but it is important to note that the
709 * \ref syncio "synchronous interface" is implemented on top of the
710 * asynchonrous interface, therefore the same considerations apply.
712 * The issue is that if two or more threads are concurrently calling poll()
713 * or select() on libusbx's file descriptors then only one of those threads
714 * will be woken up when an event arrives. The others will be completely
715 * oblivious that anything has happened.
717 * Consider the following pseudo-code, which submits an asynchronous transfer
718 * then waits for its completion. This style is one way you could implement a
719 * synchronous interface on top of the asynchronous interface (and libusbx
720 * does something similar, albeit more advanced due to the complications
721 * explained on this page).
724 void cb(struct libusb_transfer *transfer)
726 int *completed = transfer->user_data;
731 struct libusb_transfer *transfer;
732 unsigned char buffer[LIBUSB_CONTROL_SETUP_SIZE];
735 transfer = libusb_alloc_transfer(0);
736 libusb_fill_control_setup(buffer,
737 LIBUSB_REQUEST_TYPE_VENDOR | LIBUSB_ENDPOINT_OUT, 0x04, 0x01, 0, 0);
738 libusb_fill_control_transfer(transfer, dev, buffer, cb, &completed, 1000);
739 libusb_submit_transfer(transfer);
742 poll(libusbx file descriptors, 120*1000);
743 if (poll indicates activity)
744 libusb_handle_events_timeout(ctx, &zero_tv);
746 printf("completed!");
751 * Here we are <em>serializing</em> completion of an asynchronous event
752 * against a condition - the condition being completion of a specific transfer.
753 * The poll() loop has a long timeout to minimize CPU usage during situations
754 * when nothing is happening (it could reasonably be unlimited).
756 * If this is the only thread that is polling libusbx's file descriptors, there
757 * is no problem: there is no danger that another thread will swallow up the
758 * event that we are interested in. On the other hand, if there is another
759 * thread polling the same descriptors, there is a chance that it will receive
760 * the event that we were interested in. In this situation, <tt>myfunc()</tt>
761 * will only realise that the transfer has completed on the next iteration of
762 * the loop, <em>up to 120 seconds later.</em> Clearly a two-minute delay is
763 * undesirable, and don't even think about using short timeouts to circumvent
766 * The solution here is to ensure that no two threads are ever polling the
767 * file descriptors at the same time. A naive implementation of this would
768 * impact the capabilities of the library, so libusbx offers the scheme
769 * documented below to ensure no loss of functionality.
771 * Before we go any further, it is worth mentioning that all libusb-wrapped
772 * event handling procedures fully adhere to the scheme documented below.
773 * This includes libusb_handle_events() and its variants, and all the
774 * synchronous I/O functions - libusbx hides this headache from you.
776 * \section Using libusb_handle_events() from multiple threads
778 * Even when only using libusb_handle_events() and synchronous I/O functions,
779 * you can still have a race condition. You might be tempted to solve the
780 * above with libusb_handle_events() like so:
783 libusb_submit_transfer(transfer);
786 libusb_handle_events(ctx);
788 printf("completed!");
791 * This however has a race between the checking of completed and
792 * libusb_handle_events() acquiring the events lock, so another thread
793 * could have completed the transfer, resulting in this thread hanging
794 * until either a timeout or another event occurs. See also commit
795 * 6696512aade99bb15d6792af90ae329af270eba6 which fixes this in the
796 * synchronous API implementation of libusb.
798 * Fixing this race requires checking the variable completed only after
799 * taking the event lock, which defeats the concept of just calling
800 * libusb_handle_events() without worrying about locking. This is why
801 * libusb-1.0.9 introduces the new libusb_handle_events_timeout_completed()
802 * and libusb_handle_events_completed() functions, which handles doing the
803 * completion check for you after they have acquired the lock:
806 libusb_submit_transfer(transfer);
809 libusb_handle_events_completed(ctx, &completed);
811 printf("completed!");
814 * This nicely fixes the race in our example. Note that if all you want to
815 * do is submit a single transfer and wait for its completion, then using
816 * one of the synchronous I/O functions is much easier.
818 * \section eventlock The events lock
820 * The problem is when we consider the fact that libusbx exposes file
821 * descriptors to allow for you to integrate asynchronous USB I/O into
822 * existing main loops, effectively allowing you to do some work behind
823 * libusbx's back. If you do take libusbx's file descriptors and pass them to
824 * poll()/select() yourself, you need to be aware of the associated issues.
826 * The first concept to be introduced is the events lock. The events lock
827 * is used to serialize threads that want to handle events, such that only
828 * one thread is handling events at any one time.
830 * You must take the events lock before polling libusbx file descriptors,
831 * using libusb_lock_events(). You must release the lock as soon as you have
832 * aborted your poll()/select() loop, using libusb_unlock_events().
834 * \section threadwait Letting other threads do the work for you
836 * Although the events lock is a critical part of the solution, it is not
837 * enough on it's own. You might wonder if the following is sufficient...
839 libusb_lock_events(ctx);
841 poll(libusbx file descriptors, 120*1000);
842 if (poll indicates activity)
843 libusb_handle_events_timeout(ctx, &zero_tv);
845 libusb_unlock_events(ctx);
847 * ...and the answer is that it is not. This is because the transfer in the
848 * code shown above may take a long time (say 30 seconds) to complete, and
849 * the lock is not released until the transfer is completed.
851 * Another thread with similar code that wants to do event handling may be
852 * working with a transfer that completes after a few milliseconds. Despite
853 * having such a quick completion time, the other thread cannot check that
854 * status of its transfer until the code above has finished (30 seconds later)
855 * due to contention on the lock.
857 * To solve this, libusbx offers you a mechanism to determine when another
858 * thread is handling events. It also offers a mechanism to block your thread
859 * until the event handling thread has completed an event (and this mechanism
860 * does not involve polling of file descriptors).
862 * After determining that another thread is currently handling events, you
863 * obtain the <em>event waiters</em> lock using libusb_lock_event_waiters().
864 * You then re-check that some other thread is still handling events, and if
865 * so, you call libusb_wait_for_event().
867 * libusb_wait_for_event() puts your application to sleep until an event
868 * occurs, or until a thread releases the events lock. When either of these
869 * things happen, your thread is woken up, and should re-check the condition
870 * it was waiting on. It should also re-check that another thread is handling
871 * events, and if not, it should start handling events itself.
873 * This looks like the following, as pseudo-code:
876 if (libusb_try_lock_events(ctx) == 0) {
877 // we obtained the event lock: do our own event handling
879 if (!libusb_event_handling_ok(ctx)) {
880 libusb_unlock_events(ctx);
883 poll(libusbx file descriptors, 120*1000);
884 if (poll indicates activity)
885 libusb_handle_events_locked(ctx, 0);
887 libusb_unlock_events(ctx);
889 // another thread is doing event handling. wait for it to signal us that
890 // an event has completed
891 libusb_lock_event_waiters(ctx);
894 // now that we have the event waiters lock, double check that another
895 // thread is still handling events for us. (it may have ceased handling
896 // events in the time it took us to reach this point)
897 if (!libusb_event_handler_active(ctx)) {
898 // whoever was handling events is no longer doing so, try again
899 libusb_unlock_event_waiters(ctx);
903 libusb_wait_for_event(ctx, NULL);
905 libusb_unlock_event_waiters(ctx);
907 printf("completed!\n");
910 * A naive look at the above code may suggest that this can only support
911 * one event waiter (hence a total of 2 competing threads, the other doing
912 * event handling), because the event waiter seems to have taken the event
913 * waiters lock while waiting for an event. However, the system does support
914 * multiple event waiters, because libusb_wait_for_event() actually drops
915 * the lock while waiting, and reaquires it before continuing.
