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
32 #ifdef USBI_TIMERFD_AVAILABLE
33 #include <sys/timerfd.h>
39 * \page io Synchronous and asynchronous device I/O
41 * \section intro Introduction
43 * If you're using libusb in your application, you're probably wanting to
44 * perform I/O with devices - you want to perform USB data transfers.
46 * libusb offers two separate interfaces for device I/O. This page aims to
47 * introduce the two in order to help you decide which one is more suitable
48 * for your application. You can also choose to use both interfaces in your
49 * application by considering each transfer on a case-by-case basis.
51 * Once you have read through the following discussion, you should consult the
52 * detailed API documentation pages for the details:
56 * \section theory Transfers at a logical level
58 * At a logical level, USB transfers typically happen in two parts. For
59 * example, when reading data from a endpoint:
60 * -# A request for data is sent to the device
61 * -# Some time later, the incoming data is received by the host
63 * or when writing data to an endpoint:
65 * -# The data is sent to the device
66 * -# Some time later, the host receives acknowledgement from the device that
67 * the data has been transferred.
69 * There may be an indefinite delay between the two steps. Consider a
70 * fictional USB input device with a button that the user can press. In order
71 * to determine when the button is pressed, you would likely submit a request
72 * to read data on a bulk or interrupt endpoint and wait for data to arrive.
73 * Data will arrive when the button is pressed by the user, which is
74 * potentially hours later.
76 * libusb offers both a synchronous and an asynchronous interface to performing
77 * USB transfers. The main difference is that the synchronous interface
78 * combines both steps indicated above into a single function call, whereas
79 * the asynchronous interface separates them.
81 * \section sync The synchronous interface
83 * The synchronous I/O interface allows you to perform a USB transfer with
84 * a single function call. When the function call returns, the transfer has
85 * completed and you can parse the results.
87 * If you have used the libusb-0.1 before, this I/O style will seem familar to
88 * you. libusb-0.1 only offered a synchronous interface.
90 * In our input device example, to read button presses you might write code
91 * in the following style:
93 unsigned char data[4];
95 int r = libusb_bulk_transfer(handle, EP_IN, data, sizeof(data), &actual_length, 0);
96 if (r == 0 && actual_length == sizeof(data)) {
97 // results of the transaction can now be found in the data buffer
98 // parse them here and report button press
104 * The main advantage of this model is simplicity: you did everything with
105 * a single simple function call.
107 * However, this interface has its limitations. Your application will sleep
108 * inside libusb_bulk_transfer() until the transaction has completed. If it
109 * takes the user 3 hours to press the button, your application will be
110 * sleeping for that long. Execution will be tied up inside the library -
111 * the entire thread will be useless for that duration.
113 * Another issue is that by tieing up the thread with that single transaction
114 * there is no possibility of performing I/O with multiple endpoints and/or
115 * multiple devices simultaneously, unless you resort to creating one thread
118 * Additionally, there is no opportunity to cancel the transfer after the
119 * request has been submitted.
121 * For details on how to use the synchronous API, see the
122 * \ref syncio "synchronous I/O API documentation" pages.
124 * \section async The asynchronous interface
126 * Asynchronous I/O is the most significant new feature in libusb-1.0.
127 * Although it is a more complex interface, it solves all the issues detailed
130 * Instead of providing which functions that block until the I/O has complete,
131 * libusb's asynchronous interface presents non-blocking functions which
132 * begin a transfer and then return immediately. Your application passes a
133 * callback function pointer to this non-blocking function, which libusb will
134 * call with the results of the transaction when it has completed.
136 * Transfers which have been submitted through the non-blocking functions
137 * can be cancelled with a separate function call.
139 * The non-blocking nature of this interface allows you to be simultaneously
140 * performing I/O to multiple endpoints on multiple devices, without having
143 * This added flexibility does come with some complications though:
144 * - In the interest of being a lightweight library, libusb does not create
145 * threads and can only operate when your application is calling into it. Your
146 * application must call into libusb from it's main loop when events are ready
147 * to be handled, or you must use some other scheme to allow libusb to
148 * undertake whatever work needs to be done.
149 * - libusb also needs to be called into at certain fixed points in time in
150 * order to accurately handle transfer timeouts.
151 * - Memory handling becomes more complex. You cannot use stack memory unless
152 * the function with that stack is guaranteed not to return until the transfer
153 * callback has finished executing.
154 * - You generally lose some linearity from your code flow because submitting
155 * the transfer request is done in a separate function from where the transfer
156 * results are handled. This becomes particularly obvious when you want to
157 * submit a second transfer based on the results of an earlier transfer.
159 * Internally, libusb's synchronous interface is expressed in terms of function
160 * calls to the asynchronous interface.
162 * For details on how to use the asynchronous API, see the
163 * \ref asyncio "asynchronous I/O API" documentation pages.
168 * \page packetoverflow Packets and overflows
170 * \section packets Packet abstraction
172 * The USB specifications describe how data is transmitted in packets, with
173 * constraints on packet size defined by endpoint descriptors. The host must
174 * not send data payloads larger than the endpoint's maximum packet size.
176 * libusb and the underlying OS abstract out the packet concept, allowing you
177 * to request transfers of any size. Internally, the request will be divided
178 * up into correctly-sized packets. You do not have to be concerned with
179 * packet sizes, but there is one exception when considering overflows.
181 * \section overflow Bulk/interrupt transfer overflows
183 * When requesting data on a bulk endpoint, libusb requires you to supply a
184 * buffer and the maximum number of bytes of data that libusb can put in that
185 * buffer. However, the size of the buffer is not communicated to the device -
186 * the device is just asked to send any amount of data.
188 * There is no problem if the device sends an amount of data that is less than
189 * or equal to the buffer size. libusb reports this condition to you through
190 * the \ref libusb_transfer::actual_length "libusb_transfer.actual_length"
193 * Problems may occur if the device attempts to send more data than can fit in
194 * the buffer. libusb reports LIBUSB_TRANSFER_OVERFLOW for this condition but
195 * other behaviour is largely undefined: actual_length may or may not be
196 * accurate, the chunk of data that can fit in the buffer (before overflow)
197 * may or may not have been transferred.
199 * Overflows are nasty, but can be avoided. Even though you were told to
200 * ignore packets above, think about the lower level details: each transfer is
201 * split into packets (typically small, with a maximum size of 512 bytes).
202 * Overflows can only happen if the final packet in an incoming data transfer
203 * is smaller than the actual packet that the device wants to transfer.
204 * Therefore, you will never see an overflow if your transfer buffer size is a
205 * multiple of the endpoint's packet size: the final packet will either
206 * fill up completely or will be only partially filled.
210 * @defgroup asyncio Asynchronous device I/O
212 * This page details libusb's asynchronous (non-blocking) API for USB device
213 * I/O. This interface is very powerful but is also quite complex - you will
214 * need to read this page carefully to understand the necessary considerations
215 * and issues surrounding use of this interface. Simplistic applications
216 * may wish to consider the \ref syncio "synchronous I/O API" instead.
218 * The asynchronous interface is built around the idea of separating transfer
219 * submission and handling of transfer completion (the synchronous model
220 * combines both of these into one). There may be a long delay between
221 * submission and completion, however the asynchronous submission function
222 * is non-blocking so will return control to your application during that
223 * potentially long delay.
225 * \section asyncabstraction Transfer abstraction
227 * For the asynchronous I/O, libusb implements the concept of a generic
228 * transfer entity for all types of I/O (control, bulk, interrupt,
229 * isochronous). The generic transfer object must be treated slightly
230 * differently depending on which type of I/O you are performing with it.
232 * This is represented by the public libusb_transfer structure type.
234 * \section asynctrf Asynchronous transfers
236 * We can view asynchronous I/O as a 5 step process:
237 * -# <b>Allocation</b>: allocate a libusb_transfer
238 * -# <b>Filling</b>: populate the libusb_transfer instance with information
239 * about the transfer you wish to perform
240 * -# <b>Submission</b>: ask libusb to submit the transfer
241 * -# <b>Completion handling</b>: examine transfer results in the
242 * libusb_transfer structure
243 * -# <b>Deallocation</b>: clean up resources
246 * \subsection asyncalloc Allocation
248 * This step involves allocating memory for a USB transfer. This is the
249 * generic transfer object mentioned above. At this stage, the transfer
250 * is "blank" with no details about what type of I/O it will be used for.
252 * Allocation is done with the libusb_alloc_transfer() function. You must use
253 * this function rather than allocating your own transfers.
255 * \subsection asyncfill Filling
257 * This step is where you take a previously allocated transfer and fill it
258 * with information to determine the message type and direction, data buffer,
259 * callback function, etc.
261 * You can either fill the required fields yourself or you can use the
262 * helper functions: libusb_fill_control_transfer(), libusb_fill_bulk_transfer()
263 * and libusb_fill_interrupt_transfer().
265 * \subsection asyncsubmit Submission
267 * When you have allocated a transfer and filled it, you can submit it using
268 * libusb_submit_transfer(). This function returns immediately but can be
269 * regarded as firing off the I/O request in the background.
271 * \subsection asynccomplete Completion handling
273 * After a transfer has been submitted, one of four things can happen to it:
275 * - The transfer completes (i.e. some data was transferred)
276 * - The transfer has a timeout and the timeout expires before all data is
278 * - The transfer fails due to an error
279 * - The transfer is cancelled
281 * Each of these will cause the user-specified transfer callback function to
282 * be invoked. It is up to the callback function to determine which of the
283 * above actually happened and to act accordingly.
285 * The user-specified callback is passed a pointer to the libusb_transfer
286 * structure which was used to setup and submit the transfer. At completion
287 * time, libusb has populated this structure with results of the transfer:
288 * success or failure reason, number of bytes of data transferred, etc. See
289 * the libusb_transfer structure documentation for more information.
291 * \subsection Deallocation
293 * When a transfer has completed (i.e. the callback function has been invoked),
294 * you are advised to free the transfer (unless you wish to resubmit it, see
295 * below). Transfers are deallocated with libusb_free_transfer().
297 * It is undefined behaviour to free a transfer which has not completed.
299 * \section asyncresubmit Resubmission
301 * You may be wondering why allocation, filling, and submission are all
302 * separated above where they could reasonably be combined into a single
305 * The reason for separation is to allow you to resubmit transfers without
306 * having to allocate new ones every time. This is especially useful for
307 * common situations dealing with interrupt endpoints - you allocate one
308 * transfer, fill and submit it, and when it returns with results you just
309 * resubmit it for the next interrupt.
