2 * I/O functions for libusb
3 * Copyright (C) 2007-2009 Daniel Drake <dsd@gentoo.org>
4 * Copyright (c) 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 #include "os/poll_posix.h"
35 #ifdef USBI_TIMERFD_AVAILABLE
36 #include <sys/timerfd.h>
42 * \page io Synchronous and asynchronous device I/O
44 * \section intro Introduction
46 * If you're using libusb in your application, you're probably wanting to
47 * perform I/O with devices - you want to perform USB data transfers.
49 * libusb offers two separate interfaces for device I/O. This page aims to
50 * introduce the two in order to help you decide which one is more suitable
51 * for your application. You can also choose to use both interfaces in your
52 * application by considering each transfer on a case-by-case basis.
54 * Once you have read through the following discussion, you should consult the
55 * detailed API documentation pages for the details:
59 * \section theory Transfers at a logical level
61 * At a logical level, USB transfers typically happen in two parts. For
62 * example, when reading data from a endpoint:
63 * -# A request for data is sent to the device
64 * -# Some time later, the incoming data is received by the host
66 * or when writing data to an endpoint:
68 * -# The data is sent to the device
69 * -# Some time later, the host receives acknowledgement from the device that
70 * the data has been transferred.
72 * There may be an indefinite delay between the two steps. Consider a
73 * fictional USB input device with a button that the user can press. In order
74 * to determine when the button is pressed, you would likely submit a request
75 * to read data on a bulk or interrupt endpoint and wait for data to arrive.
76 * Data will arrive when the button is pressed by the user, which is
77 * potentially hours later.
79 * libusb offers both a synchronous and an asynchronous interface to performing
80 * USB transfers. The main difference is that the synchronous interface
81 * combines both steps indicated above into a single function call, whereas
82 * the asynchronous interface separates them.
84 * \section sync The synchronous interface
86 * The synchronous I/O interface allows you to perform a USB transfer with
87 * a single function call. When the function call returns, the transfer has
88 * completed and you can parse the results.
90 * If you have used the libusb-0.1 before, this I/O style will seem familar to
91 * you. libusb-0.1 only offered a synchronous interface.
93 * In our input device example, to read button presses you might write code
94 * in the following style:
96 unsigned char data[4];
98 int r = libusb_bulk_transfer(handle, EP_IN, data, sizeof(data), &actual_length, 0);
99 if (r == 0 && actual_length == sizeof(data)) {
100 // results of the transaction can now be found in the data buffer
101 // parse them here and report button press
107 * The main advantage of this model is simplicity: you did everything with
108 * a single simple function call.
110 * However, this interface has its limitations. Your application will sleep
111 * inside libusb_bulk_transfer() until the transaction has completed. If it
112 * takes the user 3 hours to press the button, your application will be
113 * sleeping for that long. Execution will be tied up inside the library -
114 * the entire thread will be useless for that duration.
116 * Another issue is that by tieing up the thread with that single transaction
117 * there is no possibility of performing I/O with multiple endpoints and/or
118 * multiple devices simultaneously, unless you resort to creating one thread
121 * Additionally, there is no opportunity to cancel the transfer after the
122 * request has been submitted.
124 * For details on how to use the synchronous API, see the
125 * \ref syncio "synchronous I/O API documentation" pages.
127 * \section async The asynchronous interface
129 * Asynchronous I/O is the most significant new feature in libusb-1.0.
130 * Although it is a more complex interface, it solves all the issues detailed
133 * Instead of providing which functions that block until the I/O has complete,
134 * libusb's asynchronous interface presents non-blocking functions which
135 * begin a transfer and then return immediately. Your application passes a
136 * callback function pointer to this non-blocking function, which libusb will
137 * call with the results of the transaction when it has completed.
139 * Transfers which have been submitted through the non-blocking functions
140 * can be cancelled with a separate function call.
142 * The non-blocking nature of this interface allows you to be simultaneously
143 * performing I/O to multiple endpoints on multiple devices, without having
146 * This added flexibility does come with some complications though:
147 * - In the interest of being a lightweight library, libusb does not create
148 * threads and can only operate when your application is calling into it. Your
149 * application must call into libusb from it's main loop when events are ready
150 * to be handled, or you must use some other scheme to allow libusb to
151 * undertake whatever work needs to be done.
152 * - libusb also needs to be called into at certain fixed points in time in
153 * order to accurately handle transfer timeouts.
154 * - Memory handling becomes more complex. You cannot use stack memory unless
155 * the function with that stack is guaranteed not to return until the transfer
156 * callback has finished executing.
157 * - You generally lose some linearity from your code flow because submitting
158 * the transfer request is done in a separate function from where the transfer
159 * results are handled. This becomes particularly obvious when you want to
160 * submit a second transfer based on the results of an earlier transfer.
162 * Internally, libusb's synchronous interface is expressed in terms of function
163 * calls to the asynchronous interface.
165 * For details on how to use the asynchronous API, see the
166 * \ref asyncio "asynchronous I/O API" documentation pages.
171 * \page packetoverflow Packets and overflows
173 * \section packets Packet abstraction
175 * The USB specifications describe how data is transmitted in packets, with
176 * constraints on packet size defined by endpoint descriptors. The host must
177 * not send data payloads larger than the endpoint's maximum packet size.
179 * libusb and the underlying OS abstract out the packet concept, allowing you
180 * to request transfers of any size. Internally, the request will be divided
181 * up into correctly-sized packets. You do not have to be concerned with
182 * packet sizes, but there is one exception when considering overflows.
184 * \section overflow Bulk/interrupt transfer overflows
186 * When requesting data on a bulk endpoint, libusb requires you to supply a
187 * buffer and the maximum number of bytes of data that libusb can put in that
188 * buffer. However, the size of the buffer is not communicated to the device -
189 * the device is just asked to send any amount of data.
191 * There is no problem if the device sends an amount of data that is less than
192 * or equal to the buffer size. libusb reports this condition to you through
193 * the \ref libusb_transfer::actual_length "libusb_transfer.actual_length"
196 * Problems may occur if the device attempts to send more data than can fit in
197 * the buffer. libusb reports LIBUSB_TRANSFER_OVERFLOW for this condition but
198 * other behaviour is largely undefined: actual_length may or may not be
199 * accurate, the chunk of data that can fit in the buffer (before overflow)
200 * may or may not have been transferred.
202 * Overflows are nasty, but can be avoided. Even though you were told to
203 * ignore packets above, think about the lower level details: each transfer is
204 * split into packets (typically small, with a maximum size of 512 bytes).
205 * Overflows can only happen if the final packet in an incoming data transfer
206 * is smaller than the actual packet that the device wants to transfer.
207 * Therefore, you will never see an overflow if your transfer buffer size is a
208 * multiple of the endpoint's packet size: the final packet will either
209 * fill up completely or will be only partially filled.
213 * @defgroup asyncio Asynchronous device I/O
215 * This page details libusb's asynchronous (non-blocking) API for USB device
216 * I/O. This interface is very powerful but is also quite complex - you will
217 * need to read this page carefully to understand the necessary considerations
218 * and issues surrounding use of this interface. Simplistic applications
219 * may wish to consider the \ref syncio "synchronous I/O API" instead.
221 * The asynchronous interface is built around the idea of separating transfer
222 * submission and handling of transfer completion (the synchronous model
223 * combines both of these into one). There may be a long delay between
224 * submission and completion, however the asynchronous submission function
225 * is non-blocking so will return control to your application during that
226 * potentially long delay.
228 * \section asyncabstraction Transfer abstraction
230 * For the asynchronous I/O, libusb implements the concept of a generic
231 * transfer entity for all types of I/O (control, bulk, interrupt,
232 * isochronous). The generic transfer object must be treated slightly
233 * differently depending on which type of I/O you are performing with it.
235 * This is represented by the public libusb_transfer structure type.
237 * \section asynctrf Asynchronous transfers
239 * We can view asynchronous I/O as a 5 step process:
240 * -# <b>Allocation</b>: allocate a libusb_transfer
241 * -# <b>Filling</b>: populate the libusb_transfer instance with information
242 * about the transfer you wish to perform
243 * -# <b>Submission</b>: ask libusb to submit the transfer
244 * -# <b>Completion handling</b>: examine transfer results in the
245 * libusb_transfer structure
246 * -# <b>Deallocation</b>: clean up resources
249 * \subsection asyncalloc Allocation
251 * This step involves allocating memory for a USB transfer. This is the
252 * generic transfer object mentioned above. At this stage, the transfer
253 * is "blank" with no details about what type of I/O it will be used for.
255 * Allocation is done with the libusb_alloc_transfer() function. You must use
256 * this function rather than allocating your own transfers.
258 * \subsection asyncfill Filling
260 * This step is where you take a previously allocated transfer and fill it
261 * with information to determine the message type and direction, data buffer,
262 * callback function, etc.
264 * You can either fill the required fields yourself or you can use the
265 * helper functions: libusb_fill_control_transfer(), libusb_fill_bulk_transfer()
266 * and libusb_fill_interrupt_transfer().
268 * \subsection asyncsubmit Submission
270 * When you have allocated a transfer and filled it, you can submit it using
271 * libusb_submit_transfer(). This function returns immediately but can be
272 * regarded as firing off the I/O request in the background.
274 * \subsection asynccomplete Completion handling
276 * After a transfer has been submitted, one of four things can happen to it:
278 * - The transfer completes (i.e. some data was transferred)
279 * - The transfer has a timeout and the timeout expires before all data is
281 * - The transfer fails due to an error
282 * - The transfer is cancelled
284 * Each of these will cause the user-specified transfer callback function to
285 * be invoked. It is up to the callback function to determine which of the
286 * above actually happened and to act accordingly.
288 * The user-specified callback is passed a pointer to the libusb_transfer
289 * structure which was used to setup and submit the transfer. At completion
290 * time, libusb has populated this structure with results of the transfer:
291 * success or failure reason, number of bytes of data transferred, etc. See
292 * the libusb_transfer structure documentation for more information.
294 * \subsection Deallocation
296 * When a transfer has completed (i.e. the callback function has been invoked),
297 * you are advised to free the transfer (unless you wish to resubmit it, see
298 * below). Transfers are deallocated with libusb_free_transfer().
300 * It is undefined behaviour to free a transfer which has not completed.