917 * We have now implemented code which can dynamically handle situations where
918 * nobody is handling events (so we should do it ourselves), and it can also
919 * handle situations where another thread is doing event handling (so we can
920 * piggyback onto them). It is also equipped to handle a combination of
921 * the two, for example, another thread is doing event handling, but for
922 * whatever reason it stops doing so before our condition is met, so we take
923 * over the event handling.
925 * Four functions were introduced in the above pseudo-code. Their importance
926 * should be apparent from the code shown above.
927 * -# libusb_try_lock_events() is a non-blocking function which attempts
928 * to acquire the events lock but returns a failure code if it is contended.
929 * -# libusb_event_handling_ok() checks that libusbx is still happy for your
930 * thread to be performing event handling. Sometimes, libusbx needs to
931 * interrupt the event handler, and this is how you can check if you have
932 * been interrupted. If this function returns 0, the correct behaviour is
933 * for you to give up the event handling lock, and then to repeat the cycle.
934 * The following libusb_try_lock_events() will fail, so you will become an
935 * events waiter. For more information on this, read \ref fullstory below.
936 * -# libusb_handle_events_locked() is a variant of
937 * libusb_handle_events_timeout() that you can call while holding the
938 * events lock. libusb_handle_events_timeout() itself implements similar
939 * logic to the above, so be sure not to call it when you are
940 * "working behind libusbx's back", as is the case here.
941 * -# libusb_event_handler_active() determines if someone is currently
942 * holding the events lock
944 * You might be wondering why there is no function to wake up all threads
945 * blocked on libusb_wait_for_event(). This is because libusbx can do this
946 * internally: it will wake up all such threads when someone calls
947 * libusb_unlock_events() or when a transfer completes (at the point after its
948 * callback has returned).
950 * \subsection fullstory The full story
952 * The above explanation should be enough to get you going, but if you're
953 * really thinking through the issues then you may be left with some more
954 * questions regarding libusbx's internals. If you're curious, read on, and if
955 * not, skip to the next section to avoid confusing yourself!
957 * The immediate question that may spring to mind is: what if one thread
958 * modifies the set of file descriptors that need to be polled while another
959 * thread is doing event handling?
961 * There are 2 situations in which this may happen.
962 * -# libusb_open() will add another file descriptor to the poll set,
963 * therefore it is desirable to interrupt the event handler so that it
964 * restarts, picking up the new descriptor.
965 * -# libusb_close() will remove a file descriptor from the poll set. There
966 * are all kinds of race conditions that could arise here, so it is
967 * important that nobody is doing event handling at this time.
969 * libusbx handles these issues internally, so application developers do not
970 * have to stop their event handlers while opening/closing devices. Here's how
971 * it works, focusing on the libusb_close() situation first:
973 * -# During initialization, libusbx opens an internal pipe, and it adds the read
974 * end of this pipe to the set of file descriptors to be polled.
975 * -# During libusb_close(), libusbx writes some dummy data on this control pipe.
976 * This immediately interrupts the event handler. libusbx also records
977 * internally that it is trying to interrupt event handlers for this
978 * high-priority event.
979 * -# At this point, some of the functions described above start behaving
981 * - libusb_event_handling_ok() starts returning 1, indicating that it is NOT
982 * OK for event handling to continue.
983 * - libusb_try_lock_events() starts returning 1, indicating that another
984 * thread holds the event handling lock, even if the lock is uncontended.
985 * - libusb_event_handler_active() starts returning 1, indicating that
986 * another thread is doing event handling, even if that is not true.
987 * -# The above changes in behaviour result in the event handler stopping and
988 * giving up the events lock very quickly, giving the high-priority
989 * libusb_close() operation a "free ride" to acquire the events lock. All
990 * threads that are competing to do event handling become event waiters.
991 * -# With the events lock held inside libusb_close(), libusbx can safely remove
992 * a file descriptor from the poll set, in the safety of knowledge that
993 * nobody is polling those descriptors or trying to access the poll set.
994 * -# After obtaining the events lock, the close operation completes very
995 * quickly (usually a matter of milliseconds) and then immediately releases
997 * -# At the same time, the behaviour of libusb_event_handling_ok() and friends
998 * reverts to the original, documented behaviour.
999 * -# The release of the events lock causes the threads that are waiting for
1000 * events to be woken up and to start competing to become event handlers
1001 * again. One of them will succeed; it will then re-obtain the list of poll
1002 * descriptors, and USB I/O will then continue as normal.
1004 * libusb_open() is similar, and is actually a more simplistic case. Upon a
1005 * call to libusb_open():
1007 * -# The device is opened and a file descriptor is added to the poll set.
1008 * -# libusbx sends some dummy data on the control pipe, and records that it
1009 * is trying to modify the poll descriptor set.
1010 * -# The event handler is interrupted, and the same behaviour change as for
1011 * libusb_close() takes effect, causing all event handling threads to become
1013 * -# The libusb_open() implementation takes its free ride to the events lock.
1014 * -# Happy that it has successfully paused the events handler, libusb_open()
1015 * releases the events lock.
1016 * -# The event waiter threads are all woken up and compete to become event
1017 * handlers again. The one that succeeds will obtain the list of poll
1018 * descriptors again, which will include the addition of the new device.
1020 * \subsection concl Closing remarks
1022 * The above may seem a little complicated, but hopefully I have made it clear
1023 * why such complications are necessary. Also, do not forget that this only
1024 * applies to applications that take libusbx's file descriptors and integrate
1025 * them into their own polling loops.
1027 * You may decide that it is OK for your multi-threaded application to ignore
1028 * some of the rules and locks detailed above, because you don't think that
1029 * two threads can ever be polling the descriptors at the same time. If that
1030 * is the case, then that's good news for you because you don't have to worry.
1031 * But be careful here; remember that the synchronous I/O functions do event
1032 * handling internally. If you have one thread doing event handling in a loop
1033 * (without implementing the rules and locking semantics documented above)
1034 * and another trying to send a synchronous USB transfer, you will end up with
1035 * two threads monitoring the same descriptors, and the above-described
1036 * undesirable behaviour occuring. The solution is for your polling thread to
1037 * play by the rules; the synchronous I/O functions do so, and this will result
1038 * in them getting along in perfect harmony.
1040 * If you do have a dedicated thread doing event handling, it is perfectly
1041 * legal for it to take the event handling lock for long periods of time. Any
1042 * synchronous I/O functions you call from other threads will transparently
1043 * fall back to the "event waiters" mechanism detailed above. The only
1044 * consideration that your event handling thread must apply is the one related
1045 * to libusb_event_handling_ok(): you must call this before every poll(), and
1046 * give up the events lock if instructed.