311 * \section asynccancel Cancellation
313 * Another advantage of using the asynchronous interface is that you have
314 * the ability to cancel transfers which have not yet completed. This is
315 * done by calling the libusb_cancel_transfer() function.
317 * libusb_cancel_transfer() is asynchronous/non-blocking in itself. When the
318 * cancellation actually completes, the transfer's callback function will
319 * be invoked, and the callback function should check the transfer status to
320 * determine that it was cancelled.
322 * Freeing the transfer after it has been cancelled but before cancellation
323 * has completed will result in undefined behaviour.
325 * When a transfer is cancelled, some of the data may have been transferred.
326 * libusb will communicate this to you in the transfer callback. Do not assume
327 * that no data was transferred.
329 * \section bulk_overflows Overflows on device-to-host bulk/interrupt endpoints
331 * If your device does not have predictable transfer sizes (or it misbehaves),
332 * your application may submit a request for data on an IN endpoint which is
333 * smaller than the data that the device wishes to send. In some circumstances
334 * this will cause an overflow, which is a nasty condition to deal with. See
335 * the \ref packetoverflow page for discussion.
337 * \section asyncctrl Considerations for control transfers
339 * The <tt>libusb_transfer</tt> structure is generic and hence does not
340 * include specific fields for the control-specific setup packet structure.
342 * In order to perform a control transfer, you must place the 8-byte setup
343 * packet at the start of the data buffer. To simplify this, you could
344 * cast the buffer pointer to type struct libusb_control_setup, or you can
345 * use the helper function libusb_fill_control_setup().
347 * The wLength field placed in the setup packet must be the length you would
348 * expect to be sent in the setup packet: the length of the payload that
349 * follows (or the expected maximum number of bytes to receive). However,
350 * the length field of the libusb_transfer object must be the length of
351 * the data buffer - i.e. it should be wLength <em>plus</em> the size of
352 * the setup packet (LIBUSB_CONTROL_SETUP_SIZE).
354 * If you use the helper functions, this is simplified for you:
355 * -# Allocate a buffer of size LIBUSB_CONTROL_SETUP_SIZE plus the size of the
356 * data you are sending/requesting.
357 * -# Call libusb_fill_control_setup() on the data buffer, using the transfer
358 * request size as the wLength value (i.e. do not include the extra space you
359 * allocated for the control setup).
360 * -# If this is a host-to-device transfer, place the data to be transferred
361 * in the data buffer, starting at offset LIBUSB_CONTROL_SETUP_SIZE.
362 * -# Call libusb_fill_control_transfer() to associate the data buffer with
363 * the transfer (and to set the remaining details such as callback and timeout).
364 * - Note that there is no parameter to set the length field of the transfer.
365 * The length is automatically inferred from the wLength field of the setup
367 * -# Submit the transfer.
369 * The multi-byte control setup fields (wValue, wIndex and wLength) must
370 * be given in little-endian byte order (the endianness of the USB bus).
371 * Endianness conversion is transparently handled by
372 * libusb_fill_control_setup() which is documented to accept host-endian
375 * Further considerations are needed when handling transfer completion in
376 * your callback function:
377 * - As you might expect, the setup packet will still be sitting at the start
378 * of the data buffer.
379 * - If this was a device-to-host transfer, the received data will be sitting
380 * at offset LIBUSB_CONTROL_SETUP_SIZE into the buffer.
381 * - The actual_length field of the transfer structure is relative to the
382 * wLength of the setup packet, rather than the size of the data buffer. So,
383 * if your wLength was 4, your transfer's <tt>length</tt> was 12, then you
384 * should expect an <tt>actual_length</tt> of 4 to indicate that the data was
385 * transferred in entirity.
387 * To simplify parsing of setup packets and obtaining the data from the
388 * correct offset, you may wish to use the libusb_control_transfer_get_data()
389 * and libusb_control_transfer_get_setup() functions within your transfer
392 * Even though control endpoints do not halt, a completed control transfer
393 * may have a LIBUSB_TRANSFER_STALL status code. This indicates the control
394 * request was not supported.
396 * \section asyncintr Considerations for interrupt transfers
398 * All interrupt transfers are performed using the polling interval presented
399 * by the bInterval value of the endpoint descriptor.
401 * \section asynciso Considerations for isochronous transfers
403 * Isochronous transfers are more complicated than transfers to
404 * non-isochronous endpoints.
406 * To perform I/O to an isochronous endpoint, allocate the transfer by calling
407 * libusb_alloc_transfer() with an appropriate number of isochronous packets.
409 * During filling, set \ref libusb_transfer::type "type" to
410 * \ref libusb_transfer_type::LIBUSB_TRANSFER_TYPE_ISOCHRONOUS
411 * "LIBUSB_TRANSFER_TYPE_ISOCHRONOUS", and set
412 * \ref libusb_transfer::num_iso_packets "num_iso_packets" to a value less than
413 * or equal to the number of packets you requested during allocation.
414 * libusb_alloc_transfer() does not set either of these fields for you, given
415 * that you might not even use the transfer on an isochronous endpoint.
417 * Next, populate the length field for the first num_iso_packets entries in
418 * the \ref libusb_transfer::iso_packet_desc "iso_packet_desc" array. Section
419 * 5.6.3 of the USB2 specifications describe how the maximum isochronous
420 * packet length is determined by the wMaxPacketSize field in the endpoint
422 * Two functions can help you here:
424 * - libusb_get_max_iso_packet_size() is an easy way to determine the max
425 * packet size for an isochronous endpoint. Note that the maximum packet
426 * size is actually the maximum number of bytes that can be transmitted in
427 * a single microframe, therefore this function multiplies the maximum number
428 * of bytes per transaction by the number of transaction opportunities per
430 * - libusb_set_iso_packet_lengths() assigns the same length to all packets
431 * within a transfer, which is usually what you want.
433 * For outgoing transfers, you'll obviously fill the buffer and populate the
434 * packet descriptors in hope that all the data gets transferred. For incoming
435 * transfers, you must ensure the buffer has sufficient capacity for
436 * the situation where all packets transfer the full amount of requested data.
438 * Completion handling requires some extra consideration. The
439 * \ref libusb_transfer::actual_length "actual_length" field of the transfer
440 * is meaningless and should not be examined; instead you must refer to the
441 * \ref libusb_iso_packet_descriptor::actual_length "actual_length" field of
442 * each individual packet.
444 * The \ref libusb_transfer::status "status" field of the transfer is also a
446 * - If the packets were submitted and the isochronous data microframes
447 * completed normally, status will have value
448 * \ref libusb_transfer_status::LIBUSB_TRANSFER_COMPLETED
449 * "LIBUSB_TRANSFER_COMPLETED". Note that bus errors and software-incurred
450 * delays are not counted as transfer errors; the transfer.status field may
451 * indicate COMPLETED even if some or all of the packets failed. Refer to
452 * the \ref libusb_iso_packet_descriptor::status "status" field of each
453 * individual packet to determine packet failures.
454 * - The status field will have value
455 * \ref libusb_transfer_status::LIBUSB_TRANSFER_ERROR
456 * "LIBUSB_TRANSFER_ERROR" only when serious errors were encountered.
457 * - Other transfer status codes occur with normal behaviour.
459 * The data for each packet will be found at an offset into the buffer that
460 * can be calculated as if each prior packet completed in full. The
461 * libusb_get_iso_packet_buffer() and libusb_get_iso_packet_buffer_simple()
462 * functions may help you here.
464 * \section asyncmem Memory caveats
466 * In most circumstances, it is not safe to use stack memory for transfer
467 * buffers. This is because the function that fired off the asynchronous
468 * transfer may return before libusb has finished using the buffer, and when
469 * the function returns it's stack gets destroyed. This is true for both
470 * host-to-device and device-to-host transfers.
472 * The only case in which it is safe to use stack memory is where you can
473 * guarantee that the function owning the stack space for the buffer does not
474 * return until after the transfer's callback function has completed. In every
475 * other case, you need to use heap memory instead.
477 * \section asyncflags Fine control
479 * Through using this asynchronous interface, you may find yourself repeating
480 * a few simple operations many times. You can apply a bitwise OR of certain
481 * flags to a transfer to simplify certain things:
482 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_SHORT_NOT_OK
483 * "LIBUSB_TRANSFER_SHORT_NOT_OK" results in transfers which transferred
484 * less than the requested amount of data being marked with status
485 * \ref libusb_transfer_status::LIBUSB_TRANSFER_ERROR "LIBUSB_TRANSFER_ERROR"
486 * (they would normally be regarded as COMPLETED)
487 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_BUFFER
488 * "LIBUSB_TRANSFER_FREE_BUFFER" allows you to ask libusb to free the transfer
489 * buffer when freeing the transfer.
490 * - \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_TRANSFER
491 * "LIBUSB_TRANSFER_FREE_TRANSFER" causes libusb to automatically free the
492 * transfer after the transfer callback returns.
494 * \section asyncevent Event handling
496 * In accordance of the aim of being a lightweight library, libusb does not
497 * create threads internally. This means that libusb code does not execute
498 * at any time other than when your application is calling a libusb function.
499 * However, an asynchronous model requires that libusb perform work at various
500 * points in time - namely processing the results of previously-submitted
501 * transfers and invoking the user-supplied callback function.
503 * This gives rise to the libusb_handle_events() function which your
504 * application must call into when libusb has work do to. This gives libusb
505 * the opportunity to reap pending transfers, invoke callbacks, etc.
507 * The first issue to discuss here is how your application can figure out
508 * when libusb has work to do. In fact, there are two naive options which
509 * do not actually require your application to know this:
510 * -# Periodically call libusb_handle_events() in non-blocking mode at fixed
511 * short intervals from your main loop
512 * -# Repeatedly call libusb_handle_events() in blocking mode from a dedicated
515 * The first option is plainly not very nice, and will cause unnecessary
516 * CPU wakeups leading to increased power usage and decreased battery life.
517 * The second option is not very nice either, but may be the nicest option
518 * available to you if the "proper" approach can not be applied to your
519 * application (read on...).
521 * The recommended option is to integrate libusb with your application main
522 * event loop. libusb exposes a set of file descriptors which allow you to do
523 * this. Your main loop is probably already calling poll() or select() or a
524 * variant on a set of file descriptors for other event sources (e.g. keyboard
525 * button presses, mouse movements, network sockets, etc). You then add
526 * libusb's file descriptors to your poll()/select() calls, and when activity
527 * is detected on such descriptors you know it is time to call
528 * libusb_handle_events().