302 * \section asyncresubmit Resubmission
304 * You may be wondering why allocation, filling, and submission are all
305 * separated above where they could reasonably be combined into a single
308 * The reason for separation is to allow you to resubmit transfers without
309 * having to allocate new ones every time. This is especially useful for
310 * common situations dealing with interrupt endpoints - you allocate one
311 * transfer, fill and submit it, and when it returns with results you just
312 * resubmit it for the next interrupt.
314 * \section asynccancel Cancellation
316 * Another advantage of using the asynchronous interface is that you have
317 * the ability to cancel transfers which have not yet completed. This is
318 * done by calling the libusb_cancel_transfer() function.
320 * libusb_cancel_transfer() is asynchronous/non-blocking in itself. When the
321 * cancellation actually completes, the transfer's callback function will
322 * be invoked, and the callback function should check the transfer status to
323 * determine that it was cancelled.
325 * Freeing the transfer after it has been cancelled but before cancellation
326 * has completed will result in undefined behaviour.
328 * When a transfer is cancelled, some of the data may have been transferred.
329 * libusb will communicate this to you in the transfer callback. Do not assume
330 * that no data was transferred.
332 * \section bulk_overflows Overflows on device-to-host bulk/interrupt endpoints
334 * If your device does not have predictable transfer sizes (or it misbehaves),
335 * your application may submit a request for data on an IN endpoint which is
336 * smaller than the data that the device wishes to send. In some circumstances
337 * this will cause an overflow, which is a nasty condition to deal with. See
338 * the \ref packetoverflow page for discussion.
340 * \section asyncctrl Considerations for control transfers
342 * The <tt>libusb_transfer</tt> structure is generic and hence does not
343 * include specific fields for the control-specific setup packet structure.
345 * In order to perform a control transfer, you must place the 8-byte setup
346 * packet at the start of the data buffer. To simplify this, you could
347 * cast the buffer pointer to type struct libusb_control_setup, or you can
348 * use the helper function libusb_fill_control_setup().
350 * The wLength field placed in the setup packet must be the length you would
351 * expect to be sent in the setup packet: the length of the payload that
352 * follows (or the expected maximum number of bytes to receive). However,
353 * the length field of the libusb_transfer object must be the length of
354 * the data buffer - i.e. it should be wLength <em>plus</em> the size of
355 * the setup packet (LIBUSB_CONTROL_SETUP_SIZE).
357 * If you use the helper functions, this is simplified for you:
358 * -# Allocate a buffer of size LIBUSB_CONTROL_SETUP_SIZE plus the size of the
359 * data you are sending/requesting.
360 * -# Call libusb_fill_control_setup() on the data buffer, using the transfer
361 * request size as the wLength value (i.e. do not include the extra space you
362 * allocated for the control setup).
363 * -# If this is a host-to-device transfer, place the data to be transferred
364 * in the data buffer, starting at offset LIBUSB_CONTROL_SETUP_SIZE.
365 * -# Call libusb_fill_control_transfer() to associate the data buffer with
366 * the transfer (and to set the remaining details such as callback and timeout).
367 * - Note that there is no parameter to set the length field of the transfer.
368 * The length is automatically inferred from the wLength field of the setup
370 * -# Submit the transfer.
372 * The multi-byte control setup fields (wValue, wIndex and wLength) must
373 * be given in little-endian byte order (the endianness of the USB bus).
374 * Endianness conversion is transparently handled by
375 * libusb_fill_control_setup() which is documented to accept host-endian
378 * Further considerations are needed when handling transfer completion in
379 * your callback function:
380 * - As you might expect, the setup packet will still be sitting at the start
381 * of the data buffer.
382 * - If this was a device-to-host transfer, the received data will be sitting
383 * at offset LIBUSB_CONTROL_SETUP_SIZE into the buffer.
384 * - The actual_length field of the transfer structure is relative to the
385 * wLength of the setup packet, rather than the size of the data buffer. So,
386 * if your wLength was 4, your transfer's <tt>length</tt> was 12, then you
387 * should expect an <tt>actual_length</tt> of 4 to indicate that the data was
388 * transferred in entirity.
390 * To simplify parsing of setup packets and obtaining the data from the
391 * correct offset, you may wish to use the libusb_control_transfer_get_data()
392 * and libusb_control_transfer_get_setup() functions within your transfer
395 * Even though control endpoints do not halt, a completed control transfer
396 * may have a LIBUSB_TRANSFER_STALL status code. This indicates the control
397 * request was not supported.
399 * \section asyncintr Considerations for interrupt transfers
401 * All interrupt transfers are performed using the polling interval presented
402 * by the bInterval value of the endpoint descriptor.
404 * \section asynciso Considerations for isochronous transfers
406 * Isochronous transfers are more complicated than transfers to
407 * non-isochronous endpoints.
409 * To perform I/O to an isochronous endpoint, allocate the transfer by calling
410 * libusb_alloc_transfer() with an appropriate number of isochronous packets.
412 * During filling, set \ref libusb_transfer::type "type" to
413 * \ref libusb_transfer_type::LIBUSB_TRANSFER_TYPE_ISOCHRONOUS
414 * "LIBUSB_TRANSFER_TYPE_ISOCHRONOUS", and set
415 * \ref libusb_transfer::num_iso_packets "num_iso_packets" to a value less than
416 * or equal to the number of packets you requested during allocation.
417 * libusb_alloc_transfer() does not set either of these fields for you, given
418 * that you might not even use the transfer on an isochronous endpoint.
420 * Next, populate the length field for the first num_iso_packets entries in
421 * the \ref libusb_transfer::iso_packet_desc "iso_packet_desc" array. Section
422 * 5.6.3 of the USB2 specifications describe how the maximum isochronous
423 * packet length is determined by the wMaxPacketSize field in the endpoint
425 * Two functions can help you here:
427 * - libusb_get_max_iso_packet_size() is an easy way to determine the max
428 * packet size for an isochronous endpoint. Note that the maximum packet
429 * size is actually the maximum number of bytes that can be transmitted in
430 * a single microframe, therefore this function multiplies the maximum number
431 * of bytes per transaction by the number of transaction opportunities per
433 * - libusb_set_iso_packet_lengths() assigns the same length to all packets
434 * within a transfer, which is usually what you want.
436 * For outgoing transfers, you'll obviously fill the buffer and populate the
437 * packet descriptors in hope that all the data gets transferred. For incoming
438 * transfers, you must ensure the buffer has sufficient capacity for
439 * the situation where all packets transfer the full amount of requested data.
441 * Completion handling requires some extra consideration. The
442 * \ref libusb_transfer::actual_length "actual_length" field of the transfer
443 * is meaningless and should not be examined; instead you must refer to the
444 * \ref libusb_iso_packet_descriptor::actual_length "actual_length" field of
445 * each individual packet.
447 * The \ref libusb_transfer::status "status" field of the transfer is also a
449 * - If the packets were submitted and the isochronous data microframes
450 * completed normally, status will have value
451 * \ref libusb_transfer_status::LIBUSB_TRANSFER_COMPLETED
452 * "LIBUSB_TRANSFER_COMPLETED". Note that bus errors and software-incurred
453 * delays are not counted as transfer errors; the transfer.status field may
454 * indicate COMPLETED even if some or all of the packets failed. Refer to
455 * the \ref libusb_iso_packet_descriptor::status "status" field of each
456 * individual packet to determine packet failures.
457 * - The status field will have value
458 * \ref libusb_transfer_status::LIBUSB_TRANSFER_ERROR
459 * "LIBUSB_TRANSFER_ERROR" only when serious errors were encountered.
460 * - Other transfer status codes occur with normal behaviour.
462 * The data for each packet will be found at an offset into the buffer that
463 * can be calculated as if each prior packet completed in full. The
464 * libusb_get_iso_packet_buffer() and libusb_get_iso_packet_buffer_simple()
465 * functions may help you here.
467 * \section asyncmem Memory caveats
469 * In most circumstances, it is not safe to use stack memory for transfer
470 * buffers. This is because the function that fired off the asynchronous
471 * transfer may return before libusb has finished using the buffer, and when
472 * the function returns it's stack gets destroyed. This is true for both
473 * host-to-device and device-to-host transfers.
475 * The only case in which it is safe to use stack memory is where you can
476 * guarantee that the function owning the stack space for the buffer does not
477 * return until after the transfer's callback function has completed. In every
478 * other case, you need to use heap memory instead.
480 * \section asyncflags Fine control
482 * Through using this asynchronous interface, you may find yourself repeating
483 * a few simple operations many times. You can apply a bitwise OR of certain
484 * flags to a transfer to simplify certain things:
485 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_SHORT_NOT_OK
486 * "LIBUSB_TRANSFER_SHORT_NOT_OK" results in transfers which transferred
487 * less than the requested amount of data being marked with status
488 * \ref libusb_transfer_status::LIBUSB_TRANSFER_ERROR "LIBUSB_TRANSFER_ERROR"
489 * (they would normally be regarded as COMPLETED)
490 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_BUFFER
491 * "LIBUSB_TRANSFER_FREE_BUFFER" allows you to ask libusb to free the transfer
492 * buffer when freeing the transfer.
493 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_TRANSFER
494 * "LIBUSB_TRANSFER_FREE_TRANSFER" causes libusb to automatically free the
495 * transfer after the transfer callback returns.
497 * \section asyncevent Event handling
499 * In accordance of the aim of being a lightweight library, libusb does not
500 * create threads internally. This means that libusb code does not execute
501 * at any time other than when your application is calling a libusb function.
502 * However, an asynchronous model requires that libusb perform work at various
503 * points in time - namely processing the results of previously-submitted
504 * transfers and invoking the user-supplied callback function.
506 * This gives rise to the libusb_handle_events() function which your
507 * application must call into when libusb has work do to. This gives libusb
508 * the opportunity to reap pending transfers, invoke callbacks, etc.
510 * The first issue to discuss here is how your application can figure out
511 * when libusb has work to do. In fact, there are two naive options which
512 * do not actually require your application to know this:
513 * -# Periodically call libusb_handle_events() in non-blocking mode at fixed
514 * short intervals from your main loop
515 * -# Repeatedly call libusb_handle_events() in blocking mode from a dedicated
518 * The first option is plainly not very nice, and will cause unnecessary
519 * CPU wakeups leading to increased power usage and decreased battery life.
520 * The second option is not very nice either, but may be the nicest option
521 * available to you if the "proper" approach can not be applied to your
522 * application (read on...).