1049 int usbi_io_init(struct libusb_context *ctx)
1053 usbi_mutex_init(&ctx->flying_transfers_lock, NULL);
1054 usbi_mutex_init(&ctx->pollfds_lock, NULL);
1055 usbi_mutex_init(&ctx->pollfd_modify_lock, NULL);
1056 usbi_mutex_init_recursive(&ctx->events_lock, NULL);
1057 usbi_mutex_init(&ctx->event_waiters_lock, NULL);
1058 usbi_cond_init(&ctx->event_waiters_cond, NULL);
1059 list_init(&ctx->flying_transfers);
1060 list_init(&ctx->pollfds);
1062 /* FIXME should use an eventfd on kernels that support it */
1063 r = usbi_pipe(ctx->ctrl_pipe);
1065 r = LIBUSB_ERROR_OTHER;
1069 r = usbi_add_pollfd(ctx, ctx->ctrl_pipe[0], POLLIN);
1071 goto err_close_pipe;
1073 #ifdef USBI_TIMERFD_AVAILABLE
1074 ctx->timerfd = timerfd_create(usbi_backend->get_timerfd_clockid(),
1076 if (ctx->timerfd >= 0) {
1077 usbi_dbg("using timerfd for timeouts");
1078 r = usbi_add_pollfd(ctx, ctx->timerfd, POLLIN);
1080 usbi_remove_pollfd(ctx, ctx->ctrl_pipe[0]);
1081 close(ctx->timerfd);
1082 goto err_close_pipe;
1085 usbi_dbg("timerfd not available (code %d error %d)", ctx->timerfd, errno);
1093 usbi_close(ctx->ctrl_pipe[0]);
1094 usbi_close(ctx->ctrl_pipe[1]);
1096 usbi_mutex_destroy(&ctx->flying_transfers_lock);
1097 usbi_mutex_destroy(&ctx->pollfds_lock);
1098 usbi_mutex_destroy(&ctx->pollfd_modify_lock);
1099 usbi_mutex_destroy(&ctx->events_lock);
1100 usbi_mutex_destroy(&ctx->event_waiters_lock);
1101 usbi_cond_destroy(&ctx->event_waiters_cond);
1105 void usbi_io_exit(struct libusb_context *ctx)
1107 usbi_remove_pollfd(ctx, ctx->ctrl_pipe[0]);
1108 usbi_close(ctx->ctrl_pipe[0]);
1109 usbi_close(ctx->ctrl_pipe[1]);
1110 #ifdef USBI_TIMERFD_AVAILABLE
1111 if (usbi_using_timerfd(ctx)) {
1112 usbi_remove_pollfd(ctx, ctx->timerfd);
1113 close(ctx->timerfd);
1116 usbi_mutex_destroy(&ctx->flying_transfers_lock);
1117 usbi_mutex_destroy(&ctx->pollfds_lock);
1118 usbi_mutex_destroy(&ctx->pollfd_modify_lock);
1119 usbi_mutex_destroy(&ctx->events_lock);
1120 usbi_mutex_destroy(&ctx->event_waiters_lock);
1121 usbi_cond_destroy(&ctx->event_waiters_cond);
1124 static int calculate_timeout(struct usbi_transfer *transfer)
1127 struct timespec current_time;
1128 unsigned int timeout =
1129 USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout;
1134 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, ¤t_time);
1136 usbi_err(ITRANSFER_CTX(transfer),
1137 "failed to read monotonic clock, errno=%d", errno);
1141 current_time.tv_sec += timeout / 1000;
1142 current_time.tv_nsec += (timeout % 1000) * 1000000;
1144 if (current_time.tv_nsec > 1000000000) {
1145 current_time.tv_nsec -= 1000000000;
1146 current_time.tv_sec++;
1149 TIMESPEC_TO_TIMEVAL(&transfer->timeout, ¤t_time);
1153 /* add a transfer to the (timeout-sorted) active transfers list.
1154 * returns 1 if the transfer has a timeout and it is the timeout next to
1156 static int add_to_flying_list(struct usbi_transfer *transfer)
1158 struct usbi_transfer *cur;
1159 struct timeval *timeout = &transfer->timeout;
1160 struct libusb_context *ctx = ITRANSFER_CTX(transfer);
1164 usbi_mutex_lock(&ctx->flying_transfers_lock);
1166 /* if we have no other flying transfers, start the list with this one */
1167 if (list_empty(&ctx->flying_transfers)) {
1168 list_add(&transfer->list, &ctx->flying_transfers);
1172 /* if we have infinite timeout, append to end of list */
1173 if (!timerisset(timeout)) {
1174 list_add_tail(&transfer->list, &ctx->flying_transfers);
1175 /* first is irrelevant in this case */
1179 /* otherwise, find appropriate place in list */
1180 list_for_each_entry(cur, &ctx->flying_transfers, list, struct usbi_transfer) {
1181 /* find first timeout that occurs after the transfer in question */
1182 struct timeval *cur_tv = &cur->timeout;
1184 if (!timerisset(cur_tv) || (cur_tv->tv_sec > timeout->tv_sec) ||
1185 (cur_tv->tv_sec == timeout->tv_sec &&
1186 cur_tv->tv_usec > timeout->tv_usec)) {
1187 list_add_tail(&transfer->list, &cur->list);
1192 /* first is 0 at this stage (list not empty) */
1194 /* otherwise we need to be inserted at the end */
1195 list_add_tail(&transfer->list, &ctx->flying_transfers);
1197 #ifdef USBI_TIMERFD_AVAILABLE
1198 if (first && usbi_using_timerfd(ctx) && timerisset(timeout)) {
1199 /* if this transfer has the lowest timeout of all active transfers,
1200 * rearm the timerfd with this transfer's timeout */
1201 const struct itimerspec it = { {0, 0},
1202 { timeout->tv_sec, timeout->tv_usec * 1000 } };
1203 usbi_dbg("arm timerfd for timeout in %dms (first in line)",
1204 USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout);
1205 r = timerfd_settime(ctx->timerfd, TFD_TIMER_ABSTIME, &it, NULL);
1207 usbi_warn(ctx, "failed to arm first timerfd (errno %d, it_value = %d:%d)",
1208 errno, it.it_value.tv_sec, it.it_value.tv_nsec);
1209 r = LIBUSB_ERROR_OTHER;
1216 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1220 /** \ingroup asyncio
1221 * Allocate a libusbx transfer with a specified number of isochronous packet
1222 * descriptors. The returned transfer is pre-initialized for you. When the new
1223 * transfer is no longer needed, it should be freed with
1224 * libusb_free_transfer().
1226 * Transfers intended for non-isochronous endpoints (e.g. control, bulk,
1227 * interrupt) should specify an iso_packets count of zero.
1229 * For transfers intended for isochronous endpoints, specify an appropriate
1230 * number of packet descriptors to be allocated as part of the transfer.
1231 * The returned transfer is not specially initialized for isochronous I/O;
1232 * you are still required to set the
1233 * \ref libusb_transfer::num_iso_packets "num_iso_packets" and
1234 * \ref libusb_transfer::type "type" fields accordingly.
1236 * It is safe to allocate a transfer with some isochronous packets and then
1237 * use it on a non-isochronous endpoint. If you do this, ensure that at time
1238 * of submission, num_iso_packets is 0 and that type is set appropriately.
1240 * \param iso_packets number of isochronous packet descriptors to allocate
1241 * \returns a newly allocated transfer, or NULL on error
1244 struct libusb_transfer * LIBUSB_CALL libusb_alloc_transfer(
1247 size_t os_alloc_size = usbi_backend->transfer_priv_size
1248 + (usbi_backend->add_iso_packet_size * iso_packets);
1249 size_t alloc_size = sizeof(struct usbi_transfer)
1250 + sizeof(struct libusb_transfer)
1251 + (sizeof(struct libusb_iso_packet_descriptor) * iso_packets)
1253 struct usbi_transfer *itransfer = calloc(1, alloc_size);
1257 itransfer->num_iso_packets = iso_packets;
1258 usbi_mutex_init(&itransfer->lock, NULL);
1259 return USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1262 /** \ingroup asyncio
1263 * Free a transfer structure. This should be called for all transfers
1264 * allocated with libusb_alloc_transfer().
1266 * If the \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_BUFFER
1267 * "LIBUSB_TRANSFER_FREE_BUFFER" flag is set and the transfer buffer is
1268 * non-NULL, this function will also free the transfer buffer using the
1269 * standard system memory allocator (e.g. free()).
1271 * It is legal to call this function with a NULL transfer. In this case,
1272 * the function will simply return safely.