530 * There is one final event handling complication. libusb supports
531 * asynchronous transfers which time out after a specified time period, and
532 * this requires that libusb is called into at or after the timeout so that
533 * the timeout can be handled. So, in addition to considering libusb's file
534 * descriptors in your main event loop, you must also consider that libusb
535 * sometimes needs to be called into at fixed points in time even when there
536 * is no file descriptor activity.
538 * For the details on retrieving the set of file descriptors and determining
539 * the next timeout, see the \ref poll "polling and timing" API documentation.
543 * @defgroup poll Polling and timing
545 * This page documents libusb's functions for polling events and timing.
546 * These functions are only necessary for users of the
547 * \ref asyncio "asynchronous API". If you are only using the simpler
548 * \ref syncio "synchronous API" then you do not need to ever call these
551 * The justification for the functionality described here has already been
552 * discussed in the \ref asyncevent "event handling" section of the
553 * asynchronous API documentation. In summary, libusb does not create internal
554 * threads for event processing and hence relies on your application calling
555 * into libusb at certain points in time so that pending events can be handled.
556 * In order to know precisely when libusb needs to be called into, libusb
557 * offers you a set of pollable file descriptors and information about when
558 * the next timeout expires.
560 * If you are using the asynchronous I/O API, you must take one of the two
561 * following options, otherwise your I/O will not complete.
563 * \section pollsimple The simple option
565 * If your application revolves solely around libusb and does not need to
566 * handle other event sources, you can have a program structure as follows:
569 // find and open device
570 // maybe fire off some initial async I/O
572 while (user_has_not_requested_exit)
573 libusb_handle_events(ctx);
578 * With such a simple main loop, you do not have to worry about managing
579 * sets of file descriptors or handling timeouts. libusb_handle_events() will
580 * handle those details internally.
582 * \section pollmain The more advanced option
584 * In more advanced applications, you will already have a main loop which
585 * is monitoring other event sources: network sockets, X11 events, mouse
586 * movements, etc. Through exposing a set of file descriptors, libusb is
587 * designed to cleanly integrate into such main loops.
589 * In addition to polling file descriptors for the other event sources, you
590 * take a set of file descriptors from libusb and monitor those too. When you
591 * detect activity on libusb's file descriptors, you call
592 * libusb_handle_events_timeout() in non-blocking mode.
594 * What's more, libusb may also need to handle events at specific moments in
595 * time. No file descriptor activity is generated at these times, so your
596 * own application needs to be continually aware of when the next one of these
597 * moments occurs (through calling libusb_get_next_timeout()), and then it
598 * needs to call libusb_handle_events_timeout() in non-blocking mode when
599 * these moments occur. This means that you need to adjust your
600 * poll()/select() timeout accordingly.
602 * libusb provides you with a set of file descriptors to poll and expects you
603 * to poll all of them, treating them as a single entity. The meaning of each
604 * file descriptor in the set is an internal implementation detail,
605 * platform-dependent and may vary from release to release. Don't try and
606 * interpret the meaning of the file descriptors, just do as libusb indicates,
607 * polling all of them at once.
609 * In pseudo-code, you want something that looks like:
613 libusb_get_pollfds(ctx)
614 while (user has not requested application exit) {
615 libusb_get_next_timeout(ctx);
616 poll(on libusb file descriptors plus any other event sources of interest,
617 using a timeout no larger than the value libusb just suggested)
618 if (poll() indicated activity on libusb file descriptors)
619 libusb_handle_events_timeout(ctx, 0);
620 if (time has elapsed to or beyond the libusb timeout)
621 libusb_handle_events_timeout(ctx, 0);
622 // handle events from other sources here
628 * \subsection polltime Notes on time-based events
630 * The above complication with having to track time and call into libusb at
631 * specific moments is a bit of a headache. For maximum compatibility, you do
632 * need to write your main loop as above, but you may decide that you can
633 * restrict the supported platforms of your application and get away with
634 * a more simplistic scheme.
636 * These time-based event complications are \b not required on the following
639 * - Linux, provided that the following version requirements are satisfied:
640 * - Linux v2.6.27 or newer, compiled with timerfd support
641 * - glibc v2.9 or newer
642 * - libusb v1.0.5 or newer
644 * Under these configurations, libusb_get_next_timeout() will \em always return
645 * 0, so your main loop can be simplified to:
649 libusb_get_pollfds(ctx)
650 while (user has not requested application exit) {
651 poll(on libusb file descriptors plus any other event sources of interest,
652 using any timeout that you like)
653 if (poll() indicated activity on libusb file descriptors)
654 libusb_handle_events_timeout(ctx, 0);
655 // handle events from other sources here
661 * Do remember that if you simplify your main loop to the above, you will
662 * lose compatibility with some platforms (including legacy Linux platforms,
663 * and <em>any future platforms supported by libusb which may have time-based
664 * event requirements</em>). The resultant problems will likely appear as
665 * strange bugs in your application.
667 * You can use the libusb_pollfds_handle_timeouts() function to do a runtime
668 * check to see if it is safe to ignore the time-based event complications.
669 * If your application has taken the shortcut of ignoring libusb's next timeout
670 * in your main loop, then you are advised to check the return value of
671 * libusb_pollfds_handle_timeouts() during application startup, and to abort
672 * if the platform does suffer from these timing complications.
674 * \subsection fdsetchange Changes in the file descriptor set
676 * The set of file descriptors that libusb uses as event sources may change
677 * during the life of your application. Rather than having to repeatedly
678 * call libusb_get_pollfds(), you can set up notification functions for when
679 * the file descriptor set changes using libusb_set_pollfd_notifiers().
681 * \subsection mtissues Multi-threaded considerations
683 * Unfortunately, the situation is complicated further when multiple threads
684 * come into play. If two threads are monitoring the same file descriptors,
685 * the fact that only one thread will be woken up when an event occurs causes
688 * The events lock, event waiters lock, and libusb_handle_events_locked()
689 * entities are added to solve these problems. You do not need to be concerned
690 * with these entities otherwise.
692 * See the extra documentation: \ref mtasync
695 /** \page mtasync Multi-threaded applications and asynchronous I/O
697 * libusb is a thread-safe library, but extra considerations must be applied
698 * to applications which interact with libusb from multiple threads.
700 * The underlying issue that must be addressed is that all libusb I/O
701 * revolves around monitoring file descriptors through the poll()/select()
702 * system calls. This is directly exposed at the
703 * \ref asyncio "asynchronous interface" but it is important to note that the
704 * \ref syncio "synchronous interface" is implemented on top of the
705 * asynchonrous interface, therefore the same considerations apply.
707 * The issue is that if two or more threads are concurrently calling poll()
708 * or select() on libusb's file descriptors then only one of those threads
709 * will be woken up when an event arrives. The others will be completely
710 * oblivious that anything has happened.
712 * Consider the following pseudo-code, which submits an asynchronous transfer
713 * then waits for its completion. This style is one way you could implement a
714 * synchronous interface on top of the asynchronous interface (and libusb
715 * does something similar, albeit more advanced due to the complications
716 * explained on this page).
719 void cb(struct libusb_transfer *transfer)
721 int *completed = transfer->user_data;
726 struct libusb_transfer *transfer;
727 unsigned char buffer[LIBUSB_CONTROL_SETUP_SIZE];
730 transfer = libusb_alloc_transfer(0);
731 libusb_fill_control_setup(buffer,
732 LIBUSB_REQUEST_TYPE_VENDOR | LIBUSB_ENDPOINT_OUT, 0x04, 0x01, 0, 0);
733 libusb_fill_control_transfer(transfer, dev, buffer, cb, &completed, 1000);
734 libusb_submit_transfer(transfer);
737 poll(libusb file descriptors, 120*1000);
738 if (poll indicates activity)
739 libusb_handle_events_timeout(ctx, 0);
741 printf("completed!");
746 * Here we are <em>serializing</em> completion of an asynchronous event
747 * against a condition - the condition being completion of a specific transfer.
748 * The poll() loop has a long timeout to minimize CPU usage during situations
749 * when nothing is happening (it could reasonably be unlimited).
751 * If this is the only thread that is polling libusb's file descriptors, there
752 * is no problem: there is no danger that another thread will swallow up the
753 * event that we are interested in. On the other hand, if there is another
754 * thread polling the same descriptors, there is a chance that it will receive
755 * the event that we were interested in. In this situation, <tt>myfunc()</tt>
756 * will only realise that the transfer has completed on the next iteration of
757 * the loop, <em>up to 120 seconds later.</em> Clearly a two-minute delay is
758 * undesirable, and don't even think about using short timeouts to circumvent
761 * The solution here is to ensure that no two threads are ever polling the
762 * file descriptors at the same time. A naive implementation of this would
763 * impact the capabilities of the library, so libusb offers the scheme
764 * documented below to ensure no loss of functionality.
766 * Before we go any further, it is worth mentioning that all libusb-wrapped
767 * event handling procedures fully adhere to the scheme documented below.
768 * This includes libusb_handle_events() and all the synchronous I/O functions -
769 * libusb hides this headache from you. You do not need to worry about any
770 * of these issues if you stick to that level.
772 * The problem is when we consider the fact that libusb exposes file
773 * descriptors to allow for you to integrate asynchronous USB I/O into
774 * existing main loops, effectively allowing you to do some work behind
775 * libusb's back. If you do take libusb's file descriptors and pass them to
776 * poll()/select() yourself, you need to be aware of the associated issues.
778 * \section eventlock The events lock
780 * The first concept to be introduced is the events lock. The events lock
781 * is used to serialize threads that want to handle events, such that only
782 * one thread is handling events at any one time.
784 * You must take the events lock before polling libusb file descriptors,
785 * using libusb_lock_events(). You must release the lock as soon as you have
786 * aborted your poll()/select() loop, using libusb_unlock_events().
788 * \section threadwait Letting other threads do the work for you
790 * Although the events lock is a critical part of the solution, it is not
791 * enough on it's own. You might wonder if the following is sufficient...
793 libusb_lock_events(ctx);
795 poll(libusb file descriptors, 120*1000);
796 if (poll indicates activity)
797 libusb_handle_events_timeout(ctx, 0);
799 libusb_unlock_events(ctx);
801 * ...and the answer is that it is not. This is because the transfer in the
802 * code shown above may take a long time (say 30 seconds) to complete, and
803 * the lock is not released until the transfer is completed.