524 * The recommended option is to integrate libusb with your application main
525 * event loop. libusb exposes a set of file descriptors which allow you to do
526 * this. Your main loop is probably already calling poll() or select() or a
527 * variant on a set of file descriptors for other event sources (e.g. keyboard
528 * button presses, mouse movements, network sockets, etc). You then add
529 * libusb's file descriptors to your poll()/select() calls, and when activity
530 * is detected on such descriptors you know it is time to call
531 * libusb_handle_events().
533 * There is one final event handling complication. libusb supports
534 * asynchronous transfers which time out after a specified time period, and
535 * this requires that libusb is called into at or after the timeout so that
536 * the timeout can be handled. So, in addition to considering libusb's file
537 * descriptors in your main event loop, you must also consider that libusb
538 * sometimes needs to be called into at fixed points in time even when there
539 * is no file descriptor activity.
541 * For the details on retrieving the set of file descriptors and determining
542 * the next timeout, see the \ref poll "polling and timing" API documentation.
546 * @defgroup poll Polling and timing
548 * This page documents libusb's functions for polling events and timing.
549 * These functions are only necessary for users of the
550 * \ref asyncio "asynchronous API". If you are only using the simpler
551 * \ref syncio "synchronous API" then you do not need to ever call these
554 * The justification for the functionality described here has already been
555 * discussed in the \ref asyncevent "event handling" section of the
556 * asynchronous API documentation. In summary, libusb does not create internal
557 * threads for event processing and hence relies on your application calling
558 * into libusb at certain points in time so that pending events can be handled.
559 * In order to know precisely when libusb needs to be called into, libusb
560 * offers you a set of pollable file descriptors and information about when
561 * the next timeout expires.
563 * If you are using the asynchronous I/O API, you must take one of the two
564 * following options, otherwise your I/O will not complete.
566 * \section pollsimple The simple option
568 * If your application revolves solely around libusb and does not need to
569 * handle other event sources, you can have a program structure as follows:
572 // find and open device
573 // maybe fire off some initial async I/O
575 while (user_has_not_requested_exit)
576 libusb_handle_events(ctx);
581 * With such a simple main loop, you do not have to worry about managing
582 * sets of file descriptors or handling timeouts. libusb_handle_events() will
583 * handle those details internally.
585 * \section pollmain The more advanced option
587 * In more advanced applications, you will already have a main loop which
588 * is monitoring other event sources: network sockets, X11 events, mouse
589 * movements, etc. Through exposing a set of file descriptors, libusb is
590 * designed to cleanly integrate into such main loops.
592 * In addition to polling file descriptors for the other event sources, you
593 * take a set of file descriptors from libusb and monitor those too. When you
594 * detect activity on libusb's file descriptors, you call
595 * libusb_handle_events_timeout() in non-blocking mode.
597 * What's more, libusb may also need to handle events at specific moments in
598 * time. No file descriptor activity is generated at these times, so your
599 * own application needs to be continually aware of when the next one of these
600 * moments occurs (through calling libusb_get_next_timeout()), and then it
601 * needs to call libusb_handle_events_timeout() in non-blocking mode when
602 * these moments occur. This means that you need to adjust your
603 * poll()/select() timeout accordingly.
605 * libusb provides you with a set of file descriptors to poll and expects you
606 * to poll all of them, treating them as a single entity. The meaning of each
607 * file descriptor in the set is an internal implementation detail,
608 * platform-dependent and may vary from release to release. Don't try and
609 * interpret the meaning of the file descriptors, just do as libusb indicates,
610 * polling all of them at once.
612 * In pseudo-code, you want something that looks like:
616 libusb_get_pollfds(ctx)
617 while (user has not requested application exit) {
618 libusb_get_next_timeout(ctx);
619 poll(on libusb file descriptors plus any other event sources of interest,
620 using a timeout no larger than the value libusb just suggested)
621 if (poll() indicated activity on libusb file descriptors)
622 libusb_handle_events_timeout(ctx, 0);
623 if (time has elapsed to or beyond the libusb timeout)
624 libusb_handle_events_timeout(ctx, 0);
625 // handle events from other sources here
631 * \subsection polltime Notes on time-based events
633 * The above complication with having to track time and call into libusb at
634 * specific moments is a bit of a headache. For maximum compatibility, you do
635 * need to write your main loop as above, but you may decide that you can
636 * restrict the supported platforms of your application and get away with
637 * a more simplistic scheme.
639 * These time-based event complications are \b not required on the following
642 * - Linux, provided that the following version requirements are satisfied:
643 * - Linux v2.6.27 or newer, compiled with timerfd support
644 * - glibc v2.9 or newer
645 * - libusb v1.0.5 or newer
647 * Under these configurations, libusb_get_next_timeout() will \em always return
648 * 0, so your main loop can be simplified to:
652 libusb_get_pollfds(ctx)
653 while (user has not requested application exit) {
654 poll(on libusb file descriptors plus any other event sources of interest,
655 using any timeout that you like)
656 if (poll() indicated activity on libusb file descriptors)
657 libusb_handle_events_timeout(ctx, 0);
658 // handle events from other sources here
664 * Do remember that if you simplify your main loop to the above, you will
665 * lose compatibility with some platforms (including legacy Linux platforms,
666 * and <em>any future platforms supported by libusb which may have time-based
667 * event requirements</em>). The resultant problems will likely appear as
668 * strange bugs in your application.
670 * You can use the libusb_pollfds_handle_timeouts() function to do a runtime
671 * check to see if it is safe to ignore the time-based event complications.
672 * If your application has taken the shortcut of ignoring libusb's next timeout
673 * in your main loop, then you are advised to check the return value of
674 * libusb_pollfds_handle_timeouts() during application startup, and to abort
675 * if the platform does suffer from these timing complications.
677 * \subsection fdsetchange Changes in the file descriptor set
679 * The set of file descriptors that libusb uses as event sources may change
680 * during the life of your application. Rather than having to repeatedly
681 * call libusb_get_pollfds(), you can set up notification functions for when
682 * the file descriptor set changes using libusb_set_pollfd_notifiers().
684 * \subsection mtissues Multi-threaded considerations
686 * Unfortunately, the situation is complicated further when multiple threads
687 * come into play. If two threads are monitoring the same file descriptors,
688 * the fact that only one thread will be woken up when an event occurs causes
691 * The events lock, event waiters lock, and libusb_handle_events_locked()
692 * entities are added to solve these problems. You do not need to be concerned
693 * with these entities otherwise.
695 * See the extra documentation: \ref mtasync
698 /** \page mtasync Multi-threaded applications and asynchronous I/O
700 * libusb is a thread-safe library, but extra considerations must be applied
701 * to applications which interact with libusb from multiple threads.
703 * The underlying issue that must be addressed is that all libusb I/O
704 * revolves around monitoring file descriptors through the poll()/select()
705 * system calls. This is directly exposed at the
706 * \ref asyncio "asynchronous interface" but it is important to note that the
707 * \ref syncio "synchronous interface" is implemented on top of the
708 * asynchonrous interface, therefore the same considerations apply.
710 * The issue is that if two or more threads are concurrently calling poll()
711 * or select() on libusb's file descriptors then only one of those threads
712 * will be woken up when an event arrives. The others will be completely
713 * oblivious that anything has happened.
715 * Consider the following pseudo-code, which submits an asynchronous transfer
716 * then waits for its completion. This style is one way you could implement a
717 * synchronous interface on top of the asynchronous interface (and libusb
718 * does something similar, albeit more advanced due to the complications
719 * explained on this page).
722 void cb(struct libusb_transfer *transfer)
724 int *completed = transfer->user_data;
729 struct libusb_transfer *transfer;
730 unsigned char buffer[LIBUSB_CONTROL_SETUP_SIZE];
733 transfer = libusb_alloc_transfer(0);
734 libusb_fill_control_setup(buffer,
735 LIBUSB_REQUEST_TYPE_VENDOR | LIBUSB_ENDPOINT_OUT, 0x04, 0x01, 0, 0);
736 libusb_fill_control_transfer(transfer, dev, buffer, cb, &completed, 1000);
737 libusb_submit_transfer(transfer);
740 poll(libusb file descriptors, 120*1000);
741 if (poll indicates activity)
742 libusb_handle_events_timeout(ctx, 0);
744 printf("completed!");
749 * Here we are <em>serializing</em> completion of an asynchronous event
750 * against a condition - the condition being completion of a specific transfer.
751 * The poll() loop has a long timeout to minimize CPU usage during situations
752 * when nothing is happening (it could reasonably be unlimited).
754 * If this is the only thread that is polling libusb's file descriptors, there
755 * is no problem: there is no danger that another thread will swallow up the
756 * event that we are interested in. On the other hand, if there is another
757 * thread polling the same descriptors, there is a chance that it will receive
758 * the event that we were interested in. In this situation, <tt>myfunc()</tt>
759 * will only realise that the transfer has completed on the next iteration of
760 * the loop, <em>up to 120 seconds later.</em> Clearly a two-minute delay is
761 * undesirable, and don't even think about using short timeouts to circumvent
764 * The solution here is to ensure that no two threads are ever polling the
765 * file descriptors at the same time. A naive implementation of this would
766 * impact the capabilities of the library, so libusb offers the scheme
767 * documented below to ensure no loss of functionality.
769 * Before we go any further, it is worth mentioning that all libusb-wrapped
770 * event handling procedures fully adhere to the scheme documented below.
771 * This includes libusb_handle_events() and all the synchronous I/O functions -
772 * libusb hides this headache from you. You do not need to worry about any
773 * of these issues if you stick to that level.
775 * The problem is when we consider the fact that libusb exposes file
776 * descriptors to allow for you to integrate asynchronous USB I/O into
777 * existing main loops, effectively allowing you to do some work behind
778 * libusb's back. If you do take libusb's file descriptors and pass them to
779 * poll()/select() yourself, you need to be aware of the associated issues.
781 * \section eventlock The events lock
783 * The first concept to be introduced is the events lock. The events lock
784 * is used to serialize threads that want to handle events, such that only
785 * one thread is handling events at any one time.
787 * You must take the events lock before polling libusb file descriptors,
788 * using libusb_lock_events(). You must release the lock as soon as you have
789 * aborted your poll()/select() loop, using libusb_unlock_events().
791 * \section threadwait Letting other threads do the work for you
793 * Although the events lock is a critical part of the solution, it is not
794 * enough on it's own. You might wonder if the following is sufficient...
796 libusb_lock_events(ctx);
798 poll(libusb file descriptors, 120*1000);
799 if (poll indicates activity)
800 libusb_handle_events_timeout(ctx, 0);
802 libusb_unlock_events(ctx);
804 * ...and the answer is that it is not. This is because the transfer in the
805 * code shown above may take a long time (say 30 seconds) to complete, and
806 * the lock is not released until the transfer is completed.