1274 * It is not legal to free an active transfer (one which has been submitted
1275 * and has not yet completed).
1277 * \param transfer the transfer to free
1279 void API_EXPORTED libusb_free_transfer(struct libusb_transfer *transfer)
1281 struct usbi_transfer *itransfer;
1285 if (transfer->flags & LIBUSB_TRANSFER_FREE_BUFFER && transfer->buffer)
1286 free(transfer->buffer);
1288 itransfer = LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1289 usbi_mutex_destroy(&itransfer->lock);
1293 #ifdef USBI_TIMERFD_AVAILABLE
1294 static int disarm_timerfd(struct libusb_context *ctx)
1296 const struct itimerspec disarm_timer = { { 0, 0 }, { 0, 0 } };
1300 r = timerfd_settime(ctx->timerfd, 0, &disarm_timer, NULL);
1302 return LIBUSB_ERROR_OTHER;
1307 /* iterates through the flying transfers, and rearms the timerfd based on the
1308 * next upcoming timeout.
1309 * must be called with flying_list locked.
1310 * returns 0 if there was no timeout to arm, 1 if the next timeout was armed,
1311 * or a LIBUSB_ERROR code on failure.
1313 static int arm_timerfd_for_next_timeout(struct libusb_context *ctx)
1315 struct usbi_transfer *transfer;
1317 list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
1318 struct timeval *cur_tv = &transfer->timeout;
1320 /* if we've reached transfers of infinite timeout, then we have no
1322 if (!timerisset(cur_tv))
1325 /* act on first transfer that is not already cancelled */
1326 if (!(transfer->flags & USBI_TRANSFER_TIMED_OUT)) {
1328 const struct itimerspec it = { {0, 0},
1329 { cur_tv->tv_sec, cur_tv->tv_usec * 1000 } };
1330 usbi_dbg("next timeout originally %dms", USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout);
1331 r = timerfd_settime(ctx->timerfd, TFD_TIMER_ABSTIME, &it, NULL);
1333 return LIBUSB_ERROR_OTHER;
1339 return disarm_timerfd(ctx);
1342 static int arm_timerfd_for_next_timeout(struct libusb_context *ctx)
1349 /** \ingroup asyncio
1350 * Submit a transfer. This function will fire off the USB transfer and then
1351 * return immediately.
1353 * \param transfer the transfer to submit
1354 * \returns 0 on success
1355 * \returns LIBUSB_ERROR_NO_DEVICE if the device has been disconnected
1356 * \returns LIBUSB_ERROR_BUSY if the transfer has already been submitted.
1357 * \returns LIBUSB_ERROR_NOT_SUPPORTED if the transfer flags are not supported
1358 * by the operating system.
1359 * \returns another LIBUSB_ERROR code on other failure
1361 int API_EXPORTED libusb_submit_transfer(struct libusb_transfer *transfer)
1363 struct libusb_context *ctx = TRANSFER_CTX(transfer);
1364 struct usbi_transfer *itransfer =
1365 LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1369 usbi_mutex_lock(&itransfer->lock);
1370 itransfer->transferred = 0;
1371 itransfer->flags = 0;
1372 r = calculate_timeout(itransfer);
1374 r = LIBUSB_ERROR_OTHER;
1378 r = add_to_flying_list(itransfer);
1381 r = usbi_backend->submit_transfer(itransfer);
1383 usbi_mutex_lock(&ctx->flying_transfers_lock);
1384 list_del(&itransfer->list);
1385 arm_timerfd_for_next_timeout(ctx);
1386 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1390 updated_fds = (itransfer->flags & USBI_TRANSFER_UPDATED_FDS);
1391 usbi_mutex_unlock(&itransfer->lock);
1393 usbi_fd_notification(ctx);
1397 /** \ingroup asyncio
1398 * Asynchronously cancel a previously submitted transfer.
1399 * This function returns immediately, but this does not indicate cancellation
1400 * is complete. Your callback function will be invoked at some later time
1401 * with a transfer status of
1402 * \ref libusb_transfer_status::LIBUSB_TRANSFER_CANCELLED
1403 * "LIBUSB_TRANSFER_CANCELLED."
1405 * \param transfer the transfer to cancel
1406 * \returns 0 on success
1407 * \returns LIBUSB_ERROR_NOT_FOUND if the transfer is already complete or
1409 * \returns a LIBUSB_ERROR code on failure
1411 int API_EXPORTED libusb_cancel_transfer(struct libusb_transfer *transfer)
1413 struct usbi_transfer *itransfer =
1414 LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1418 usbi_mutex_lock(&itransfer->lock);
1419 r = usbi_backend->cancel_transfer(itransfer);
1421 if (r != LIBUSB_ERROR_NOT_FOUND)
1422 usbi_err(TRANSFER_CTX(transfer),
1423 "cancel transfer failed error %d", r);
1425 usbi_dbg("cancel transfer failed error %d", r);
1427 if (r == LIBUSB_ERROR_NO_DEVICE)
1428 itransfer->flags |= USBI_TRANSFER_DEVICE_DISAPPEARED;
1431 itransfer->flags |= USBI_TRANSFER_CANCELLING;
1433 usbi_mutex_unlock(&itransfer->lock);
1437 /* Handle completion of a transfer (completion might be an error condition).
1438 * This will invoke the user-supplied callback function, which may end up
1439 * freeing the transfer. Therefore you cannot use the transfer structure
1440 * after calling this function, and you should free all backend-specific
1441 * data before calling it.
1442 * Do not call this function with the usbi_transfer lock held. User-specified
1443 * callback functions may attempt to directly resubmit the transfer, which
1444 * will attempt to take the lock. */
1445 int usbi_handle_transfer_completion(struct usbi_transfer *itransfer,
1446 enum libusb_transfer_status status)
1448 struct libusb_transfer *transfer =
1449 USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1450 struct libusb_context *ctx = TRANSFER_CTX(transfer);
1454 /* FIXME: could be more intelligent with the timerfd here. we don't need
1455 * to disarm the timerfd if there was no timer running, and we only need
1456 * to rearm the timerfd if the transfer that expired was the one with
1457 * the shortest timeout. */
1459 usbi_mutex_lock(&ctx->flying_transfers_lock);
1460 list_del(&itransfer->list);
1461 if (usbi_using_timerfd(ctx))
1462 r = arm_timerfd_for_next_timeout(ctx);
1463 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1464 if (usbi_using_timerfd(ctx) && (r < 0))
1467 if (status == LIBUSB_TRANSFER_COMPLETED
1468 && transfer->flags & LIBUSB_TRANSFER_SHORT_NOT_OK) {
1469 int rqlen = transfer->length;
1470 if (transfer->type == LIBUSB_TRANSFER_TYPE_CONTROL)
1471 rqlen -= LIBUSB_CONTROL_SETUP_SIZE;
1472 if (rqlen != itransfer->transferred) {
1473 usbi_dbg("interpreting short transfer as error");
1474 status = LIBUSB_TRANSFER_ERROR;
1478 flags = transfer->flags;
1479 transfer->status = status;
1480 transfer->actual_length = itransfer->transferred;
1481 usbi_dbg("transfer %p has callback %p", transfer, transfer->callback);
1482 if (transfer->callback)
1483 transfer->callback(transfer);
1484 /* transfer might have been freed by the above call, do not use from
1486 if (flags & LIBUSB_TRANSFER_FREE_TRANSFER)
1487 libusb_free_transfer(transfer);
1488 usbi_mutex_lock(&ctx->event_waiters_lock);
1489 usbi_cond_broadcast(&ctx->event_waiters_cond);
1490 usbi_mutex_unlock(&ctx->event_waiters_lock);
1494 /* Similar to usbi_handle_transfer_completion() but exclusively for transfers
1495 * that were asynchronously cancelled. The same concerns w.r.t. freeing of
1496 * transfers exist here.