805 * Another thread with similar code that wants to do event handling may be
806 * working with a transfer that completes after a few milliseconds. Despite
807 * having such a quick completion time, the other thread cannot check that
808 * status of its transfer until the code above has finished (30 seconds later)
809 * due to contention on the lock.
811 * To solve this, libusb offers you a mechanism to determine when another
812 * thread is handling events. It also offers a mechanism to block your thread
813 * until the event handling thread has completed an event (and this mechanism
814 * does not involve polling of file descriptors).
816 * After determining that another thread is currently handling events, you
817 * obtain the <em>event waiters</em> lock using libusb_lock_event_waiters().
818 * You then re-check that some other thread is still handling events, and if
819 * so, you call libusb_wait_for_event().
821 * libusb_wait_for_event() puts your application to sleep until an event
822 * occurs, or until a thread releases the events lock. When either of these
823 * things happen, your thread is woken up, and should re-check the condition
824 * it was waiting on. It should also re-check that another thread is handling
825 * events, and if not, it should start handling events itself.
827 * This looks like the following, as pseudo-code:
830 if (libusb_try_lock_events(ctx) == 0) {
831 // we obtained the event lock: do our own event handling
833 if (!libusb_event_handling_ok(ctx)) {
834 libusb_unlock_events(ctx);
837 poll(libusb file descriptors, 120*1000);
838 if (poll indicates activity)
839 libusb_handle_events_locked(ctx, 0);
841 libusb_unlock_events(ctx);
843 // another thread is doing event handling. wait for it to signal us that
844 // an event has completed
845 libusb_lock_event_waiters(ctx);
848 // now that we have the event waiters lock, double check that another
849 // thread is still handling events for us. (it may have ceased handling
850 // events in the time it took us to reach this point)
851 if (!libusb_event_handler_active(ctx)) {
852 // whoever was handling events is no longer doing so, try again
853 libusb_unlock_event_waiters(ctx);
857 libusb_wait_for_event(ctx);
859 libusb_unlock_event_waiters(ctx);
861 printf("completed!\n");
864 * A naive look at the above code may suggest that this can only support
865 * one event waiter (hence a total of 2 competing threads, the other doing
866 * event handling), because the event waiter seems to have taken the event
867 * waiters lock while waiting for an event. However, the system does support
868 * multiple event waiters, because libusb_wait_for_event() actually drops
869 * the lock while waiting, and reaquires it before continuing.
871 * We have now implemented code which can dynamically handle situations where
872 * nobody is handling events (so we should do it ourselves), and it can also
873 * handle situations where another thread is doing event handling (so we can
874 * piggyback onto them). It is also equipped to handle a combination of
875 * the two, for example, another thread is doing event handling, but for
876 * whatever reason it stops doing so before our condition is met, so we take
877 * over the event handling.
879 * Four functions were introduced in the above pseudo-code. Their importance
880 * should be apparent from the code shown above.
881 * -# libusb_try_lock_events() is a non-blocking function which attempts
882 * to acquire the events lock but returns a failure code if it is contended.
883 * -# libusb_event_handling_ok() checks that libusb is still happy for your
884 * thread to be performing event handling. Sometimes, libusb needs to
885 * interrupt the event handler, and this is how you can check if you have
886 * been interrupted. If this function returns 0, the correct behaviour is
887 * for you to give up the event handling lock, and then to repeat the cycle.
888 * The following libusb_try_lock_events() will fail, so you will become an
889 * events waiter. For more information on this, read \ref fullstory below.
890 * -# libusb_handle_events_locked() is a variant of
891 * libusb_handle_events_timeout() that you can call while holding the
892 * events lock. libusb_handle_events_timeout() itself implements similar
893 * logic to the above, so be sure not to call it when you are
894 * "working behind libusb's back", as is the case here.
895 * -# libusb_event_handler_active() determines if someone is currently
896 * holding the events lock
898 * You might be wondering why there is no function to wake up all threads
899 * blocked on libusb_wait_for_event(). This is because libusb can do this
900 * internally: it will wake up all such threads when someone calls
901 * libusb_unlock_events() or when a transfer completes (at the point after its
902 * callback has returned).
904 * \subsection fullstory The full story
906 * The above explanation should be enough to get you going, but if you're
907 * really thinking through the issues then you may be left with some more
908 * questions regarding libusb's internals. If you're curious, read on, and if
909 * not, skip to the next section to avoid confusing yourself!
911 * The immediate question that may spring to mind is: what if one thread
912 * modifies the set of file descriptors that need to be polled while another
913 * thread is doing event handling?
915 * There are 2 situations in which this may happen.
916 * -# libusb_open() will add another file descriptor to the poll set,
917 * therefore it is desirable to interrupt the event handler so that it
918 * restarts, picking up the new descriptor.
919 * -# libusb_close() will remove a file descriptor from the poll set. There
920 * are all kinds of race conditions that could arise here, so it is
921 * important that nobody is doing event handling at this time.
923 * libusb handles these issues internally, so application developers do not
924 * have to stop their event handlers while opening/closing devices. Here's how
925 * it works, focusing on the libusb_close() situation first:
927 * -# During initialization, libusb opens an internal pipe, and it adds the read
928 * end of this pipe to the set of file descriptors to be polled.
929 * -# During libusb_close(), libusb writes some dummy data on this control pipe.
930 * This immediately interrupts the event handler. libusb also records
931 * internally that it is trying to interrupt event handlers for this
932 * high-priority event.
933 * -# At this point, some of the functions described above start behaving
935 * - libusb_event_handling_ok() starts returning 1, indicating that it is NOT
936 * OK for event handling to continue.
937 * - libusb_try_lock_events() starts returning 1, indicating that another
938 * thread holds the event handling lock, even if the lock is uncontended.
939 * - libusb_event_handler_active() starts returning 1, indicating that
940 * another thread is doing event handling, even if that is not true.
941 * -# The above changes in behaviour result in the event handler stopping and
942 * giving up the events lock very quickly, giving the high-priority
943 * libusb_close() operation a "free ride" to acquire the events lock. All
944 * threads that are competing to do event handling become event waiters.
945 * -# With the events lock held inside libusb_close(), libusb can safely remove
946 * a file descriptor from the poll set, in the safety of knowledge that
947 * nobody is polling those descriptors or trying to access the poll set.
948 * -# After obtaining the events lock, the close operation completes very
949 * quickly (usually a matter of milliseconds) and then immediately releases
951 * -# At the same time, the behaviour of libusb_event_handling_ok() and friends
952 * reverts to the original, documented behaviour.
953 * -# The release of the events lock causes the threads that are waiting for
954 * events to be woken up and to start competing to become event handlers
955 * again. One of them will succeed; it will then re-obtain the list of poll
956 * descriptors, and USB I/O will then continue as normal.
958 * libusb_open() is similar, and is actually a more simplistic case. Upon a
959 * call to libusb_open():
961 * -# The device is opened and a file descriptor is added to the poll set.
962 * -# libusb sends some dummy data on the control pipe, and records that it
963 * is trying to modify the poll descriptor set.
964 * -# The event handler is interrupted, and the same behaviour change as for
965 * libusb_close() takes effect, causing all event handling threads to become
967 * -# The libusb_open() implementation takes its free ride to the events lock.
968 * -# Happy that it has successfully paused the events handler, libusb_open()
969 * releases the events lock.
970 * -# The event waiter threads are all woken up and compete to become event
971 * handlers again. The one that succeeds will obtain the list of poll
972 * descriptors again, which will include the addition of the new device.
974 * \subsection concl Closing remarks
976 * The above may seem a little complicated, but hopefully I have made it clear
977 * why such complications are necessary. Also, do not forget that this only
978 * applies to applications that take libusb's file descriptors and integrate
979 * them into their own polling loops.
981 * You may decide that it is OK for your multi-threaded application to ignore
982 * some of the rules and locks detailed above, because you don't think that
983 * two threads can ever be polling the descriptors at the same time. If that
984 * is the case, then that's good news for you because you don't have to worry.
985 * But be careful here; remember that the synchronous I/O functions do event
986 * handling internally. If you have one thread doing event handling in a loop
987 * (without implementing the rules and locking semantics documented above)
988 * and another trying to send a synchronous USB transfer, you will end up with
989 * two threads monitoring the same descriptors, and the above-described
990 * undesirable behaviour occuring. The solution is for your polling thread to
991 * play by the rules; the synchronous I/O functions do so, and this will result
992 * in them getting along in perfect harmony.
994 * If you do have a dedicated thread doing event handling, it is perfectly
995 * legal for it to take the event handling lock for long periods of time. Any
996 * synchronous I/O functions you call from other threads will transparently
997 * fall back to the "event waiters" mechanism detailed above. The only
998 * consideration that your event handling thread must apply is the one related
999 * to libusb_event_handling_ok(): you must call this before every poll(), and
1000 * give up the events lock if instructed.