808 * Another thread with similar code that wants to do event handling may be
809 * working with a transfer that completes after a few milliseconds. Despite
810 * having such a quick completion time, the other thread cannot check that
811 * status of its transfer until the code above has finished (30 seconds later)
812 * due to contention on the lock.
814 * To solve this, libusb offers you a mechanism to determine when another
815 * thread is handling events. It also offers a mechanism to block your thread
816 * until the event handling thread has completed an event (and this mechanism
817 * does not involve polling of file descriptors).
819 * After determining that another thread is currently handling events, you
820 * obtain the <em>event waiters</em> lock using libusb_lock_event_waiters().
821 * You then re-check that some other thread is still handling events, and if
822 * so, you call libusb_wait_for_event().
824 * libusb_wait_for_event() puts your application to sleep until an event
825 * occurs, or until a thread releases the events lock. When either of these
826 * things happen, your thread is woken up, and should re-check the condition
827 * it was waiting on. It should also re-check that another thread is handling
828 * events, and if not, it should start handling events itself.
830 * This looks like the following, as pseudo-code:
833 if (libusb_try_lock_events(ctx) == 0) {
834 // we obtained the event lock: do our own event handling
836 if (!libusb_event_handling_ok(ctx)) {
837 libusb_unlock_events(ctx);
840 poll(libusb file descriptors, 120*1000);
841 if (poll indicates activity)
842 libusb_handle_events_locked(ctx, 0);
844 libusb_unlock_events(ctx);
846 // another thread is doing event handling. wait for it to signal us that
847 // an event has completed
848 libusb_lock_event_waiters(ctx);
851 // now that we have the event waiters lock, double check that another
852 // thread is still handling events for us. (it may have ceased handling
853 // events in the time it took us to reach this point)
854 if (!libusb_event_handler_active(ctx)) {
855 // whoever was handling events is no longer doing so, try again
856 libusb_unlock_event_waiters(ctx);
860 libusb_wait_for_event(ctx);
862 libusb_unlock_event_waiters(ctx);
864 printf("completed!\n");
867 * A naive look at the above code may suggest that this can only support
868 * one event waiter (hence a total of 2 competing threads, the other doing
869 * event handling), because the event waiter seems to have taken the event
870 * waiters lock while waiting for an event. However, the system does support
871 * multiple event waiters, because libusb_wait_for_event() actually drops
872 * the lock while waiting, and reaquires it before continuing.
874 * We have now implemented code which can dynamically handle situations where
875 * nobody is handling events (so we should do it ourselves), and it can also
876 * handle situations where another thread is doing event handling (so we can
877 * piggyback onto them). It is also equipped to handle a combination of
878 * the two, for example, another thread is doing event handling, but for
879 * whatever reason it stops doing so before our condition is met, so we take
880 * over the event handling.
882 * Four functions were introduced in the above pseudo-code. Their importance
883 * should be apparent from the code shown above.
884 * -# libusb_try_lock_events() is a non-blocking function which attempts
885 * to acquire the events lock but returns a failure code if it is contended.
886 * -# libusb_event_handling_ok() checks that libusb is still happy for your
887 * thread to be performing event handling. Sometimes, libusb needs to
888 * interrupt the event handler, and this is how you can check if you have
889 * been interrupted. If this function returns 0, the correct behaviour is
890 * for you to give up the event handling lock, and then to repeat the cycle.
891 * The following libusb_try_lock_events() will fail, so you will become an
892 * events waiter. For more information on this, read \ref fullstory below.
893 * -# libusb_handle_events_locked() is a variant of
894 * libusb_handle_events_timeout() that you can call while holding the
895 * events lock. libusb_handle_events_timeout() itself implements similar
896 * logic to the above, so be sure not to call it when you are
897 * "working behind libusb's back", as is the case here.
898 * -# libusb_event_handler_active() determines if someone is currently
899 * holding the events lock
901 * You might be wondering why there is no function to wake up all threads
902 * blocked on libusb_wait_for_event(). This is because libusb can do this
903 * internally: it will wake up all such threads when someone calls
904 * libusb_unlock_events() or when a transfer completes (at the point after its
905 * callback has returned).
907 * \subsection fullstory The full story
909 * The above explanation should be enough to get you going, but if you're
910 * really thinking through the issues then you may be left with some more
911 * questions regarding libusb's internals. If you're curious, read on, and if
912 * not, skip to the next section to avoid confusing yourself!
914 * The immediate question that may spring to mind is: what if one thread
915 * modifies the set of file descriptors that need to be polled while another
916 * thread is doing event handling?
918 * There are 2 situations in which this may happen.
919 * -# libusb_open() will add another file descriptor to the poll set,
920 * therefore it is desirable to interrupt the event handler so that it
921 * restarts, picking up the new descriptor.
922 * -# libusb_close() will remove a file descriptor from the poll set. There
923 * are all kinds of race conditions that could arise here, so it is
924 * important that nobody is doing event handling at this time.
926 * libusb handles these issues internally, so application developers do not
927 * have to stop their event handlers while opening/closing devices. Here's how
928 * it works, focusing on the libusb_close() situation first:
930 * -# During initialization, libusb opens an internal pipe, and it adds the read
931 * end of this pipe to the set of file descriptors to be polled.
932 * -# During libusb_close(), libusb writes some dummy data on this control pipe.
933 * This immediately interrupts the event handler. libusb also records
934 * internally that it is trying to interrupt event handlers for this
935 * high-priority event.
936 * -# At this point, some of the functions described above start behaving
938 * - libusb_event_handling_ok() starts returning 1, indicating that it is NOT
939 * OK for event handling to continue.
940 * - libusb_try_lock_events() starts returning 1, indicating that another
941 * thread holds the event handling lock, even if the lock is uncontended.
942 * - libusb_event_handler_active() starts returning 1, indicating that
943 * another thread is doing event handling, even if that is not true.
944 * -# The above changes in behaviour result in the event handler stopping and
945 * giving up the events lock very quickly, giving the high-priority
946 * libusb_close() operation a "free ride" to acquire the events lock. All
947 * threads that are competing to do event handling become event waiters.
948 * -# With the events lock held inside libusb_close(), libusb can safely remove
949 * a file descriptor from the poll set, in the safety of knowledge that
950 * nobody is polling those descriptors or trying to access the poll set.
951 * -# After obtaining the events lock, the close operation completes very
952 * quickly (usually a matter of milliseconds) and then immediately releases
954 * -# At the same time, the behaviour of libusb_event_handling_ok() and friends
955 * reverts to the original, documented behaviour.
956 * -# The release of the events lock causes the threads that are waiting for
957 * events to be woken up and to start competing to become event handlers
958 * again. One of them will succeed; it will then re-obtain the list of poll
959 * descriptors, and USB I/O will then continue as normal.
961 * libusb_open() is similar, and is actually a more simplistic case. Upon a
962 * call to libusb_open():
964 * -# The device is opened and a file descriptor is added to the poll set.
965 * -# libusb sends some dummy data on the control pipe, and records that it
966 * is trying to modify the poll descriptor set.
967 * -# The event handler is interrupted, and the same behaviour change as for
968 * libusb_close() takes effect, causing all event handling threads to become
970 * -# The libusb_open() implementation takes its free ride to the events lock.
971 * -# Happy that it has successfully paused the events handler, libusb_open()
972 * releases the events lock.
973 * -# The event waiter threads are all woken up and compete to become event
974 * handlers again. The one that succeeds will obtain the list of poll
975 * descriptors again, which will include the addition of the new device.
977 * \subsection concl Closing remarks
979 * The above may seem a little complicated, but hopefully I have made it clear
980 * why such complications are necessary. Also, do not forget that this only
981 * applies to applications that take libusb's file descriptors and integrate
982 * them into their own polling loops.
984 * You may decide that it is OK for your multi-threaded application to ignore
985 * some of the rules and locks detailed above, because you don't think that
986 * two threads can ever be polling the descriptors at the same time. If that
987 * is the case, then that's good news for you because you don't have to worry.
988 * But be careful here; remember that the synchronous I/O functions do event
989 * handling internally. If you have one thread doing event handling in a loop
990 * (without implementing the rules and locking semantics documented above)
991 * and another trying to send a synchronous USB transfer, you will end up with
992 * two threads monitoring the same descriptors, and the above-described
993 * undesirable behaviour occuring. The solution is for your polling thread to
994 * play by the rules; the synchronous I/O functions do so, and this will result
995 * in them getting along in perfect harmony.
997 * If you do have a dedicated thread doing event handling, it is perfectly
998 * legal for it to take the event handling lock for long periods of time. Any
999 * synchronous I/O functions you call from other threads will transparently
1000 * fall back to the "event waiters" mechanism detailed above. The only
1001 * consideration that your event handling thread must apply is the one related
1002 * to libusb_event_handling_ok(): you must call this before every poll(), and
1003 * give up the events lock if instructed.