1497 * Do not call this function with the usbi_transfer lock held. User-specified
1498 * callback functions may attempt to directly resubmit the transfer, which
1499 * will attempt to take the lock. */
1500 int usbi_handle_transfer_cancellation(struct usbi_transfer *transfer)
1502 /* if the URB was cancelled due to timeout, report timeout to the user */
1503 if (transfer->flags & USBI_TRANSFER_TIMED_OUT) {
1504 usbi_dbg("detected timeout cancellation");
1505 return usbi_handle_transfer_completion(transfer, LIBUSB_TRANSFER_TIMED_OUT);
1508 /* otherwise its a normal async cancel */
1509 return usbi_handle_transfer_completion(transfer, LIBUSB_TRANSFER_CANCELLED);
1513 * Attempt to acquire the event handling lock. This lock is used to ensure that
1514 * only one thread is monitoring libusbx event sources at any one time.
1516 * You only need to use this lock if you are developing an application
1517 * which calls poll() or select() on libusbx's file descriptors directly.
1518 * If you stick to libusbx's event handling loop functions (e.g.
1519 * libusb_handle_events()) then you do not need to be concerned with this
1522 * While holding this lock, you are trusted to actually be handling events.
1523 * If you are no longer handling events, you must call libusb_unlock_events()
1524 * as soon as possible.
1526 * \param ctx the context to operate on, or NULL for the default context
1527 * \returns 0 if the lock was obtained successfully
1528 * \returns 1 if the lock was not obtained (i.e. another thread holds the lock)
1531 int API_EXPORTED libusb_try_lock_events(libusb_context *ctx)
1534 USBI_GET_CONTEXT(ctx);
1536 /* is someone else waiting to modify poll fds? if so, don't let this thread
1537 * start event handling */
1538 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1539 r = ctx->pollfd_modify;
1540 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1542 usbi_dbg("someone else is modifying poll fds");
1546 r = usbi_mutex_trylock(&ctx->events_lock);
1550 ctx->event_handler_active = 1;
1555 * Acquire the event handling lock, blocking until successful acquisition if
1556 * it is contended. This lock is used to ensure that only one thread is
1557 * monitoring libusbx event sources at any one time.
1559 * You only need to use this lock if you are developing an application
1560 * which calls poll() or select() on libusbx's file descriptors directly.
1561 * If you stick to libusbx's event handling loop functions (e.g.
1562 * libusb_handle_events()) then you do not need to be concerned with this
1565 * While holding this lock, you are trusted to actually be handling events.
1566 * If you are no longer handling events, you must call libusb_unlock_events()
1567 * as soon as possible.
1569 * \param ctx the context to operate on, or NULL for the default context
1572 void API_EXPORTED libusb_lock_events(libusb_context *ctx)
1574 USBI_GET_CONTEXT(ctx);
1575 usbi_mutex_lock(&ctx->events_lock);
1576 ctx->event_handler_active = 1;
1580 * Release the lock previously acquired with libusb_try_lock_events() or
1581 * libusb_lock_events(). Releasing this lock will wake up any threads blocked
1582 * on libusb_wait_for_event().
1584 * \param ctx the context to operate on, or NULL for the default context
1587 void API_EXPORTED libusb_unlock_events(libusb_context *ctx)
1589 USBI_GET_CONTEXT(ctx);
1590 ctx->event_handler_active = 0;
1591 usbi_mutex_unlock(&ctx->events_lock);
1593 /* FIXME: perhaps we should be a bit more efficient by not broadcasting
1594 * the availability of the events lock when we are modifying pollfds
1595 * (check ctx->pollfd_modify)? */
1596 usbi_mutex_lock(&ctx->event_waiters_lock);
1597 usbi_cond_broadcast(&ctx->event_waiters_cond);
1598 usbi_mutex_unlock(&ctx->event_waiters_lock);
1602 * Determine if it is still OK for this thread to be doing event handling.
1604 * Sometimes, libusbx needs to temporarily pause all event handlers, and this
1605 * is the function you should use before polling file descriptors to see if
1608 * If this function instructs your thread to give up the events lock, you
1609 * should just continue the usual logic that is documented in \ref mtasync.
1610 * On the next iteration, your thread will fail to obtain the events lock,
1611 * and will hence become an event waiter.
1613 * This function should be called while the events lock is held: you don't
1614 * need to worry about the results of this function if your thread is not
1615 * the current event handler.
1617 * \param ctx the context to operate on, or NULL for the default context
1618 * \returns 1 if event handling can start or continue
1619 * \returns 0 if this thread must give up the events lock
1620 * \see \ref fullstory "Multi-threaded I/O: the full story"
1622 int API_EXPORTED libusb_event_handling_ok(libusb_context *ctx)
1625 USBI_GET_CONTEXT(ctx);
1627 /* is someone else waiting to modify poll fds? if so, don't let this thread
1628 * continue event handling */
1629 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1630 r = ctx->pollfd_modify;
1631 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1633 usbi_dbg("someone else is modifying poll fds");
1642 * Determine if an active thread is handling events (i.e. if anyone is holding
1643 * the event handling lock).
1645 * \param ctx the context to operate on, or NULL for the default context
1646 * \returns 1 if a thread is handling events
1647 * \returns 0 if there are no threads currently handling events
1650 int API_EXPORTED libusb_event_handler_active(libusb_context *ctx)
1653 USBI_GET_CONTEXT(ctx);
1655 /* is someone else waiting to modify poll fds? if so, don't let this thread
1656 * start event handling -- indicate that event handling is happening */
1657 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1658 r = ctx->pollfd_modify;
1659 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1661 usbi_dbg("someone else is modifying poll fds");
1665 return ctx->event_handler_active;
1669 * Acquire the event waiters lock. This lock is designed to be obtained under
1670 * the situation where you want to be aware when events are completed, but
1671 * some other thread is event handling so calling libusb_handle_events() is not
1674 * You then obtain this lock, re-check that another thread is still handling
1675 * events, then call libusb_wait_for_event().
1677 * You only need to use this lock if you are developing an application
1678 * which calls poll() or select() on libusbx's file descriptors directly,
1679 * <b>and</b> may potentially be handling events from 2 threads simultaenously.
1680 * If you stick to libusbx's event handling loop functions (e.g.
1681 * libusb_handle_events()) then you do not need to be concerned with this
1684 * \param ctx the context to operate on, or NULL for the default context
1687 void API_EXPORTED libusb_lock_event_waiters(libusb_context *ctx)
1689 USBI_GET_CONTEXT(ctx);
1690 usbi_mutex_lock(&ctx->event_waiters_lock);
1694 * Release the event waiters lock.
1695 * \param ctx the context to operate on, or NULL for the default context
1698 void API_EXPORTED libusb_unlock_event_waiters(libusb_context *ctx)
1700 USBI_GET_CONTEXT(ctx);
1701 usbi_mutex_unlock(&ctx->event_waiters_lock);
1705 * Wait for another thread to signal completion of an event. Must be called
1706 * with the event waiters lock held, see libusb_lock_event_waiters().
1708 * This function will block until any of the following conditions are met:
1709 * -# The timeout expires
1710 * -# A transfer completes
1711 * -# A thread releases the event handling lock through libusb_unlock_events()
1713 * Condition 1 is obvious. Condition 2 unblocks your thread <em>after</em>
1714 * the callback for the transfer has completed. Condition 3 is important
1715 * because it means that the thread that was previously handling events is no
1716 * longer doing so, so if any events are to complete, another thread needs to
1717 * step up and start event handling.