1003 int usbi_io_init(struct libusb_context *ctx)
1007 usbi_mutex_init(&ctx->flying_transfers_lock, NULL);
1008 usbi_mutex_init(&ctx->pollfds_lock, NULL);
1009 usbi_mutex_init(&ctx->pollfd_modify_lock, NULL);
1010 usbi_mutex_init(&ctx->events_lock, NULL);
1011 usbi_mutex_init(&ctx->event_waiters_lock, NULL);
1012 usbi_cond_init(&ctx->event_waiters_cond, NULL);
1013 list_init(&ctx->flying_transfers);
1014 list_init(&ctx->pollfds);
1016 /* FIXME should use an eventfd on kernels that support it */
1017 r = pipe(ctx->ctrl_pipe);
1019 r = LIBUSB_ERROR_OTHER;
1023 r = usbi_add_pollfd(ctx, ctx->ctrl_pipe[0], POLLIN);
1025 goto err_close_pipe;
1027 #ifdef USBI_TIMERFD_AVAILABLE
1028 ctx->timerfd = timerfd_create(usbi_backend->get_timerfd_clockid(),
1030 if (ctx->timerfd >= 0) {
1031 usbi_dbg("using timerfd for timeouts");
1032 r = usbi_add_pollfd(ctx, ctx->timerfd, POLLIN);
1034 close(ctx->timerfd);
1035 goto err_close_pipe;
1038 usbi_dbg("timerfd not available (code %d error %d)", ctx->timerfd, errno);
1046 close(ctx->ctrl_pipe[0]);
1047 close(ctx->ctrl_pipe[1]);
1049 usbi_mutex_destroy(&ctx->flying_transfers_lock);
1050 usbi_mutex_destroy(&ctx->pollfds_lock);
1051 usbi_mutex_destroy(&ctx->pollfd_modify_lock);
1052 usbi_mutex_destroy(&ctx->events_lock);
1053 usbi_mutex_destroy(&ctx->event_waiters_lock);
1054 usbi_cond_destroy(&ctx->event_waiters_cond);
1058 void usbi_io_exit(struct libusb_context *ctx)
1060 usbi_remove_pollfd(ctx, ctx->ctrl_pipe[0]);
1061 close(ctx->ctrl_pipe[0]);
1062 close(ctx->ctrl_pipe[1]);
1063 #ifdef USBI_TIMERFD_AVAILABLE
1064 if (usbi_using_timerfd(ctx)) {
1065 usbi_remove_pollfd(ctx, ctx->timerfd);
1066 close(ctx->timerfd);
1069 usbi_mutex_destroy(&ctx->flying_transfers_lock);
1070 usbi_mutex_destroy(&ctx->pollfds_lock);
1071 usbi_mutex_destroy(&ctx->pollfd_modify_lock);
1072 usbi_mutex_destroy(&ctx->events_lock);
1073 usbi_mutex_destroy(&ctx->event_waiters_lock);
1074 usbi_cond_destroy(&ctx->event_waiters_cond);
1077 static int calculate_timeout(struct usbi_transfer *transfer)
1080 struct timespec current_time;
1081 unsigned int timeout =
1082 __USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout;
1087 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, ¤t_time);
1089 usbi_err(ITRANSFER_CTX(transfer),
1090 "failed to read monotonic clock, errno=%d", errno);
1094 current_time.tv_sec += timeout / 1000;
1095 current_time.tv_nsec += (timeout % 1000) * 1000000;
1097 if (current_time.tv_nsec > 1000000000) {
1098 current_time.tv_nsec -= 1000000000;
1099 current_time.tv_sec++;
1102 TIMESPEC_TO_TIMEVAL(&transfer->timeout, ¤t_time);
1106 /* add a transfer to the (timeout-sorted) active transfers list.
1107 * returns 1 if the transfer has a timeout and it is the timeout next to
1109 static int add_to_flying_list(struct usbi_transfer *transfer)
1111 struct usbi_transfer *cur;
1112 struct timeval *timeout = &transfer->timeout;
1113 struct libusb_context *ctx = ITRANSFER_CTX(transfer);
1117 usbi_mutex_lock(&ctx->flying_transfers_lock);
1119 /* if we have no other flying transfers, start the list with this one */
1120 if (list_empty(&ctx->flying_transfers)) {
1121 list_add(&transfer->list, &ctx->flying_transfers);
1122 if (timerisset(timeout))
1127 /* if we have infinite timeout, append to end of list */
1128 if (!timerisset(timeout)) {
1129 list_add_tail(&transfer->list, &ctx->flying_transfers);
1133 /* otherwise, find appropriate place in list */
1134 list_for_each_entry(cur, &ctx->flying_transfers, list) {
1135 /* find first timeout that occurs after the transfer in question */
1136 struct timeval *cur_tv = &cur->timeout;
1138 if (!timerisset(cur_tv) || (cur_tv->tv_sec > timeout->tv_sec) ||
1139 (cur_tv->tv_sec == timeout->tv_sec &&
1140 cur_tv->tv_usec > timeout->tv_usec)) {
1141 list_add_tail(&transfer->list, &cur->list);
1148 /* otherwise we need to be inserted at the end */
1149 list_add_tail(&transfer->list, &ctx->flying_transfers);
1151 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1155 /** \ingroup asyncio
1156 * Allocate a libusb transfer with a specified number of isochronous packet
1157 * descriptors. The returned transfer is pre-initialized for you. When the new
1158 * transfer is no longer needed, it should be freed with
1159 * libusb_free_transfer().
1161 * Transfers intended for non-isochronous endpoints (e.g. control, bulk,
1162 * interrupt) should specify an iso_packets count of zero.
1164 * For transfers intended for isochronous endpoints, specify an appropriate
1165 * number of packet descriptors to be allocated as part of the transfer.
1166 * The returned transfer is not specially initialized for isochronous I/O;
1167 * you are still required to set the
1168 * \ref libusb_transfer::num_iso_packets "num_iso_packets" and
1169 * \ref libusb_transfer::type "type" fields accordingly.
1171 * It is safe to allocate a transfer with some isochronous packets and then
1172 * use it on a non-isochronous endpoint. If you do this, ensure that at time
1173 * of submission, num_iso_packets is 0 and that type is set appropriately.
1175 * \param iso_packets number of isochronous packet descriptors to allocate
1176 * \returns a newly allocated transfer, or NULL on error
1178 API_EXPORTED struct libusb_transfer *libusb_alloc_transfer(int iso_packets)
1180 size_t os_alloc_size = usbi_backend->transfer_priv_size
1181 + (usbi_backend->add_iso_packet_size * iso_packets);
1182 size_t alloc_size = sizeof(struct usbi_transfer)
1183 + sizeof(struct libusb_transfer)
1184 + (sizeof(struct libusb_iso_packet_descriptor) * iso_packets)
1186 struct usbi_transfer *itransfer = malloc(alloc_size);
1190 memset(itransfer, 0, alloc_size);
1191 itransfer->num_iso_packets = iso_packets;
1192 usbi_mutex_init(&itransfer->lock, NULL);
1193 return __USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1196 /** \ingroup asyncio
1197 * Free a transfer structure. This should be called for all transfers
1198 * allocated with libusb_alloc_transfer().
1200 * If the \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_BUFFER
1201 * "LIBUSB_TRANSFER_FREE_BUFFER" flag is set and the transfer buffer is
1202 * non-NULL, this function will also free the transfer buffer using the
1203 * standard system memory allocator (e.g. free()).
1205 * It is legal to call this function with a NULL transfer. In this case,
1206 * the function will simply return safely.
1208 * It is not legal to free an active transfer (one which has been submitted
1209 * and has not yet completed).
1211 * \param transfer the transfer to free
1213 API_EXPORTED void libusb_free_transfer(struct libusb_transfer *transfer)
1215 struct usbi_transfer *itransfer;
1219 if (transfer->flags & LIBUSB_TRANSFER_FREE_BUFFER && transfer->buffer)
1220 free(transfer->buffer);
1222 itransfer = __LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1223 usbi_mutex_destroy(&itransfer->lock);
1227 /** \ingroup asyncio
1228 * Submit a transfer. This function will fire off the USB transfer and then
1229 * return immediately.
1231 * \param transfer the transfer to submit
1232 * \returns 0 on success
1233 * \returns LIBUSB_ERROR_NO_DEVICE if the device has been disconnected
1234 * \returns LIBUSB_ERROR_BUSY if the transfer has already been submitted.
1235 * \returns another LIBUSB_ERROR code on other failure
1237 API_EXPORTED int libusb_submit_transfer(struct libusb_transfer *transfer)
1239 struct libusb_context *ctx = TRANSFER_CTX(transfer);
1240 struct usbi_transfer *itransfer =
1241 __LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1245 usbi_mutex_lock(&itransfer->lock);
1246 itransfer->transferred = 0;
1247 itransfer->flags = 0;
1248 r = calculate_timeout(itransfer);
1250 r = LIBUSB_ERROR_OTHER;
1254 first = add_to_flying_list(itransfer);
1255 r = usbi_backend->submit_transfer(itransfer);
1257 usbi_mutex_lock(&ctx->flying_transfers_lock);
1258 list_del(&itransfer->list);
1259 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1261 #ifdef USBI_TIMERFD_AVAILABLE
1262 else if (first && usbi_using_timerfd(ctx)) {
1263 /* if this transfer has the lowest timeout of all active transfers,
1264 * rearm the timerfd with this transfer's timeout */
1265 const struct itimerspec it = { {0, 0},
1266 { itransfer->timeout.tv_sec, itransfer->timeout.tv_usec * 1000 } };
1267 usbi_dbg("arm timerfd for timeout in %dms (first in line)", transfer->timeout);
1268 r = timerfd_settime(ctx->timerfd, TFD_TIMER_ABSTIME, &it, NULL);
1270 r = LIBUSB_ERROR_OTHER;
1275 usbi_mutex_unlock(&itransfer->lock);
1279 /** \ingroup asyncio
1280 * Asynchronously cancel a previously submitted transfer.
1281 * This function returns immediately, but this does not indicate cancellation
1282 * is complete. Your callback function will be invoked at some later time
1283 * with a transfer status of
1284 * \ref libusb_transfer_status::LIBUSB_TRANSFER_CANCELLED
1285 * "LIBUSB_TRANSFER_CANCELLED."
1287 * \param transfer the transfer to cancel
1288 * \returns 0 on success
1289 * \returns LIBUSB_ERROR_NOT_FOUND if the transfer is already complete or
1291 * \returns a LIBUSB_ERROR code on failure
1293 API_EXPORTED int libusb_cancel_transfer(struct libusb_transfer *transfer)
1295 struct usbi_transfer *itransfer =
1296 __LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1300 usbi_mutex_lock(&itransfer->lock);
1301 r = usbi_backend->cancel_transfer(itransfer);
1303 usbi_err(TRANSFER_CTX(transfer),
1304 "cancel transfer failed error %d", r);
1305 usbi_mutex_unlock(&itransfer->lock);
1309 #ifdef USBI_TIMERFD_AVAILABLE
1310 static int disarm_timerfd(struct libusb_context *ctx)
1312 const struct itimerspec disarm_timer = { { 0, 0 }, { 0, 0 } };
1316 r = timerfd_settime(ctx->timerfd, 0, &disarm_timer, NULL);
1318 return LIBUSB_ERROR_OTHER;
1323 /* iterates through the flying transfers, and rearms the timerfd based on the
1324 * next upcoming timeout.
1325 * must be called with flying_list locked.
1326 * returns 0 if there was no timeout to arm, 1 if the next timeout was armed,
1327 * or a LIBUSB_ERROR code on failure.