1006 int usbi_io_init(struct libusb_context *ctx)
1010 usbi_mutex_init(&ctx->flying_transfers_lock, NULL);
1011 usbi_mutex_init(&ctx->pollfds_lock, NULL);
1012 usbi_mutex_init(&ctx->pollfd_modify_lock, NULL);
1013 usbi_mutex_init(&ctx->events_lock, NULL);
1014 usbi_mutex_init(&ctx->event_waiters_lock, NULL);
1015 usbi_cond_init(&ctx->event_waiters_cond, NULL);
1016 list_init(&ctx->flying_transfers);
1017 list_init(&ctx->pollfds);
1019 /* FIXME should use an eventfd on kernels that support it */
1020 r = usbi_pipe(ctx->ctrl_pipe);
1022 r = LIBUSB_ERROR_OTHER;
1026 r = usbi_add_pollfd(ctx, ctx->ctrl_pipe[0], POLLIN);
1028 goto err_close_pipe;
1030 #ifdef USBI_TIMERFD_AVAILABLE
1031 ctx->timerfd = timerfd_create(usbi_backend->get_timerfd_clockid(),
1033 if (ctx->timerfd >= 0) {
1034 usbi_dbg("using timerfd for timeouts");
1035 r = usbi_add_pollfd(ctx, ctx->timerfd, POLLIN);
1037 usbi_remove_pollfd(ctx, ctx->ctrl_pipe[0]);
1038 close(ctx->timerfd);
1039 goto err_close_pipe;
1042 usbi_dbg("timerfd not available (code %d error %d)", ctx->timerfd, errno);
1050 usbi_close(ctx->ctrl_pipe[0]);
1051 usbi_close(ctx->ctrl_pipe[1]);
1053 usbi_mutex_destroy(&ctx->flying_transfers_lock);
1054 usbi_mutex_destroy(&ctx->pollfds_lock);
1055 usbi_mutex_destroy(&ctx->pollfd_modify_lock);
1056 usbi_mutex_destroy(&ctx->events_lock);
1057 usbi_mutex_destroy(&ctx->event_waiters_lock);
1058 usbi_cond_destroy(&ctx->event_waiters_cond);
1062 void usbi_io_exit(struct libusb_context *ctx)
1064 usbi_remove_pollfd(ctx, ctx->ctrl_pipe[0]);
1065 usbi_close(ctx->ctrl_pipe[0]);
1066 usbi_close(ctx->ctrl_pipe[1]);
1067 #ifdef USBI_TIMERFD_AVAILABLE
1068 if (usbi_using_timerfd(ctx)) {
1069 usbi_remove_pollfd(ctx, ctx->timerfd);
1070 close(ctx->timerfd);
1073 usbi_mutex_destroy(&ctx->flying_transfers_lock);
1074 usbi_mutex_destroy(&ctx->pollfds_lock);
1075 usbi_mutex_destroy(&ctx->pollfd_modify_lock);
1076 usbi_mutex_destroy(&ctx->events_lock);
1077 usbi_mutex_destroy(&ctx->event_waiters_lock);
1078 usbi_cond_destroy(&ctx->event_waiters_cond);
1081 static int calculate_timeout(struct usbi_transfer *transfer)
1084 struct timespec current_time;
1085 unsigned int timeout =
1086 __USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout;
1091 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, ¤t_time);
1093 usbi_err(ITRANSFER_CTX(transfer),
1094 "failed to read monotonic clock, errno=%d", errno);
1098 current_time.tv_sec += timeout / 1000;
1099 current_time.tv_nsec += (timeout % 1000) * 1000000;
1101 if (current_time.tv_nsec > 1000000000) {
1102 current_time.tv_nsec -= 1000000000;
1103 current_time.tv_sec++;
1106 TIMESPEC_TO_TIMEVAL(&transfer->timeout, ¤t_time);
1110 /* add a transfer to the (timeout-sorted) active transfers list.
1111 * returns 1 if the transfer has a timeout and it is the timeout next to
1113 static int add_to_flying_list(struct usbi_transfer *transfer)
1115 struct usbi_transfer *cur;
1116 struct timeval *timeout = &transfer->timeout;
1117 struct libusb_context *ctx = ITRANSFER_CTX(transfer);
1121 usbi_mutex_lock(&ctx->flying_transfers_lock);
1123 /* if we have no other flying transfers, start the list with this one */
1124 if (list_empty(&ctx->flying_transfers)) {
1125 list_add(&transfer->list, &ctx->flying_transfers);
1126 if (timerisset(timeout))
1131 /* if we have infinite timeout, append to end of list */
1132 if (!timerisset(timeout)) {
1133 list_add_tail(&transfer->list, &ctx->flying_transfers);
1137 /* otherwise, find appropriate place in list */
1138 list_for_each_entry(cur, &ctx->flying_transfers, list, struct usbi_transfer) {
1139 /* find first timeout that occurs after the transfer in question */
1140 struct timeval *cur_tv = &cur->timeout;
1142 if (!timerisset(cur_tv) || (cur_tv->tv_sec > timeout->tv_sec) ||
1143 (cur_tv->tv_sec == timeout->tv_sec &&
1144 cur_tv->tv_usec > timeout->tv_usec)) {
1145 list_add_tail(&transfer->list, &cur->list);
1152 /* otherwise we need to be inserted at the end */
1153 list_add_tail(&transfer->list, &ctx->flying_transfers);
1155 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1159 /** \ingroup asyncio
1160 * Allocate a libusb transfer with a specified number of isochronous packet
1161 * descriptors. The returned transfer is pre-initialized for you. When the new
1162 * transfer is no longer needed, it should be freed with
1163 * libusb_free_transfer().
1165 * Transfers intended for non-isochronous endpoints (e.g. control, bulk,
1166 * interrupt) should specify an iso_packets count of zero.
1168 * For transfers intended for isochronous endpoints, specify an appropriate
1169 * number of packet descriptors to be allocated as part of the transfer.
1170 * The returned transfer is not specially initialized for isochronous I/O;
1171 * you are still required to set the
1172 * \ref libusb_transfer::num_iso_packets "num_iso_packets" and
1173 * \ref libusb_transfer::type "type" fields accordingly.
1175 * It is safe to allocate a transfer with some isochronous packets and then
1176 * use it on a non-isochronous endpoint. If you do this, ensure that at time
1177 * of submission, num_iso_packets is 0 and that type is set appropriately.
1179 * \param iso_packets number of isochronous packet descriptors to allocate
1180 * \returns a newly allocated transfer, or NULL on error
1182 API_EXPORTED struct libusb_transfer *libusb_alloc_transfer(int iso_packets)
1184 size_t os_alloc_size = usbi_backend->transfer_priv_size
1185 + (usbi_backend->add_iso_packet_size * iso_packets);
1186 size_t alloc_size = sizeof(struct usbi_transfer)
1187 + sizeof(struct libusb_transfer)
1188 + (sizeof(struct libusb_iso_packet_descriptor) * iso_packets)
1190 struct usbi_transfer *itransfer = malloc(alloc_size);
1194 memset(itransfer, 0, alloc_size);
1195 itransfer->num_iso_packets = iso_packets;
1196 usbi_mutex_init(&itransfer->lock, NULL);
1197 return __USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1200 /** \ingroup asyncio
1201 * Free a transfer structure. This should be called for all transfers
1202 * allocated with libusb_alloc_transfer().
1204 * If the \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_BUFFER
1205 * "LIBUSB_TRANSFER_FREE_BUFFER" flag is set and the transfer buffer is
1206 * non-NULL, this function will also free the transfer buffer using the
1207 * standard system memory allocator (e.g. free()).
1209 * It is legal to call this function with a NULL transfer. In this case,
1210 * the function will simply return safely.
1212 * It is not legal to free an active transfer (one which has been submitted
1213 * and has not yet completed).
1215 * \param transfer the transfer to free
1217 API_EXPORTED void libusb_free_transfer(struct libusb_transfer *transfer)
1219 struct usbi_transfer *itransfer;
1223 if (transfer->flags & LIBUSB_TRANSFER_FREE_BUFFER && transfer->buffer)
1224 free(transfer->buffer);
1226 itransfer = __LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1227 usbi_mutex_destroy(&itransfer->lock);
1231 /** \ingroup asyncio
1232 * Submit a transfer. This function will fire off the USB transfer and then
1233 * return immediately.
1235 * \param transfer the transfer to submit
1236 * \returns 0 on success
1237 * \returns LIBUSB_ERROR_NO_DEVICE if the device has been disconnected
1238 * \returns LIBUSB_ERROR_BUSY if the transfer has already been submitted.
1239 * \returns another LIBUSB_ERROR code on other failure
1241 API_EXPORTED int libusb_submit_transfer(struct libusb_transfer *transfer)
1243 struct libusb_context *ctx = TRANSFER_CTX(transfer);
1244 struct usbi_transfer *itransfer =
1245 __LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1249 usbi_mutex_lock(&itransfer->lock);
1250 itransfer->transferred = 0;
1251 itransfer->flags = 0;
1252 r = calculate_timeout(itransfer);
1254 r = LIBUSB_ERROR_OTHER;
1258 first = add_to_flying_list(itransfer);
1259 r = usbi_backend->submit_transfer(itransfer);
1261 usbi_mutex_lock(&ctx->flying_transfers_lock);
1262 list_del(&itransfer->list);
1263 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1265 #ifdef USBI_TIMERFD_AVAILABLE
1266 else if (first && usbi_using_timerfd(ctx)) {
1267 /* if this transfer has the lowest timeout of all active transfers,
1268 * rearm the timerfd with this transfer's timeout */
1269 const struct itimerspec it = { {0, 0},
1270 { itransfer->timeout.tv_sec, itransfer->timeout.tv_usec * 1000 } };
1271 usbi_dbg("arm timerfd for timeout in %dms (first in line)", transfer->timeout);
1272 r = timerfd_settime(ctx->timerfd, TFD_TIMER_ABSTIME, &it, NULL);
1274 r = LIBUSB_ERROR_OTHER;
1279 usbi_mutex_unlock(&itransfer->lock);
1283 /** \ingroup asyncio
1284 * Asynchronously cancel a previously submitted transfer.
1285 * This function returns immediately, but this does not indicate cancellation
1286 * is complete. Your callback function will be invoked at some later time
1287 * with a transfer status of
1288 * \ref libusb_transfer_status::LIBUSB_TRANSFER_CANCELLED
1289 * "LIBUSB_TRANSFER_CANCELLED."
1291 * \param transfer the transfer to cancel
1292 * \returns 0 on success
1293 * \returns LIBUSB_ERROR_NOT_FOUND if the transfer is already complete or
1295 * \returns a LIBUSB_ERROR code on failure
1297 API_EXPORTED int libusb_cancel_transfer(struct libusb_transfer *transfer)
1299 struct usbi_transfer *itransfer =
1300 __LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1304 usbi_mutex_lock(&itransfer->lock);
1305 r = usbi_backend->cancel_transfer(itransfer);
1307 usbi_err(TRANSFER_CTX(transfer),
1308 "cancel transfer failed error %d", r);
1309 usbi_mutex_unlock(&itransfer->lock);
1313 #ifdef USBI_TIMERFD_AVAILABLE
1314 static int disarm_timerfd(struct libusb_context *ctx)
1316 const struct itimerspec disarm_timer = { { 0, 0 }, { 0, 0 } };
1320 r = timerfd_settime(ctx->timerfd, 0, &disarm_timer, NULL);
1322 return LIBUSB_ERROR_OTHER;
1327 /* iterates through the flying transfers, and rearms the timerfd based on the
1328 * next upcoming timeout.
1329 * must be called with flying_list locked.
1330 * returns 0 if there was no timeout to arm, 1 if the next timeout was armed,
1331 * or a LIBUSB_ERROR code on failure.