1719 * This function releases the event waiters lock before putting your thread
1720 * to sleep, and reacquires the lock as it is being woken up.
1722 * \param ctx the context to operate on, or NULL for the default context
1723 * \param tv maximum timeout for this blocking function. A NULL value
1724 * indicates unlimited timeout.
1725 * \returns 0 after a transfer completes or another thread stops event handling
1726 * \returns 1 if the timeout expired
1729 int API_EXPORTED libusb_wait_for_event(libusb_context *ctx, struct timeval *tv)
1731 struct timespec timeout;
1734 USBI_GET_CONTEXT(ctx);
1736 usbi_cond_wait(&ctx->event_waiters_cond, &ctx->event_waiters_lock);
1740 r = usbi_backend->clock_gettime(USBI_CLOCK_REALTIME, &timeout);
1742 usbi_err(ctx, "failed to read realtime clock, error %d", errno);
1743 return LIBUSB_ERROR_OTHER;
1746 timeout.tv_sec += tv->tv_sec;
1747 timeout.tv_nsec += tv->tv_usec * 1000;
1748 if (timeout.tv_nsec > 1000000000) {
1749 timeout.tv_nsec -= 1000000000;
1753 r = usbi_cond_timedwait(&ctx->event_waiters_cond,
1754 &ctx->event_waiters_lock, &timeout);
1755 return (r == ETIMEDOUT);
1758 static void handle_timeout(struct usbi_transfer *itransfer)
1760 struct libusb_transfer *transfer =
1761 USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1764 itransfer->flags |= USBI_TRANSFER_TIMED_OUT;
1765 r = libusb_cancel_transfer(transfer);
1767 usbi_warn(TRANSFER_CTX(transfer),
1768 "async cancel failed %d errno=%d", r, errno);
1771 static int handle_timeouts_locked(struct libusb_context *ctx)
1774 struct timespec systime_ts;
1775 struct timeval systime;
1776 struct usbi_transfer *transfer;
1778 if (list_empty(&ctx->flying_transfers))
1781 /* get current time */
1782 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, &systime_ts);
1786 TIMESPEC_TO_TIMEVAL(&systime, &systime_ts);
1788 /* iterate through flying transfers list, finding all transfers that
1789 * have expired timeouts */
1790 list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
1791 struct timeval *cur_tv = &transfer->timeout;
1793 /* if we've reached transfers of infinite timeout, we're all done */
1794 if (!timerisset(cur_tv))
1797 /* ignore timeouts we've already handled */
1798 if (transfer->flags & (USBI_TRANSFER_TIMED_OUT | USBI_TRANSFER_OS_HANDLES_TIMEOUT))
1801 /* if transfer has non-expired timeout, nothing more to do */
1802 if ((cur_tv->tv_sec > systime.tv_sec) ||
1803 (cur_tv->tv_sec == systime.tv_sec &&
1804 cur_tv->tv_usec > systime.tv_usec))
1807 /* otherwise, we've got an expired timeout to handle */
1808 handle_timeout(transfer);
1813 static int handle_timeouts(struct libusb_context *ctx)
1816 USBI_GET_CONTEXT(ctx);
1817 usbi_mutex_lock(&ctx->flying_transfers_lock);
1818 r = handle_timeouts_locked(ctx);
1819 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1823 #ifdef USBI_TIMERFD_AVAILABLE
1824 static int handle_timerfd_trigger(struct libusb_context *ctx)
1828 usbi_mutex_lock(&ctx->flying_transfers_lock);
1830 /* process the timeout that just happened */
1831 r = handle_timeouts_locked(ctx);
1835 /* arm for next timeout*/
1836 r = arm_timerfd_for_next_timeout(ctx);
1839 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1844 /* do the actual event handling. assumes that no other thread is concurrently
1845 * doing the same thing. */
1846 static int handle_events(struct libusb_context *ctx, struct timeval *tv)
1849 struct usbi_pollfd *ipollfd;
1850 POLL_NFDS_TYPE nfds = 0;
1851 struct pollfd *fds = NULL;
1855 usbi_mutex_lock(&ctx->pollfds_lock);
1856 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
1859 /* TODO: malloc when number of fd's changes, not on every poll */
1861 fds = malloc(sizeof(*fds) * nfds);
1863 usbi_mutex_unlock(&ctx->pollfds_lock);
1864 return LIBUSB_ERROR_NO_MEM;
1867 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd) {
1868 struct libusb_pollfd *pollfd = &ipollfd->pollfd;
1869 int fd = pollfd->fd;
1872 fds[i].events = pollfd->events;
1875 usbi_mutex_unlock(&ctx->pollfds_lock);
1877 timeout_ms = (tv->tv_sec * 1000) + (tv->tv_usec / 1000);
1879 /* round up to next millisecond */
1880 if (tv->tv_usec % 1000)
1883 usbi_dbg("poll() %d fds with timeout in %dms", nfds, timeout_ms);
1884 r = usbi_poll(fds, nfds, timeout_ms);
1885 usbi_dbg("poll() returned %d", r);
1888 return handle_timeouts(ctx);
1889 } else if (r == -1 && errno == EINTR) {
1891 return LIBUSB_ERROR_INTERRUPTED;
1894 usbi_err(ctx, "poll failed %d err=%d\n", r, errno);
1895 return LIBUSB_ERROR_IO;
1898 /* fd[0] is always the ctrl pipe */
1899 if (fds[0].revents) {
1900 /* another thread wanted to interrupt event handling, and it succeeded!
1901 * handle any other events that cropped up at the same time, and
1903 usbi_dbg("caught a fish on the control pipe");
1909 /* prevent OS backend from trying to handle events on ctrl pipe */
1915 #ifdef USBI_TIMERFD_AVAILABLE
1916 /* on timerfd configurations, fds[1] is the timerfd */
1917 if (usbi_using_timerfd(ctx) && fds[1].revents) {
1918 /* timerfd indicates that a timeout has expired */
1920 usbi_dbg("timerfd triggered");
1922 ret = handle_timerfd_trigger(ctx);
1924 /* return error code */
1927 } else if (r == 1) {
1928 /* no more active file descriptors, nothing more to do */
1932 /* more events pending...
1933 * prevent OS backend from trying to handle events on timerfd */
1940 r = usbi_backend->handle_events(ctx, fds, nfds, r);
1942 usbi_err(ctx, "backend handle_events failed with error %d", r);
1949 /* returns the smallest of:
1950 * 1. timeout of next URB
1951 * 2. user-supplied timeout
1952 * returns 1 if there is an already-expired timeout, otherwise returns 0
1955 static int get_next_timeout(libusb_context *ctx, struct timeval *tv,
1956 struct timeval *out)
1958 struct timeval timeout;
1959 int r = libusb_get_next_timeout(ctx, &timeout);
1961 /* timeout already expired? */
1962 if (!timerisset(&timeout))
1965 /* choose the smallest of next URB timeout or user specified timeout */
1966 if (timercmp(&timeout, tv, <))
1977 * Handle any pending events.
1979 * libusbx determines "pending events" by checking if any timeouts have expired
1980 * and by checking the set of file descriptors for activity.
1982 * If a zero timeval is passed, this function will handle any already-pending
1983 * events and then immediately return in non-blocking style.
1985 * If a non-zero timeval is passed and no events are currently pending, this
1986 * function will block waiting for events to handle up until the specified
1987 * timeout. If an event arrives or a signal is raised, this function will
1990 * If the parameter completed is not NULL then <em>after obtaining the event
1991 * handling lock</em> this function will return immediately if the integer
1992 * pointed to is not 0. This allows for race free waiting for the completion
1993 * of a specific transfer.