1329 static int arm_timerfd_for_next_timeout(struct libusb_context *ctx)
1331 struct usbi_transfer *transfer;
1333 list_for_each_entry(transfer, &ctx->flying_transfers, list) {
1334 struct timeval *cur_tv = &transfer->timeout;
1336 /* if we've reached transfers of infinite timeout, then we have no
1338 if (!timerisset(cur_tv))
1341 /* act on first transfer that is not already cancelled */
1342 if (!(transfer->flags & USBI_TRANSFER_TIMED_OUT)) {
1344 const struct itimerspec it = { {0, 0},
1345 { cur_tv->tv_sec, cur_tv->tv_usec * 1000 } };
1346 usbi_dbg("next timeout originally %dms", __USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout);
1347 r = timerfd_settime(ctx->timerfd, TFD_TIMER_ABSTIME, &it, NULL);
1349 return LIBUSB_ERROR_OTHER;
1357 static int disarm_timerfd(struct libusb_context *ctx)
1361 static int arm_timerfd_for_next_timeout(struct libusb_context *ctx)
1367 /* Handle completion of a transfer (completion might be an error condition).
1368 * This will invoke the user-supplied callback function, which may end up
1369 * freeing the transfer. Therefore you cannot use the transfer structure
1370 * after calling this function, and you should free all backend-specific
1371 * data before calling it.
1372 * Do not call this function with the usbi_transfer lock held. User-specified
1373 * callback functions may attempt to directly resubmit the transfer, which
1374 * will attempt to take the lock. */
1375 int usbi_handle_transfer_completion(struct usbi_transfer *itransfer,
1376 enum libusb_transfer_status status)
1378 struct libusb_transfer *transfer =
1379 __USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1380 struct libusb_context *ctx = TRANSFER_CTX(transfer);
1384 /* FIXME: could be more intelligent with the timerfd here. we don't need
1385 * to disarm the timerfd if there was no timer running, and we only need
1386 * to rearm the timerfd if the transfer that expired was the one with
1387 * the shortest timeout. */
1389 usbi_mutex_lock(&ctx->flying_transfers_lock);
1390 list_del(&itransfer->list);
1391 r = arm_timerfd_for_next_timeout(ctx);
1392 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1396 } else if (r == 0) {
1397 r = disarm_timerfd(ctx);
1402 if (status == LIBUSB_TRANSFER_COMPLETED
1403 && transfer->flags & LIBUSB_TRANSFER_SHORT_NOT_OK) {
1404 int rqlen = transfer->length;
1405 if (transfer->type == LIBUSB_TRANSFER_TYPE_CONTROL)
1406 rqlen -= LIBUSB_CONTROL_SETUP_SIZE;
1407 if (rqlen != itransfer->transferred) {
1408 usbi_dbg("interpreting short transfer as error");
1409 status = LIBUSB_TRANSFER_ERROR;
1413 flags = transfer->flags;
1414 transfer->status = status;
1415 transfer->actual_length = itransfer->transferred;
1416 if (transfer->callback)
1417 transfer->callback(transfer);
1418 /* transfer might have been freed by the above call, do not use from
1420 if (flags & LIBUSB_TRANSFER_FREE_TRANSFER)
1421 libusb_free_transfer(transfer);
1422 usbi_mutex_lock(&ctx->event_waiters_lock);
1423 usbi_cond_broadcast(&ctx->event_waiters_cond);
1424 usbi_mutex_unlock(&ctx->event_waiters_lock);
1428 /* Similar to usbi_handle_transfer_completion() but exclusively for transfers
1429 * that were asynchronously cancelled. The same concerns w.r.t. freeing of
1430 * transfers exist here.
1431 * Do not call this function with the usbi_transfer lock held. User-specified
1432 * callback functions may attempt to directly resubmit the transfer, which
1433 * will attempt to take the lock. */
1434 int usbi_handle_transfer_cancellation(struct usbi_transfer *transfer)
1436 /* if the URB was cancelled due to timeout, report timeout to the user */
1437 if (transfer->flags & USBI_TRANSFER_TIMED_OUT) {
1438 usbi_dbg("detected timeout cancellation");
1439 return usbi_handle_transfer_completion(transfer, LIBUSB_TRANSFER_TIMED_OUT);
1442 /* otherwise its a normal async cancel */
1443 return usbi_handle_transfer_completion(transfer, LIBUSB_TRANSFER_CANCELLED);
1447 * Attempt to acquire the event handling lock. This lock is used to ensure that
1448 * only one thread is monitoring libusb event sources at any one time.
1450 * You only need to use this lock if you are developing an application
1451 * which calls poll() or select() on libusb's file descriptors directly.
1452 * If you stick to libusb's event handling loop functions (e.g.
1453 * libusb_handle_events()) then you do not need to be concerned with this
1456 * While holding this lock, you are trusted to actually be handling events.
1457 * If you are no longer handling events, you must call libusb_unlock_events()
1458 * as soon as possible.
1460 * \param ctx the context to operate on, or NULL for the default context
1461 * \returns 0 if the lock was obtained successfully
1462 * \returns 1 if the lock was not obtained (i.e. another thread holds the lock)
1465 API_EXPORTED int libusb_try_lock_events(libusb_context *ctx)
1468 USBI_GET_CONTEXT(ctx);
1470 /* is someone else waiting to modify poll fds? if so, don't let this thread
1471 * start event handling */
1472 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1473 r = ctx->pollfd_modify;
1474 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1476 usbi_dbg("someone else is modifying poll fds");
1480 r = usbi_mutex_trylock(&ctx->events_lock);
1484 ctx->event_handler_active = 1;
1489 * Acquire the event handling lock, blocking until successful acquisition if
1490 * it is contended. This lock is used to ensure that only one thread is
1491 * monitoring libusb event sources at any one time.
1493 * You only need to use this lock if you are developing an application
1494 * which calls poll() or select() on libusb's file descriptors directly.
1495 * If you stick to libusb's event handling loop functions (e.g.
1496 * libusb_handle_events()) then you do not need to be concerned with this
1499 * While holding this lock, you are trusted to actually be handling events.
1500 * If you are no longer handling events, you must call libusb_unlock_events()
1501 * as soon as possible.
1503 * \param ctx the context to operate on, or NULL for the default context
1506 API_EXPORTED void libusb_lock_events(libusb_context *ctx)
1508 USBI_GET_CONTEXT(ctx);
1509 usbi_mutex_lock(&ctx->events_lock);
1510 ctx->event_handler_active = 1;
1514 * Release the lock previously acquired with libusb_try_lock_events() or
1515 * libusb_lock_events(). Releasing this lock will wake up any threads blocked
1516 * on libusb_wait_for_event().
1518 * \param ctx the context to operate on, or NULL for the default context
1521 API_EXPORTED void libusb_unlock_events(libusb_context *ctx)
1523 USBI_GET_CONTEXT(ctx);
1524 ctx->event_handler_active = 0;
1525 usbi_mutex_unlock(&ctx->events_lock);
1527 /* FIXME: perhaps we should be a bit more efficient by not broadcasting
1528 * the availability of the events lock when we are modifying pollfds
1529 * (check ctx->pollfd_modify)? */
1530 usbi_mutex_lock(&ctx->event_waiters_lock);
1531 usbi_cond_broadcast(&ctx->event_waiters_cond);
1532 usbi_mutex_unlock(&ctx->event_waiters_lock);
1536 * Determine if it is still OK for this thread to be doing event handling.
1538 * Sometimes, libusb needs to temporarily pause all event handlers, and this
1539 * is the function you should use before polling file descriptors to see if
1542 * If this function instructs your thread to give up the events lock, you
1543 * should just continue the usual logic that is documented in \ref mtasync.
1544 * On the next iteration, your thread will fail to obtain the events lock,
1545 * and will hence become an event waiter.
1547 * This function should be called while the events lock is held: you don't
1548 * need to worry about the results of this function if your thread is not
1549 * the current event handler.
1551 * \param ctx the context to operate on, or NULL for the default context
1552 * \returns 1 if event handling can start or continue
1553 * \returns 0 if this thread must give up the events lock
1554 * \see \ref fullstory "Multi-threaded I/O: the full story"
1556 API_EXPORTED int libusb_event_handling_ok(libusb_context *ctx)
1559 USBI_GET_CONTEXT(ctx);
1561 /* is someone else waiting to modify poll fds? if so, don't let this thread
1562 * continue event handling */
1563 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1564 r = ctx->pollfd_modify;
1565 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1567 usbi_dbg("someone else is modifying poll fds");
1576 * Determine if an active thread is handling events (i.e. if anyone is holding
1577 * the event handling lock).
1579 * \param ctx the context to operate on, or NULL for the default context
1580 * \returns 1 if a thread is handling events
1581 * \returns 0 if there are no threads currently handling events
1584 API_EXPORTED int libusb_event_handler_active(libusb_context *ctx)
1587 USBI_GET_CONTEXT(ctx);
1589 /* is someone else waiting to modify poll fds? if so, don't let this thread
1590 * start event handling -- indicate that event handling is happening */
1591 usbi_mutex_lock(&ctx->pollfd_modify_lock);
1592 r = ctx->pollfd_modify;
1593 usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1595 usbi_dbg("someone else is modifying poll fds");
1599 return ctx->event_handler_active;
1603 * Acquire the event waiters lock. This lock is designed to be obtained under
1604 * the situation where you want to be aware when events are completed, but
1605 * some other thread is event handling so calling libusb_handle_events() is not
1608 * You then obtain this lock, re-check that another thread is still handling
1609 * events, then call libusb_wait_for_event().
1611 * You only need to use this lock if you are developing an application
1612 * which calls poll() or select() on libusb's file descriptors directly,
1613 * <b>and</b> may potentially be handling events from 2 threads simultaenously.
1614 * If you stick to libusb's event handling loop functions (e.g.
1615 * libusb_handle_events()) then you do not need to be concerned with this
1618 * \param ctx the context to operate on, or NULL for the default context
1621 API_EXPORTED void libusb_lock_event_waiters(libusb_context *ctx)
1623 USBI_GET_CONTEXT(ctx);
1624 usbi_mutex_lock(&ctx->event_waiters_lock);
1628 * Release the event waiters lock.
1629 * \param ctx the context to operate on, or NULL for the default context
1632 API_EXPORTED void libusb_unlock_event_waiters(libusb_context *ctx)
1634 USBI_GET_CONTEXT(ctx);
1635 usbi_mutex_unlock(&ctx->event_waiters_lock);
1639 * Wait for another thread to signal completion of an event. Must be called
1640 * with the event waiters lock held, see libusb_lock_event_waiters().
1642 * This function will block until any of the following conditions are met:
1643 * -# The timeout expires
1644 * -# A transfer completes
1645 * -# A thread releases the event handling lock through libusb_unlock_events()
1647 * Condition 1 is obvious. Condition 2 unblocks your thread <em>after</em>
1648 * the callback for the transfer has completed. Condition 3 is important
1649 * because it means that the thread that was previously handling events is no
1650 * longer doing so, so if any events are to complete, another thread needs to
1651 * step up and start event handling.