1333 static int arm_timerfd_for_next_timeout(struct libusb_context *ctx)
1335 struct usbi_transfer *transfer;
1337 list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
1338 struct timeval *cur_tv = &transfer->timeout;
1340 /* if we've reached transfers of infinite timeout, then we have no
1342 if (!timerisset(cur_tv))
1345 /* act on first transfer that is not already cancelled */
1346 if (!(transfer->flags & USBI_TRANSFER_TIMED_OUT)) {
1348 const struct itimerspec it = { {0, 0},
1349 { cur_tv->tv_sec, cur_tv->tv_usec * 1000 } };
1350 usbi_dbg("next timeout originally %dms", __USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout);
1351 r = timerfd_settime(ctx->timerfd, TFD_TIMER_ABSTIME, &it, NULL);
1353 return LIBUSB_ERROR_OTHER;
1361 static int disarm_timerfd(struct libusb_context *ctx)
1365 static int arm_timerfd_for_next_timeout(struct libusb_context *ctx)
1371 /* Handle completion of a transfer (completion might be an error condition).
1372 * This will invoke the user-supplied callback function, which may end up
1373 * freeing the transfer. Therefore you cannot use the transfer structure
1374 * after calling this function, and you should free all backend-specific
1375 * data before calling it.
1376 * Do not call this function with the usbi_transfer lock held. User-specified
1377 * callback functions may attempt to directly resubmit the transfer, which
1378 * will attempt to take the lock. */
1379 int usbi_handle_transfer_completion(struct usbi_transfer *itransfer,
1380 enum libusb_transfer_status status)
1382 struct libusb_transfer *transfer =
1383 __USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1384 struct libusb_context *ctx = TRANSFER_CTX(transfer);
1388 /* FIXME: could be more intelligent with the timerfd here. we don't need
1389 * to disarm the timerfd if there was no timer running, and we only need
1390 * to rearm the timerfd if the transfer that expired was the one with
1391 * the shortest timeout. */
1393 usbi_mutex_lock(&ctx->flying_transfers_lock);
1394 list_del(&itransfer->list);
1395 r = arm_timerfd_for_next_timeout(ctx);
1396 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1400 } else if (r == 0) {
1401 r = disarm_timerfd(ctx);
1406 if (status == LIBUSB_TRANSFER_COMPLETED
1407 && transfer->flags & LIBUSB_TRANSFER_SHORT_NOT_OK) {
1408 int rqlen = transfer->length;
1409 if (transfer->type == LIBUSB_TRANSFER_TYPE_CONTROL)
1410 rqlen -= LIBUSB_CONTROL_SETUP_SIZE;
1411 if (rqlen != itransfer->transferred) {
1412 usbi_dbg("interpreting short transfer as error");
1413 status = LIBUSB_TRANSFER_ERROR;
1417 flags = transfer->flags;
1418 transfer->status = status;
1419 transfer->actual_length = itransfer->transferred;
1420 if (transfer->callback)
1421 transfer->callback(transfer);
1422 /* transfer might have been freed by the above call, do not use from
1424 if (flags & LIBUSB_TRANSFER_FREE_TRANSFER)
1425 libusb_free_transfer(transfer);
1426 usbi_mutex_lock(&ctx->event_waiters_lock);
1427 usbi_cond_broadcast(&ctx->event_waiters_cond);
1428 usbi_mutex_unlock(&ctx->event_waiters_lock);
1432 /* Similar to usbi_handle_transfer_completion() but exclusively for transfers
1433 * that were asynchronously cancelled. The same concerns w.r.t. freeing of
1434 * transfers exist here.
1435 * Do not call this function with the usbi_transfer lock held. User-specified
1436 * callback functions may attempt to directly resubmit the transfer, which
1437 * will attempt to take the lock. */
1438 int usbi_handle_transfer_cancellation(struct usbi_transfer *transfer)
1440 /* if the URB was cancelled due to timeout, report timeout to the user */
1441 if (transfer->flags & USBI_TRANSFER_TIMED_OUT) {
1442 usbi_dbg("detected timeout cancellation");
1443 return usbi_handle_transfer_completion(transfer, LIBUSB_TRANSFER_TIMED_OUT);
1446 /* otherwise its a normal async cancel */
1447 return usbi_handle_transfer_completion(transfer, LIBUSB_TRANSFER_CANCELLED);
1451 * Attempt to acquire the event handling lock. This lock is used to ensure that
1452 * only one thread is monitoring libusb event sources at any one time.
1454 * You only need to use this lock if you are developing an application
1455 * which calls poll() or select() on libusb's file descriptors directly.
1456 * If you stick to libusb's event handling loop functions (e.g.
1457 * libusb_handle_events()) then you do not need to be concerned with this
1460 * While holding this lock, you are trusted to actually be handling events.
1461 * If you are no longer handling events, you must call libusb_unlock_events()
1462 * as soon as possible.
1464 * \param ctx the context to operate on, or NULL for the default context
1465 * \returns 0 if the lock was obtained successfully
1466 * \returns 1 if the lock was not obtained (i.e. another thread holds the lock)
1469 API_EXPORTED int libusb_try_lock_events(libusb_context *ctx)
1472 USBI_GET_CONTEXT(ctx);
1474 /* is someone else waiting to modify poll fds? if so, don't let this thread
1475 * start event handling */
1476 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1477 r = ctx->pollfd_modify;
1478 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1480 usbi_dbg("someone else is modifying poll fds");
1484 r = usbi_mutex_trylock(&ctx->events_lock);
1488 ctx->event_handler_active = 1;
1493 * Acquire the event handling lock, blocking until successful acquisition if
1494 * it is contended. This lock is used to ensure that only one thread is
1495 * monitoring libusb event sources at any one time.
1497 * You only need to use this lock if you are developing an application
1498 * which calls poll() or select() on libusb's file descriptors directly.
1499 * If you stick to libusb's event handling loop functions (e.g.
1500 * libusb_handle_events()) then you do not need to be concerned with this
1503 * While holding this lock, you are trusted to actually be handling events.
1504 * If you are no longer handling events, you must call libusb_unlock_events()
1505 * as soon as possible.
1507 * \param ctx the context to operate on, or NULL for the default context
1510 API_EXPORTED void libusb_lock_events(libusb_context *ctx)
1512 USBI_GET_CONTEXT(ctx);
1513 usbi_mutex_lock(&ctx->events_lock);
1514 ctx->event_handler_active = 1;
1518 * Release the lock previously acquired with libusb_try_lock_events() or
1519 * libusb_lock_events(). Releasing this lock will wake up any threads blocked
1520 * on libusb_wait_for_event().
1522 * \param ctx the context to operate on, or NULL for the default context
1525 API_EXPORTED void libusb_unlock_events(libusb_context *ctx)
1527 USBI_GET_CONTEXT(ctx);
1528 ctx->event_handler_active = 0;
1529 usbi_mutex_unlock(&ctx->events_lock);
1531 /* FIXME: perhaps we should be a bit more efficient by not broadcasting
1532 * the availability of the events lock when we are modifying pollfds
1533 * (check ctx->pollfd_modify)? */
1534 usbi_mutex_lock(&ctx->event_waiters_lock);
1535 usbi_cond_broadcast(&ctx->event_waiters_cond);
1536 usbi_mutex_unlock(&ctx->event_waiters_lock);
1540 * Determine if it is still OK for this thread to be doing event handling.
1542 * Sometimes, libusb needs to temporarily pause all event handlers, and this
1543 * is the function you should use before polling file descriptors to see if
1546 * If this function instructs your thread to give up the events lock, you
1547 * should just continue the usual logic that is documented in \ref mtasync.
1548 * On the next iteration, your thread will fail to obtain the events lock,
1549 * and will hence become an event waiter.
1551 * This function should be called while the events lock is held: you don't
1552 * need to worry about the results of this function if your thread is not
1553 * the current event handler.
1555 * \param ctx the context to operate on, or NULL for the default context
1556 * \returns 1 if event handling can start or continue
1557 * \returns 0 if this thread must give up the events lock
1558 * \see \ref fullstory "Multi-threaded I/O: the full story"
1560 API_EXPORTED int libusb_event_handling_ok(libusb_context *ctx)
1563 USBI_GET_CONTEXT(ctx);
1565 /* is someone else waiting to modify poll fds? if so, don't let this thread
1566 * continue event handling */
1567 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1568 r = ctx->pollfd_modify;
1569 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1571 usbi_dbg("someone else is modifying poll fds");
1580 * Determine if an active thread is handling events (i.e. if anyone is holding
1581 * the event handling lock).
1583 * \param ctx the context to operate on, or NULL for the default context
1584 * \returns 1 if a thread is handling events
1585 * \returns 0 if there are no threads currently handling events
1588 API_EXPORTED int libusb_event_handler_active(libusb_context *ctx)
1591 USBI_GET_CONTEXT(ctx);
1593 /* is someone else waiting to modify poll fds? if so, don't let this thread
1594 * start event handling -- indicate that event handling is happening */
1595 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1596 r = ctx->pollfd_modify;
1597 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1599 usbi_dbg("someone else is modifying poll fds");
1603 return ctx->event_handler_active;
1607 * Acquire the event waiters lock. This lock is designed to be obtained under
1608 * the situation where you want to be aware when events are completed, but
1609 * some other thread is event handling so calling libusb_handle_events() is not
1612 * You then obtain this lock, re-check that another thread is still handling
1613 * events, then call libusb_wait_for_event().
1615 * You only need to use this lock if you are developing an application
1616 * which calls poll() or select() on libusb's file descriptors directly,
1617 * <b>and</b> may potentially be handling events from 2 threads simultaenously.
1618 * If you stick to libusb's event handling loop functions (e.g.
1619 * libusb_handle_events()) then you do not need to be concerned with this
1622 * \param ctx the context to operate on, or NULL for the default context
1625 API_EXPORTED void libusb_lock_event_waiters(libusb_context *ctx)
1627 USBI_GET_CONTEXT(ctx);
1628 usbi_mutex_lock(&ctx->event_waiters_lock);
1632 * Release the event waiters lock.
1633 * \param ctx the context to operate on, or NULL for the default context
1636 API_EXPORTED void libusb_unlock_event_waiters(libusb_context *ctx)
1638 USBI_GET_CONTEXT(ctx);
1639 usbi_mutex_unlock(&ctx->event_waiters_lock);
1643 * Wait for another thread to signal completion of an event. Must be called
1644 * with the event waiters lock held, see libusb_lock_event_waiters().