1995 * \param ctx the context to operate on, or NULL for the default context
1996 * \param tv the maximum time to block waiting for events, or an all zero
1997 * timeval struct for non-blocking mode
1998 * \param completed pointer to completion integer to check, or NULL
1999 * \returns 0 on success, or a LIBUSB_ERROR code on failure
2002 int API_EXPORTED libusb_handle_events_timeout_completed(libusb_context *ctx,
2003 struct timeval *tv, int *completed)
2006 struct timeval poll_timeout;
2008 USBI_GET_CONTEXT(ctx);
2009 r = get_next_timeout(ctx, tv, &poll_timeout);
2011 /* timeout already expired */
2012 return handle_timeouts(ctx);
2016 if (libusb_try_lock_events(ctx) == 0) {
2017 if (completed == NULL || !*completed) {
2018 /* we obtained the event lock: do our own event handling */
2019 usbi_dbg("doing our own event handling");
2020 r = handle_events(ctx, &poll_timeout);
2022 libusb_unlock_events(ctx);
2026 /* another thread is doing event handling. wait for thread events that
2027 * notify event completion. */
2028 libusb_lock_event_waiters(ctx);
2030 if (completed && *completed)
2033 if (!libusb_event_handler_active(ctx)) {
2034 /* we hit a race: whoever was event handling earlier finished in the
2035 * time it took us to reach this point. try the cycle again. */
2036 libusb_unlock_event_waiters(ctx);
2037 usbi_dbg("event handler was active but went away, retrying");
2041 usbi_dbg("another thread is doing event handling");
2042 r = libusb_wait_for_event(ctx, &poll_timeout);
2045 libusb_unlock_event_waiters(ctx);
2050 return handle_timeouts(ctx);
2056 * Handle any pending events
2058 * Like libusb_handle_events_timeout_completed(), but without the completed
2059 * parameter, calling this function is equivalent to calling
2060 * libusb_handle_events_timeout_completed() with a NULL completed parameter.
2062 * This function is kept primarily for backwards compatibility.
2063 * All new code should call libusb_handle_events_completed() or
2064 * libusb_handle_events_timeout_completed() to avoid race conditions.
2066 * \param ctx the context to operate on, or NULL for the default context
2067 * \param tv the maximum time to block waiting for events, or an all zero
2068 * timeval struct for non-blocking mode
2069 * \returns 0 on success, or a LIBUSB_ERROR code on failure
2071 int API_EXPORTED libusb_handle_events_timeout(libusb_context *ctx,
2074 return libusb_handle_events_timeout_completed(ctx, tv, NULL);
2078 * Handle any pending events in blocking mode. There is currently a timeout
2079 * hardcoded at 60 seconds but we plan to make it unlimited in future. For
2080 * finer control over whether this function is blocking or non-blocking, or
2081 * for control over the timeout, use libusb_handle_events_timeout_completed()
2084 * This function is kept primarily for backwards compatibility.
2085 * All new code should call libusb_handle_events_completed() or
2086 * libusb_handle_events_timeout_completed() to avoid race conditions.
2088 * \param ctx the context to operate on, or NULL for the default context
2089 * \returns 0 on success, or a LIBUSB_ERROR code on failure
2091 int API_EXPORTED libusb_handle_events(libusb_context *ctx)
2096 return libusb_handle_events_timeout_completed(ctx, &tv, NULL);
2100 * Handle any pending events in blocking mode.
2102 * Like libusb_handle_events(), with the addition of a completed parameter
2103 * to allow for race free waiting for the completion of a specific transfer.
2105 * See libusb_handle_events_timeout_completed() for details on the completed
2108 * \param ctx the context to operate on, or NULL for the default context
2109 * \param completed pointer to completion integer to check, or NULL
2110 * \returns 0 on success, or a LIBUSB_ERROR code on failure
2113 int API_EXPORTED libusb_handle_events_completed(libusb_context *ctx,
2119 return libusb_handle_events_timeout_completed(ctx, &tv, completed);
2123 * Handle any pending events by polling file descriptors, without checking if
2124 * any other threads are already doing so. Must be called with the event lock
2125 * held, see libusb_lock_events().
2127 * This function is designed to be called under the situation where you have
2128 * taken the event lock and are calling poll()/select() directly on libusbx's
2129 * file descriptors (as opposed to using libusb_handle_events() or similar).
2130 * You detect events on libusbx's descriptors, so you then call this function
2131 * with a zero timeout value (while still holding the event lock).
2133 * \param ctx the context to operate on, or NULL for the default context
2134 * \param tv the maximum time to block waiting for events, or zero for
2136 * \returns 0 on success, or a LIBUSB_ERROR code on failure
2139 int API_EXPORTED libusb_handle_events_locked(libusb_context *ctx,
2143 struct timeval poll_timeout;
2145 USBI_GET_CONTEXT(ctx);
2146 r = get_next_timeout(ctx, tv, &poll_timeout);
2148 /* timeout already expired */
2149 return handle_timeouts(ctx);
2152 return handle_events(ctx, &poll_timeout);
2156 * Determines whether your application must apply special timing considerations
2157 * when monitoring libusbx's file descriptors.
2159 * This function is only useful for applications which retrieve and poll
2160 * libusbx's file descriptors in their own main loop (\ref pollmain).
2162 * Ordinarily, libusbx's event handler needs to be called into at specific
2163 * moments in time (in addition to times when there is activity on the file
2164 * descriptor set). The usual approach is to use libusb_get_next_timeout()
2165 * to learn about when the next timeout occurs, and to adjust your
2166 * poll()/select() timeout accordingly so that you can make a call into the
2167 * library at that time.
2169 * Some platforms supported by libusbx do not come with this baggage - any
2170 * events relevant to timing will be represented by activity on the file
2171 * descriptor set, and libusb_get_next_timeout() will always return 0.
2172 * This function allows you to detect whether you are running on such a
2177 * \param ctx the context to operate on, or NULL for the default context
2178 * \returns 0 if you must call into libusbx at times determined by
2179 * libusb_get_next_timeout(), or 1 if all timeout events are handled internally
2180 * or through regular activity on the file descriptors.
2181 * \see \ref pollmain "Polling libusbx file descriptors for event handling"
2183 int API_EXPORTED libusb_pollfds_handle_timeouts(libusb_context *ctx)
2185 #if defined(USBI_TIMERFD_AVAILABLE)
2186 USBI_GET_CONTEXT(ctx);
2187 return usbi_using_timerfd(ctx);
2195 * Determine the next internal timeout that libusbx needs to handle. You only
2196 * need to use this function if you are calling poll() or select() or similar
2197 * on libusbx's file descriptors yourself - you do not need to use it if you
2198 * are calling libusb_handle_events() or a variant directly.
2200 * You should call this function in your main loop in order to determine how
2201 * long to wait for select() or poll() to return results. libusbx needs to be
2202 * called into at this timeout, so you should use it as an upper bound on
2203 * your select() or poll() call.
2205 * When the timeout has expired, call into libusb_handle_events_timeout()
2206 * (perhaps in non-blocking mode) so that libusbx can handle the timeout.
2208 * This function may return 1 (success) and an all-zero timeval. If this is
2209 * the case, it indicates that libusbx has a timeout that has already expired
2210 * so you should call libusb_handle_events_timeout() or similar immediately.
2211 * A return code of 0 indicates that there are no pending timeouts.
2213 * On some platforms, this function will always returns 0 (no pending
2214 * timeouts). See \ref polltime.