1653 * This function releases the event waiters lock before putting your thread
1654 * to sleep, and reacquires the lock as it is being woken up.
1656 * \param ctx the context to operate on, or NULL for the default context
1657 * \param tv maximum timeout for this blocking function. A NULL value
1658 * indicates unlimited timeout.
1659 * \returns 0 after a transfer completes or another thread stops event handling
1660 * \returns 1 if the timeout expired
1663 API_EXPORTED int libusb_wait_for_event(libusb_context *ctx, struct timeval *tv)
1665 struct timespec timeout;
1668 USBI_GET_CONTEXT(ctx);
1670 usbi_cond_wait(&ctx->event_waiters_cond, &ctx->event_waiters_lock);
1674 r = usbi_backend->clock_gettime(USBI_CLOCK_REALTIME, &timeout);
1676 usbi_err(ctx, "failed to read realtime clock, error %d", errno);
1677 return LIBUSB_ERROR_OTHER;
1680 timeout.tv_sec += tv->tv_sec;
1681 timeout.tv_nsec += tv->tv_usec * 1000;
1682 if (timeout.tv_nsec > 1000000000) {
1683 timeout.tv_nsec -= 1000000000;
1687 r = usbi_cond_timedwait(&ctx->event_waiters_cond,
1688 &ctx->event_waiters_lock, &timeout);
1689 return (r == ETIMEDOUT);
1692 static void handle_timeout(struct usbi_transfer *itransfer)
1694 struct libusb_transfer *transfer =
1695 __USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1698 itransfer->flags |= USBI_TRANSFER_TIMED_OUT;
1699 r = libusb_cancel_transfer(transfer);
1701 usbi_warn(TRANSFER_CTX(transfer),
1702 "async cancel failed %d errno=%d", r, errno);
1705 #ifdef USBI_OS_HANDLES_TIMEOUT
1706 static int handle_timeouts_locked(struct libusb_context *ctx)
1710 static int handle_timeouts(struct libusb_context *ctx)
1715 static int handle_timeouts_locked(struct libusb_context *ctx)
1718 struct timespec systime_ts;
1719 struct timeval systime;
1720 struct usbi_transfer *transfer;
1722 if (list_empty(&ctx->flying_transfers))
1725 /* get current time */
1726 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, &systime_ts);
1730 TIMESPEC_TO_TIMEVAL(&systime, &systime_ts);
1732 /* iterate through flying transfers list, finding all transfers that
1733 * have expired timeouts */
1734 list_for_each_entry(transfer, &ctx->flying_transfers, list) {
1735 struct timeval *cur_tv = &transfer->timeout;
1737 /* if we've reached transfers of infinite timeout, we're all done */
1738 if (!timerisset(cur_tv))
1741 /* ignore timeouts we've already handled */
1742 if (transfer->flags & USBI_TRANSFER_TIMED_OUT)
1745 /* if transfer has non-expired timeout, nothing more to do */
1746 if ((cur_tv->tv_sec > systime.tv_sec) ||
1747 (cur_tv->tv_sec == systime.tv_sec &&
1748 cur_tv->tv_usec > systime.tv_usec))
1751 /* otherwise, we've got an expired timeout to handle */
1752 handle_timeout(transfer);
1757 static int handle_timeouts(struct libusb_context *ctx)
1760 USBI_GET_CONTEXT(ctx);
1761 usbi_mutex_lock(&ctx->flying_transfers_lock);
1762 r = handle_timeouts_locked(ctx);
1763 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1768 #ifdef USBI_TIMERFD_AVAILABLE
1769 static int handle_timerfd_trigger(struct libusb_context *ctx)
1773 r = disarm_timerfd(ctx);
1777 usbi_mutex_lock(&ctx->flying_transfers_lock);
1779 /* process the timeout that just happened */
1780 r = handle_timeouts_locked(ctx);
1784 /* arm for next timeout*/
1785 r = arm_timerfd_for_next_timeout(ctx);
1788 usbi_mutex_unlock(&ctx->flying_transfers_lock);
1793 /* do the actual event handling. assumes that no other thread is concurrently
1794 * doing the same thing. */
1795 static int handle_events(struct libusb_context *ctx, struct timeval *tv)
1798 struct usbi_pollfd *ipollfd;
1804 usbi_mutex_lock(&ctx->pollfds_lock);
1805 list_for_each_entry(ipollfd, &ctx->pollfds, list)
1808 /* TODO: malloc when number of fd's changes, not on every poll */
1809 fds = malloc(sizeof(*fds) * nfds);
1811 usbi_mutex_unlock(&ctx->pollfds_lock);
1812 return LIBUSB_ERROR_NO_MEM;
1815 list_for_each_entry(ipollfd, &ctx->pollfds, list) {
1816 struct libusb_pollfd *pollfd = &ipollfd->pollfd;
1817 int fd = pollfd->fd;
1820 fds[i].events = pollfd->events;
1823 usbi_mutex_unlock(&ctx->pollfds_lock);
1825 timeout_ms = (tv->tv_sec * 1000) + (tv->tv_usec / 1000);
1827 /* round up to next millisecond */
1828 if (tv->tv_usec % 1000)
1831 usbi_dbg("poll() %d fds with timeout in %dms", nfds, timeout_ms);
1832 r = poll(fds, nfds, timeout_ms);
1833 usbi_dbg("poll() returned %d", r);
1836 return handle_timeouts(ctx);
1837 } else if (r == -1 && errno == EINTR) {
1839 return LIBUSB_ERROR_INTERRUPTED;
1842 usbi_err(ctx, "poll failed %d err=%d\n", r, errno);
1843 return LIBUSB_ERROR_IO;
1846 /* fd[0] is always the ctrl pipe */
1847 if (fds[0].revents) {
1848 /* another thread wanted to interrupt event handling, and it succeeded!
1849 * handle any other events that cropped up at the same time, and
1851 usbi_dbg("caught a fish on the control pipe");
1857 /* prevent OS backend from trying to handle events on ctrl pipe */
1863 #ifdef USBI_TIMERFD_AVAILABLE
1864 /* on timerfd configurations, fds[1] is the timerfd */
1865 if (usbi_using_timerfd(ctx) && fds[1].revents) {
1866 /* timerfd indicates that a timeout has expired */
1868 usbi_dbg("timerfd triggered");
1870 ret = handle_timerfd_trigger(ctx);
1872 /* return error code */
1875 } else if (r == 1) {
1876 /* no more active file descriptors, nothing more to do */
1880 /* more events pending...
1881 * prevent OS backend from trying to handle events on timerfd */
1888 r = usbi_backend->handle_events(ctx, fds, nfds, r);
1890 usbi_err(ctx, "backend handle_events failed with error %d", r);
1897 /* returns the smallest of:
1898 * 1. timeout of next URB
1899 * 2. user-supplied timeout
1900 * returns 1 if there is an already-expired timeout, otherwise returns 0
1903 static int get_next_timeout(libusb_context *ctx, struct timeval *tv,
1904 struct timeval *out)
1906 struct timeval timeout;
1907 int r = libusb_get_next_timeout(ctx, &timeout);
1909 /* timeout already expired? */
1910 if (!timerisset(&timeout))
1913 /* choose the smallest of next URB timeout or user specified timeout */
1914 if (timercmp(&timeout, tv, <))
1925 * Handle any pending events.
1927 * libusb determines "pending events" by checking if any timeouts have expired
1928 * and by checking the set of file descriptors for activity.
1930 * If a zero timeval is passed, this function will handle any already-pending
1931 * events and then immediately return in non-blocking style.
1933 * If a non-zero timeval is passed and no events are currently pending, this
1934 * function will block waiting for events to handle up until the specified
1935 * timeout. If an event arrives or a signal is raised, this function will
1938 * \param ctx the context to operate on, or NULL for the default context
1939 * \param tv the maximum time to block waiting for events, or zero for
1941 * \returns 0 on success, or a LIBUSB_ERROR code on failure
1943 API_EXPORTED int libusb_handle_events_timeout(libusb_context *ctx,
1947 struct timeval poll_timeout;
1949 USBI_GET_CONTEXT(ctx);
1950 r = get_next_timeout(ctx, tv, &poll_timeout);
1952 /* timeout already expired */
1953 return handle_timeouts(ctx);
1957 if (libusb_try_lock_events(ctx) == 0) {
1958 /* we obtained the event lock: do our own event handling */
1959 r = handle_events(ctx, &poll_timeout);
1960 libusb_unlock_events(ctx);
1964 /* another thread is doing event handling. wait for pthread events that
1965 * notify event completion. */
1966 libusb_lock_event_waiters(ctx);
1968 if (!libusb_event_handler_active(ctx)) {
1969 /* we hit a race: whoever was event handling earlier finished in the
1970 * time it took us to reach this point. try the cycle again. */
1971 libusb_unlock_event_waiters(ctx);
1972 usbi_dbg("event handler was active but went away, retrying");
1976 usbi_dbg("another thread is doing event handling");
1977 r = libusb_wait_for_event(ctx, &poll_timeout);
1978 libusb_unlock_event_waiters(ctx);
1983 return handle_timeouts(ctx);
1989 * Handle any pending events in blocking mode. There is currently a timeout
1990 * hardcoded at 60 seconds but we plan to make it unlimited in future. For
1991 * finer control over whether this function is blocking or non-blocking, or
1992 * for control over the timeout, use libusb_handle_events_timeout() instead.
1994 * \param ctx the context to operate on, or NULL for the default context
1995 * \returns 0 on success, or a LIBUSB_ERROR code on failure
1997 API_EXPORTED int libusb_handle_events(libusb_context *ctx)
2002 return libusb_handle_events_timeout(ctx, &tv);
2006 * Handle any pending events by polling file descriptors, without checking if
2007 * any other threads are already doing so. Must be called with the event lock
2008 * held, see libusb_lock_events().
2010 * This function is designed to be called under the situation where you have
2011 * taken the event lock and are calling poll()/select() directly on libusb's
2012 * file descriptors (as opposed to using libusb_handle_events() or similar).
2013 * You detect events on libusb's descriptors, so you then call this function
2014 * with a zero timeout value (while still holding the event lock).