1646 * This function will block until any of the following conditions are met:
1647 * -# The timeout expires
1648 * -# A transfer completes
1649 * -# A thread releases the event handling lock through libusb_unlock_events()
1651 * Condition 1 is obvious. Condition 2 unblocks your thread <em>after</em>
1652 * the callback for the transfer has completed. Condition 3 is important
1653 * because it means that the thread that was previously handling events is no
1654 * longer doing so, so if any events are to complete, another thread needs to
1655 * step up and start event handling.
1657 * This function releases the event waiters lock before putting your thread
1658 * to sleep, and reacquires the lock as it is being woken up.
1660 * \param ctx the context to operate on, or NULL for the default context
1661 * \param tv maximum timeout for this blocking function. A NULL value
1662 * indicates unlimited timeout.
1663 * \returns 0 after a transfer completes or another thread stops event handling
1664 * \returns 1 if the timeout expired
1667 API_EXPORTED int libusb_wait_for_event(libusb_context *ctx, struct timeval *tv)
1669 struct timespec timeout;
1672 USBI_GET_CONTEXT(ctx);
1674 usbi_cond_wait(&ctx->event_waiters_cond, &ctx->event_waiters_lock);
1678 r = usbi_backend->clock_gettime(USBI_CLOCK_REALTIME, &timeout);
1680 usbi_err(ctx, "failed to read realtime clock, error %d", errno);
1681 return LIBUSB_ERROR_OTHER;
1684 timeout.tv_sec += tv->tv_sec;
1685 timeout.tv_nsec += tv->tv_usec * 1000;
1686 if (timeout.tv_nsec > 1000000000) {
1687 timeout.tv_nsec -= 1000000000;
1691 r = usbi_cond_timedwait(&ctx->event_waiters_cond,
1692 &ctx->event_waiters_lock, &timeout);
1693 return (r == ETIMEDOUT);
1696 static void handle_timeout(struct usbi_transfer *itransfer)
1698 struct libusb_transfer *transfer =
1699 __USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1702 itransfer->flags |= USBI_TRANSFER_TIMED_OUT;
1703 r = libusb_cancel_transfer(transfer);
1705 usbi_warn(TRANSFER_CTX(transfer),
1706 "async cancel failed %d errno=%d", r, errno);
1709 #ifdef USBI_OS_HANDLES_TIMEOUT
1710 static int handle_timeouts_locked(struct libusb_context *ctx)
1714 static int handle_timeouts(struct libusb_context *ctx)
1719 static int handle_timeouts_locked(struct libusb_context *ctx)
1722 struct timespec systime_ts;
1723 struct timeval systime;
1724 struct usbi_transfer *transfer;
1726 if (list_empty(&ctx->flying_transfers))
1729 /* get current time */
1730 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, &systime_ts);
1734 TIMESPEC_TO_TIMEVAL(&systime, &systime_ts);
1736 /* iterate through flying transfers list, finding all transfers that
1737 * have expired timeouts */
1738 list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
1739 struct timeval *cur_tv = &transfer->timeout;
1741 /* if we've reached transfers of infinite timeout, we're all done */
1742 if (!timerisset(cur_tv))
1745 /* ignore timeouts we've already handled */
1746 if (transfer->flags & USBI_TRANSFER_TIMED_OUT)
1749 /* if transfer has non-expired timeout, nothing more to do */
1750 if ((cur_tv->tv_sec > systime.tv_sec) ||
1751 (cur_tv->tv_sec == systime.tv_sec &&
1752 cur_tv->tv_usec > systime.tv_usec))
1755 /* otherwise, we've got an expired timeout to handle */
1756 handle_timeout(transfer);
1761 static int handle_timeouts(struct libusb_context *ctx)
1764 USBI_GET_CONTEXT(ctx);
1765 usbi_mutex_lock(&ctx->flying_transfers_lock);
1766 r = handle_timeouts_locked(ctx);
1767 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1772 #ifdef USBI_TIMERFD_AVAILABLE
1773 static int handle_timerfd_trigger(struct libusb_context *ctx)
1777 r = disarm_timerfd(ctx);
1781 usbi_mutex_lock(&ctx->flying_transfers_lock);
1783 /* process the timeout that just happened */
1784 r = handle_timeouts_locked(ctx);
1788 /* arm for next timeout*/
1789 r = arm_timerfd_for_next_timeout(ctx);
1792 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1797 /* do the actual event handling. assumes that no other thread is concurrently
1798 * doing the same thing. */
1799 static int handle_events(struct libusb_context *ctx, struct timeval *tv)
1802 struct usbi_pollfd *ipollfd;
1808 usbi_mutex_lock(&ctx->pollfds_lock);
1809 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
1812 /* TODO: malloc when number of fd's changes, not on every poll */
1813 fds = malloc(sizeof(*fds) * nfds);
1815 usbi_mutex_unlock(&ctx->pollfds_lock);
1816 return LIBUSB_ERROR_NO_MEM;
1819 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd) {
1820 struct libusb_pollfd *pollfd = &ipollfd->pollfd;
1821 int fd = pollfd->fd;
1824 fds[i].events = pollfd->events;
1827 usbi_mutex_unlock(&ctx->pollfds_lock);
1829 timeout_ms = (tv->tv_sec * 1000) + (tv->tv_usec / 1000);
1831 /* round up to next millisecond */
1832 if (tv->tv_usec % 1000)
1835 usbi_dbg("poll() %d fds with timeout in %dms", nfds, timeout_ms);
1836 r = usbi_poll(fds, nfds, timeout_ms);
1837 usbi_dbg("poll() returned %d", r);
1840 return handle_timeouts(ctx);
1841 } else if (r == -1 && errno == EINTR) {
1843 return LIBUSB_ERROR_INTERRUPTED;
1846 usbi_err(ctx, "poll failed %d err=%d\n", r, errno);
1847 return LIBUSB_ERROR_IO;
1850 /* fd[0] is always the ctrl pipe */
1851 if (fds[0].revents) {
1852 /* another thread wanted to interrupt event handling, and it succeeded!
1853 * handle any other events that cropped up at the same time, and
1855 usbi_dbg("caught a fish on the control pipe");
1861 /* prevent OS backend from trying to handle events on ctrl pipe */
1867 #ifdef USBI_TIMERFD_AVAILABLE
1868 /* on timerfd configurations, fds[1] is the timerfd */
1869 if (usbi_using_timerfd(ctx) && fds[1].revents) {
1870 /* timerfd indicates that a timeout has expired */
1872 usbi_dbg("timerfd triggered");
1874 ret = handle_timerfd_trigger(ctx);
1876 /* return error code */
1879 } else if (r == 1) {
1880 /* no more active file descriptors, nothing more to do */
1884 /* more events pending...
1885 * prevent OS backend from trying to handle events on timerfd */
1892 r = usbi_backend->handle_events(ctx, fds, nfds, r);
1894 usbi_err(ctx, "backend handle_events failed with error %d", r);
1901 /* returns the smallest of:
1902 * 1. timeout of next URB
1903 * 2. user-supplied timeout
1904 * returns 1 if there is an already-expired timeout, otherwise returns 0
1907 static int get_next_timeout(libusb_context *ctx, struct timeval *tv,
1908 struct timeval *out)
1910 struct timeval timeout;
1911 int r = libusb_get_next_timeout(ctx, &timeout);
1913 /* timeout already expired? */
1914 if (!timerisset(&timeout))
1917 /* choose the smallest of next URB timeout or user specified timeout */
1918 if (timercmp(&timeout, tv, <))
1929 * Handle any pending events.
1931 * libusb determines "pending events" by checking if any timeouts have expired
1932 * and by checking the set of file descriptors for activity.
1934 * If a zero timeval is passed, this function will handle any already-pending
1935 * events and then immediately return in non-blocking style.
1937 * If a non-zero timeval is passed and no events are currently pending, this
1938 * function will block waiting for events to handle up until the specified
1939 * timeout. If an event arrives or a signal is raised, this function will
1942 * \param ctx the context to operate on, or NULL for the default context
1943 * \param tv the maximum time to block waiting for events, or zero for
1945 * \returns 0 on success, or a LIBUSB_ERROR code on failure
1947 API_EXPORTED int libusb_handle_events_timeout(libusb_context *ctx,
1951 struct timeval poll_timeout;
1953 USBI_GET_CONTEXT(ctx);
1954 r = get_next_timeout(ctx, tv, &poll_timeout);
1956 /* timeout already expired */
1957 return handle_timeouts(ctx);
1961 if (libusb_try_lock_events(ctx) == 0) {
1962 /* we obtained the event lock: do our own event handling */
1963 r = handle_events(ctx, &poll_timeout);
1964 libusb_unlock_events(ctx);
1968 /* another thread is doing event handling. wait for pthread events that
1969 * notify event completion. */
1970 libusb_lock_event_waiters(ctx);
1972 if (!libusb_event_handler_active(ctx)) {
1973 /* we hit a race: whoever was event handling earlier finished in the
1974 * time it took us to reach this point. try the cycle again. */
1975 libusb_unlock_event_waiters(ctx);
1976 usbi_dbg("event handler was active but went away, retrying");
1980 usbi_dbg("another thread is doing event handling");
1981 r = libusb_wait_for_event(ctx, &poll_timeout);
1982 libusb_unlock_event_waiters(ctx);
1987 return handle_timeouts(ctx);
1993 * Handle any pending events in blocking mode. There is currently a timeout
1994 * hardcoded at 60 seconds but we plan to make it unlimited in future. For
1995 * finer control over whether this function is blocking or non-blocking, or
1996 * for control over the timeout, use libusb_handle_events_timeout() instead.
1998 * \param ctx the context to operate on, or NULL for the default context
1999 * \returns 0 on success, or a LIBUSB_ERROR code on failure
2001 API_EXPORTED int libusb_handle_events(libusb_context *ctx)
2006 return libusb_handle_events_timeout(ctx, &tv);
2010 * Handle any pending events by polling file descriptors, without checking if
2011 * any other threads are already doing so. Must be called with the event lock
2012 * held, see libusb_lock_events().
2014 * This function is designed to be called under the situation where you have
2015 * taken the event lock and are calling poll()/select() directly on libusb's
2016 * file descriptors (as opposed to using libusb_handle_events() or similar).
2017 * You detect events on libusb's descriptors, so you then call this function
2018 * with a zero timeout value (while still holding the event lock).