2216 * \param ctx the context to operate on, or NULL for the default context
2217 * \param tv output location for a relative time against the current
2218 * clock in which libusbx must be called into in order to process timeout events
2219 * \returns 0 if there are no pending timeouts, 1 if a timeout was returned,
2220 * or LIBUSB_ERROR_OTHER on failure
2222 int API_EXPORTED libusb_get_next_timeout(libusb_context *ctx,
2225 struct usbi_transfer *transfer;
2226 struct timespec cur_ts;
2227 struct timeval cur_tv;
2228 struct timeval *next_timeout;
2232 USBI_GET_CONTEXT(ctx);
2233 if (usbi_using_timerfd(ctx))
2236 usbi_mutex_lock(&ctx->flying_transfers_lock);
2237 if (list_empty(&ctx->flying_transfers)) {
2238 usbi_mutex_unlock(&ctx->flying_transfers_lock);
2239 usbi_dbg("no URBs, no timeout!");
2243 /* find next transfer which hasn't already been processed as timed out */
2244 list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
2245 if (transfer->flags & (USBI_TRANSFER_TIMED_OUT | USBI_TRANSFER_OS_HANDLES_TIMEOUT))
2248 /* no timeout for this transfer? */
2249 if (!timerisset(&transfer->timeout))
2255 usbi_mutex_unlock(&ctx->flying_transfers_lock);
2258 usbi_dbg("no URB with timeout or all handled by OS; no timeout!");
2262 next_timeout = &transfer->timeout;
2264 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, &cur_ts);
2266 usbi_err(ctx, "failed to read monotonic clock, errno=%d", errno);
2269 TIMESPEC_TO_TIMEVAL(&cur_tv, &cur_ts);
2271 if (!timercmp(&cur_tv, next_timeout, <)) {
2272 usbi_dbg("first timeout already expired");
2275 timersub(next_timeout, &cur_tv, tv);
2276 usbi_dbg("next timeout in %d.%06ds", tv->tv_sec, tv->tv_usec);
2283 * Register notification functions for file descriptor additions/removals.
2284 * These functions will be invoked for every new or removed file descriptor
2285 * that libusbx uses as an event source.
2287 * To remove notifiers, pass NULL values for the function pointers.
2289 * Note that file descriptors may have been added even before you register
2290 * these notifiers (e.g. at libusb_init() time).
2292 * Additionally, note that the removal notifier may be called during
2293 * libusb_exit() (e.g. when it is closing file descriptors that were opened
2294 * and added to the poll set at libusb_init() time). If you don't want this,
2295 * remove the notifiers immediately before calling libusb_exit().
2297 * \param ctx the context to operate on, or NULL for the default context
2298 * \param added_cb pointer to function for addition notifications
2299 * \param removed_cb pointer to function for removal notifications
2300 * \param user_data User data to be passed back to callbacks (useful for
2301 * passing context information)
2303 void API_EXPORTED libusb_set_pollfd_notifiers(libusb_context *ctx,
2304 libusb_pollfd_added_cb added_cb, libusb_pollfd_removed_cb removed_cb,
2307 USBI_GET_CONTEXT(ctx);
2308 ctx->fd_added_cb = added_cb;
2309 ctx->fd_removed_cb = removed_cb;
2310 ctx->fd_cb_user_data = user_data;
2313 /* Add a file descriptor to the list of file descriptors to be monitored.
2314 * events should be specified as a bitmask of events passed to poll(), e.g.
2315 * POLLIN and/or POLLOUT. */
2316 int usbi_add_pollfd(struct libusb_context *ctx, int fd, short events)
2318 struct usbi_pollfd *ipollfd = malloc(sizeof(*ipollfd));
2320 return LIBUSB_ERROR_NO_MEM;
2322 usbi_dbg("add fd %d events %d", fd, events);
2323 ipollfd->pollfd.fd = fd;
2324 ipollfd->pollfd.events = events;
2325 usbi_mutex_lock(&ctx->pollfds_lock);
2326 list_add_tail(&ipollfd->list, &ctx->pollfds);
2327 usbi_mutex_unlock(&ctx->pollfds_lock);
2329 if (ctx->fd_added_cb)
2330 ctx->fd_added_cb(fd, events, ctx->fd_cb_user_data);
2334 /* Remove a file descriptor from the list of file descriptors to be polled. */
2335 void usbi_remove_pollfd(struct libusb_context *ctx, int fd)
2337 struct usbi_pollfd *ipollfd;
2340 usbi_dbg("remove fd %d", fd);
2341 usbi_mutex_lock(&ctx->pollfds_lock);
2342 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2343 if (ipollfd->pollfd.fd == fd) {
2349 usbi_dbg("couldn't find fd %d to remove", fd);
2350 usbi_mutex_unlock(&ctx->pollfds_lock);
2354 list_del(&ipollfd->list);
2355 usbi_mutex_unlock(&ctx->pollfds_lock);
2357 if (ctx->fd_removed_cb)
2358 ctx->fd_removed_cb(fd, ctx->fd_cb_user_data);
2362 * Retrieve a list of file descriptors that should be polled by your main loop
2363 * as libusbx event sources.
2365 * The returned list is NULL-terminated and should be freed with free() when
2366 * done. The actual list contents must not be touched.
2368 * As file descriptors are a Unix-specific concept, this function is not
2369 * available on Windows and will always return NULL.
2371 * \param ctx the context to operate on, or NULL for the default context
2372 * \returns a NULL-terminated list of libusb_pollfd structures
2373 * \returns NULL on error
2374 * \returns NULL on platforms where the functionality is not available
2377 const struct libusb_pollfd ** LIBUSB_CALL libusb_get_pollfds(
2378 libusb_context *ctx)
2381 struct libusb_pollfd **ret = NULL;
2382 struct usbi_pollfd *ipollfd;
2385 USBI_GET_CONTEXT(ctx);
2387 usbi_mutex_lock(&ctx->pollfds_lock);
2388 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2391 ret = calloc(cnt + 1, sizeof(struct libusb_pollfd *));
2395 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2396 ret[i++] = (struct libusb_pollfd *) ipollfd;
2400 usbi_mutex_unlock(&ctx->pollfds_lock);
2401 return (const struct libusb_pollfd **) ret;
2403 usbi_err(ctx, "external polling of libusbx's internal descriptors "\
2404 "is not yet supported on Windows platforms");
2409 /* Backends call this from handle_events to report disconnection of a device.
2410 * The transfers get cancelled appropriately.
2412 void usbi_handle_disconnect(struct libusb_device_handle *handle)
2414 struct usbi_transfer *cur;
2415 struct usbi_transfer *to_cancel;
2417 usbi_dbg("device %d.%d",
2418 handle->dev->bus_number, handle->dev->device_address);
2420 /* terminate all pending transfers with the LIBUSB_TRANSFER_NO_DEVICE
2423 * this is a bit tricky because:
2424 * 1. we can't do transfer completion while holding flying_transfers_lock
2425 * 2. the transfers list can change underneath us - if we were to build a
2426 * list of transfers to complete (while holding look), the situation
2427 * might be different by the time we come to free them
2429 * so we resort to a loop-based approach as below
2430 * FIXME: is this still potentially racy?
2434 usbi_mutex_lock(&HANDLE_CTX(handle)->flying_transfers_lock);
2436 list_for_each_entry(cur, &HANDLE_CTX(handle)->flying_transfers, list, struct usbi_transfer)
2437 if (USBI_TRANSFER_TO_LIBUSB_TRANSFER(cur)->dev_handle == handle) {
2441 usbi_mutex_unlock(&HANDLE_CTX(handle)->flying_transfers_lock);
2446 usbi_backend->clear_transfer_priv(to_cancel);
2447 usbi_handle_transfer_completion(to_cancel, LIBUSB_TRANSFER_NO_DEVICE);