2020 * \param ctx the context to operate on, or NULL for the default context
2021 * \param tv the maximum time to block waiting for events, or zero for
2023 * \returns 0 on success, or a LIBUSB_ERROR code on failure
2026 API_EXPORTED int libusb_handle_events_locked(libusb_context *ctx,
2030 struct timeval poll_timeout;
2032 USBI_GET_CONTEXT(ctx);
2033 r = get_next_timeout(ctx, tv, &poll_timeout);
2035 /* timeout already expired */
2036 return handle_timeouts(ctx);
2039 return handle_events(ctx, &poll_timeout);
2043 * Determines whether your application must apply special timing considerations
2044 * when monitoring libusb's file descriptors.
2046 * This function is only useful for applications which retrieve and poll
2047 * libusb's file descriptors in their own main loop (\ref pollmain).
2049 * Ordinarily, libusb's event handler needs to be called into at specific
2050 * moments in time (in addition to times when there is activity on the file
2051 * descriptor set). The usual approach is to use libusb_get_next_timeout()
2052 * to learn about when the next timeout occurs, and to adjust your
2053 * poll()/select() timeout accordingly so that you can make a call into the
2054 * library at that time.
2056 * Some platforms supported by libusb do not come with this baggage - any
2057 * events relevant to timing will be represented by activity on the file
2058 * descriptor set, and libusb_get_next_timeout() will always return 0.
2059 * This function allows you to detect whether you are running on such a
2064 * \param ctx the context to operate on, or NULL for the default context
2065 * \returns 0 if you must call into libusb at times determined by
2066 * libusb_get_next_timeout(), or 1 if all timeout events are handled internally
2067 * or through regular activity on the file descriptors.
2068 * \see \ref pollmain "Polling libusb file descriptors for event handling"
2070 API_EXPORTED int libusb_pollfds_handle_timeouts(libusb_context *ctx)
2072 #if defined(USBI_OS_HANDLES_TIMEOUT)
2074 #elif defined(USBI_TIMERFD_AVAILABLE)
2075 USBI_GET_CONTEXT(ctx);
2076 return usbi_using_timerfd(ctx);
2083 * Determine the next internal timeout that libusb needs to handle. You only
2084 * need to use this function if you are calling poll() or select() or similar
2085 * on libusb's file descriptors yourself - you do not need to use it if you
2086 * are calling libusb_handle_events() or a variant directly.
2088 * You should call this function in your main loop in order to determine how
2089 * long to wait for select() or poll() to return results. libusb needs to be
2090 * called into at this timeout, so you should use it as an upper bound on
2091 * your select() or poll() call.
2093 * When the timeout has expired, call into libusb_handle_events_timeout()
2094 * (perhaps in non-blocking mode) so that libusb can handle the timeout.
2096 * This function may return 1 (success) and an all-zero timeval. If this is
2097 * the case, it indicates that libusb has a timeout that has already expired
2098 * so you should call libusb_handle_events_timeout() or similar immediately.
2099 * A return code of 0 indicates that there are no pending timeouts.
2101 * On some platforms, this function will always returns 0 (no pending
2102 * timeouts). See \ref polltime.
2104 * \param ctx the context to operate on, or NULL for the default context
2105 * \param tv output location for a relative time against the current
2106 * clock in which libusb must be called into in order to process timeout events
2107 * \returns 0 if there are no pending timeouts, 1 if a timeout was returned,
2108 * or LIBUSB_ERROR_OTHER on failure
2110 API_EXPORTED int libusb_get_next_timeout(libusb_context *ctx,
2113 #ifndef USBI_OS_HANDLES_TIMEOUT
2114 struct usbi_transfer *transfer;
2115 struct timespec cur_ts;
2116 struct timeval cur_tv;
2117 struct timeval *next_timeout;
2121 USBI_GET_CONTEXT(ctx);
2122 if (usbi_using_timerfd(ctx))
2125 usbi_mutex_lock(&ctx->flying_transfers_lock);
2126 if (list_empty(&ctx->flying_transfers)) {
2127 usbi_mutex_unlock(&ctx->flying_transfers_lock);
2128 usbi_dbg("no URBs, no timeout!");
2132 /* find next transfer which hasn't already been processed as timed out */
2133 list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
2134 if (!(transfer->flags & USBI_TRANSFER_TIMED_OUT)) {
2139 usbi_mutex_unlock(&ctx->flying_transfers_lock);
2142 usbi_dbg("all URBs have already been processed for timeouts");
2146 next_timeout = &transfer->timeout;
2148 /* no timeout for next transfer */
2149 if (!timerisset(next_timeout)) {
2150 usbi_dbg("no URBs with timeouts, no timeout!");
2154 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, &cur_ts);
2156 usbi_err(ctx, "failed to read monotonic clock, errno=%d", errno);
2157 return LIBUSB_ERROR_OTHER;
2159 TIMESPEC_TO_TIMEVAL(&cur_tv, &cur_ts);
2161 if (!timercmp(&cur_tv, next_timeout, <)) {
2162 usbi_dbg("first timeout already expired");
2165 timersub(next_timeout, &cur_tv, tv);
2166 usbi_dbg("next timeout in %d.%06ds", tv->tv_sec, tv->tv_usec);
2176 * Register notification functions for file descriptor additions/removals.
2177 * These functions will be invoked for every new or removed file descriptor
2178 * that libusb uses as an event source.
2180 * To remove notifiers, pass NULL values for the function pointers.
2182 * Note that file descriptors may have been added even before you register
2183 * these notifiers (e.g. at libusb_init() time).
2185 * Additionally, note that the removal notifier may be called during
2186 * libusb_exit() (e.g. when it is closing file descriptors that were opened
2187 * and added to the poll set at libusb_init() time). If you don't want this,
2188 * remove the notifiers immediately before calling libusb_exit().
2190 * \param ctx the context to operate on, or NULL for the default context
2191 * \param added_cb pointer to function for addition notifications
2192 * \param removed_cb pointer to function for removal notifications
2193 * \param user_data User data to be passed back to callbacks (useful for
2194 * passing context information)
2196 API_EXPORTED void libusb_set_pollfd_notifiers(libusb_context *ctx,
2197 libusb_pollfd_added_cb added_cb, libusb_pollfd_removed_cb removed_cb,
2200 USBI_GET_CONTEXT(ctx);
2201 ctx->fd_added_cb = added_cb;
2202 ctx->fd_removed_cb = removed_cb;
2203 ctx->fd_cb_user_data = user_data;
2206 /* Add a file descriptor to the list of file descriptors to be monitored.
2207 * events should be specified as a bitmask of events passed to poll(), e.g.
2208 * POLLIN and/or POLLOUT. */
2209 int usbi_add_pollfd(struct libusb_context *ctx, int fd, short events)
2211 struct usbi_pollfd *ipollfd = malloc(sizeof(*ipollfd));
2213 return LIBUSB_ERROR_NO_MEM;
2215 usbi_dbg("add fd %d events %d", fd, events);
2216 ipollfd->pollfd.fd = fd;
2217 ipollfd->pollfd.events = events;
2218 usbi_mutex_lock(&ctx->pollfds_lock);
2219 list_add_tail(&ipollfd->list, &ctx->pollfds);
2220 usbi_mutex_unlock(&ctx->pollfds_lock);
2222 if (ctx->fd_added_cb)
2223 ctx->fd_added_cb(fd, events, ctx->fd_cb_user_data);
2227 /* Remove a file descriptor from the list of file descriptors to be polled. */
2228 void usbi_remove_pollfd(struct libusb_context *ctx, int fd)
2230 struct usbi_pollfd *ipollfd;
2233 usbi_dbg("remove fd %d", fd);
2234 usbi_mutex_lock(&ctx->pollfds_lock);
2235 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2236 if (ipollfd->pollfd.fd == fd) {
2242 usbi_dbg("couldn't find fd %d to remove", fd);
2243 usbi_mutex_unlock(&ctx->pollfds_lock);
2247 list_del(&ipollfd->list);
2248 usbi_mutex_unlock(&ctx->pollfds_lock);
2250 if (ctx->fd_removed_cb)
2251 ctx->fd_removed_cb(fd, ctx->fd_cb_user_data);
2255 * Retrieve a list of file descriptors that should be polled by your main loop
2256 * as libusb event sources.
2258 * The returned list is NULL-terminated and should be freed with free() when
2259 * done. The actual list contents must not be touched.
2261 * \param ctx the context to operate on, or NULL for the default context
2262 * \returns a NULL-terminated list of libusb_pollfd structures, or NULL on
2265 API_EXPORTED const struct libusb_pollfd **libusb_get_pollfds(
2266 libusb_context *ctx)
2268 struct libusb_pollfd **ret = NULL;
2269 struct usbi_pollfd *ipollfd;
2272 USBI_GET_CONTEXT(ctx);
2274 usbi_mutex_lock(&ctx->pollfds_lock);
2275 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2278 ret = calloc(cnt + 1, sizeof(struct libusb_pollfd *));
2282 list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2283 ret[i++] = (struct libusb_pollfd *) ipollfd;
2287 usbi_mutex_unlock(&ctx->pollfds_lock);
2288 return (const struct libusb_pollfd **) ret;
2291 /* Backends call this from handle_events to report disconnection of a device.
2292 * The transfers get cancelled appropriately.
2294 void usbi_handle_disconnect(struct libusb_device_handle *handle)
2296 struct usbi_transfer *cur;
2297 struct usbi_transfer *to_cancel;
2299 usbi_dbg("device %d.%d",
2300 handle->dev->bus_number, handle->dev->device_address);
2302 /* terminate all pending transfers with the LIBUSB_TRANSFER_NO_DEVICE
2305 * this is a bit tricky because:
2306 * 1. we can't do transfer completion while holding flying_transfers_lock
2307 * 2. the transfers list can change underneath us - if we were to build a
2308 * list of transfers to complete (while holding look), the situation
2309 * might be different by the time we come to free them
2311 * so we resort to a loop-based approach as below
2312 * FIXME: is this still potentially racy?
2316 usbi_mutex_lock(&HANDLE_CTX(handle)->flying_transfers_lock);
2318 list_for_each_entry(cur, &HANDLE_CTX(handle)->flying_transfers, list, struct usbi_transfer)
2319 if (__USBI_TRANSFER_TO_LIBUSB_TRANSFER(cur)->dev_handle == handle) {
2323 usbi_mutex_unlock(&HANDLE_CTX(handle)->flying_transfers_lock);
2328 usbi_backend->clear_transfer_priv(to_cancel);
2329 usbi_handle_transfer_completion(to_cancel, LIBUSB_TRANSFER_NO_DEVICE);