2016 * \param ctx the context to operate on, or NULL for the default context
2017 * \param tv the maximum time to block waiting for events, or zero for
2019 * \returns 0 on success, or a LIBUSB_ERROR code on failure
2022 API_EXPORTED int libusb_handle_events_locked(libusb_context *ctx,
2026 struct timeval poll_timeout;
2028 USBI_GET_CONTEXT(ctx);
2029 r = get_next_timeout(ctx, tv, &poll_timeout);
2031 /* timeout already expired */
2032 return handle_timeouts(ctx);
2035 return handle_events(ctx, &poll_timeout);
2039 * Determines whether your application must apply special timing considerations
2040 * when monitoring libusb's file descriptors.
2042 * This function is only useful for applications which retrieve and poll
2043 * libusb's file descriptors in their own main loop (\ref pollmain).
2045 * Ordinarily, libusb's event handler needs to be called into at specific
2046 * moments in time (in addition to times when there is activity on the file
2047 * descriptor set). The usual approach is to use libusb_get_next_timeout()
2048 * to learn about when the next timeout occurs, and to adjust your
2049 * poll()/select() timeout accordingly so that you can make a call into the
2050 * library at that time.
2052 * Some platforms supported by libusb do not come with this baggage - any
2053 * events relevant to timing will be represented by activity on the file
2054 * descriptor set, and libusb_get_next_timeout() will always return 0.
2055 * This function allows you to detect whether you are running on such a
2060 * \param ctx the context to operate on, or NULL for the default context
2061 * \returns 0 if you must call into libusb at times determined by
2062 * libusb_get_next_timeout(), or 1 if all timeout events are handled internally
2063 * or through regular activity on the file descriptors.
2064 * \see \ref pollmain "Polling libusb file descriptors for event handling"
2066 API_EXPORTED int libusb_pollfds_handle_timeouts(libusb_context *ctx)
2068 #if defined(USBI_OS_HANDLES_TIMEOUT)
2070 #elif defined(USBI_TIMERFD_AVAILABLE)
2071 USBI_GET_CONTEXT(ctx);
2072 return usbi_using_timerfd(ctx);
2079 * Determine the next internal timeout that libusb needs to handle. You only
2080 * need to use this function if you are calling poll() or select() or similar
2081 * on libusb's file descriptors yourself - you do not need to use it if you
2082 * are calling libusb_handle_events() or a variant directly.
2084 * You should call this function in your main loop in order to determine how
2085 * long to wait for select() or poll() to return results. libusb needs to be
2086 * called into at this timeout, so you should use it as an upper bound on
2087 * your select() or poll() call.
2089 * When the timeout has expired, call into libusb_handle_events_timeout()
2090 * (perhaps in non-blocking mode) so that libusb can handle the timeout.
2092 * This function may return 1 (success) and an all-zero timeval. If this is
2093 * the case, it indicates that libusb has a timeout that has already expired
2094 * so you should call libusb_handle_events_timeout() or similar immediately.
2095 * A return code of 0 indicates that there are no pending timeouts.
2097 * On some platforms, this function will always returns 0 (no pending
2098 * timeouts). See \ref polltime.
2100 * \param ctx the context to operate on, or NULL for the default context
2101 * \param tv output location for a relative time against the current
2102 * clock in which libusb must be called into in order to process timeout events
2103 * \returns 0 if there are no pending timeouts, 1 if a timeout was returned,
2104 * or LIBUSB_ERROR_OTHER on failure
2106 API_EXPORTED int libusb_get_next_timeout(libusb_context *ctx,
2109 #ifndef USBI_OS_HANDLES_TIMEOUT
2110 struct usbi_transfer *transfer;
2111 struct timespec cur_ts;
2112 struct timeval cur_tv;
2113 struct timeval *next_timeout;
2117 USBI_GET_CONTEXT(ctx);
2118 if (usbi_using_timerfd(ctx))
2121 usbi_mutex_lock(&ctx->flying_transfers_lock);
2122 if (list_empty(&ctx->flying_transfers)) {
2123 usbi_mutex_unlock(&ctx->flying_transfers_lock);
2124 usbi_dbg("no URBs, no timeout!");
2128 /* find next transfer which hasn't already been processed as timed out */
2129 list_for_each_entry(transfer, &ctx->flying_transfers, list) {
2130 if (!(transfer->flags & USBI_TRANSFER_TIMED_OUT)) {
2135 usbi_mutex_unlock(&ctx->flying_transfers_lock);
2138 usbi_dbg("all URBs have already been processed for timeouts");
2142 next_timeout = &transfer->timeout;
2144 /* no timeout for next transfer */
2145 if (!timerisset(next_timeout)) {
2146 usbi_dbg("no URBs with timeouts, no timeout!");
2150 r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, &cur_ts);
2152 usbi_err(ctx, "failed to read monotonic clock, errno=%d", errno);
2153 return LIBUSB_ERROR_OTHER;
2155 TIMESPEC_TO_TIMEVAL(&cur_tv, &cur_ts);
2157 if (timercmp(&cur_tv, next_timeout, >=)) {
2158 usbi_dbg("first timeout already expired");
2161 timersub(next_timeout, &cur_tv, tv);
2162 usbi_dbg("next timeout in %d.%06ds", tv->tv_sec, tv->tv_usec);
2172 * Register notification functions for file descriptor additions/removals.
2173 * These functions will be invoked for every new or removed file descriptor
2174 * that libusb uses as an event source.
2176 * To remove notifiers, pass NULL values for the function pointers.
2178 * Note that file descriptors may have been added even before you register
2179 * these notifiers (e.g. at libusb_init() time).
2181 * Additionally, note that the removal notifier may be called during
2182 * libusb_exit() (e.g. when it is closing file descriptors that were opened
2183 * and added to the poll set at libusb_init() time). If you don't want this,
2184 * remove the notifiers immediately before calling libusb_exit().
2186 * \param ctx the context to operate on, or NULL for the default context
2187 * \param added_cb pointer to function for addition notifications
2188 * \param removed_cb pointer to function for removal notifications
2189 * \param user_data User data to be passed back to callbacks (useful for
2190 * passing context information)
2192 API_EXPORTED void libusb_set_pollfd_notifiers(libusb_context *ctx,
2193 libusb_pollfd_added_cb added_cb, libusb_pollfd_removed_cb removed_cb,
2196 USBI_GET_CONTEXT(ctx);
2197 ctx->fd_added_cb = added_cb;
2198 ctx->fd_removed_cb = removed_cb;
2199 ctx->fd_cb_user_data = user_data;
2202 /* Add a file descriptor to the list of file descriptors to be monitored.
2203 * events should be specified as a bitmask of events passed to poll(), e.g.
2204 * POLLIN and/or POLLOUT. */
2205 int usbi_add_pollfd(struct libusb_context *ctx, int fd, short events)
2207 struct usbi_pollfd *ipollfd = malloc(sizeof(*ipollfd));
2209 return LIBUSB_ERROR_NO_MEM;
2211 usbi_dbg("add fd %d events %d", fd, events);
2212 ipollfd->pollfd.fd = fd;
2213 ipollfd->pollfd.events = events;
2214 usbi_mutex_lock(&ctx->pollfds_lock);
2215 list_add_tail(&ipollfd->list, &ctx->pollfds);
2216 usbi_mutex_unlock(&ctx->pollfds_lock);
2218 if (ctx->fd_added_cb)
2219 ctx->fd_added_cb(fd, events, ctx->fd_cb_user_data);
2223 /* Remove a file descriptor from the list of file descriptors to be polled. */
2224 void usbi_remove_pollfd(struct libusb_context *ctx, int fd)
2226 struct usbi_pollfd *ipollfd;
2229 usbi_dbg("remove fd %d", fd);
2230 usbi_mutex_lock(&ctx->pollfds_lock);
2231 list_for_each_entry(ipollfd, &ctx->pollfds, list)
2232 if (ipollfd->pollfd.fd == fd) {
2238 usbi_dbg("couldn't find fd %d to remove", fd);
2239 usbi_mutex_unlock(&ctx->pollfds_lock);
2243 list_del(&ipollfd->list);
2244 usbi_mutex_unlock(&ctx->pollfds_lock);
2246 if (ctx->fd_removed_cb)
2247 ctx->fd_removed_cb(fd, ctx->fd_cb_user_data);
2251 * Retrieve a list of file descriptors that should be polled by your main loop
2252 * as libusb event sources.
2254 * The returned list is NULL-terminated and should be freed with free() when
2255 * done. The actual list contents must not be touched.
2257 * \param ctx the context to operate on, or NULL for the default context
2258 * \returns a NULL-terminated list of libusb_pollfd structures, or NULL on
2261 API_EXPORTED const struct libusb_pollfd **libusb_get_pollfds(
2262 libusb_context *ctx)
2264 struct libusb_pollfd **ret = NULL;
2265 struct usbi_pollfd *ipollfd;
2268 USBI_GET_CONTEXT(ctx);
2270 usbi_mutex_lock(&ctx->pollfds_lock);
2271 list_for_each_entry(ipollfd, &ctx->pollfds, list)
2274 ret = calloc(cnt + 1, sizeof(struct libusb_pollfd *));
2278 list_for_each_entry(ipollfd, &ctx->pollfds, list)
2279 ret[i++] = (struct libusb_pollfd *) ipollfd;
2283 usbi_mutex_unlock(&ctx->pollfds_lock);
2284 return (const struct libusb_pollfd **) ret;
2287 /* Backends call this from handle_events to report disconnection of a device.
2288 * The transfers get cancelled appropriately.
2290 void usbi_handle_disconnect(struct libusb_device_handle *handle)
2292 struct usbi_transfer *cur;
2293 struct usbi_transfer *to_cancel;
2295 usbi_dbg("device %d.%d",
2296 handle->dev->bus_number, handle->dev->device_address);
2298 /* terminate all pending transfers with the LIBUSB_TRANSFER_NO_DEVICE
2301 * this is a bit tricky because:
2302 * 1. we can't do transfer completion while holding flying_transfers_lock
2303 * 2. the transfers list can change underneath us - if we were to build a
2304 * list of transfers to complete (while holding look), the situation
2305 * might be different by the time we come to free them
2307 * so we resort to a loop-based approach as below
2308 * FIXME: is this still potentially racy?
2312 usbi_mutex_lock(&HANDLE_CTX(handle)->flying_transfers_lock);
2314 list_for_each_entry(cur, &HANDLE_CTX(handle)->flying_transfers, list)
2315 if (__USBI_TRANSFER_TO_LIBUSB_TRANSFER(cur)->dev_handle == handle) {
2319 usbi_mutex_unlock(&HANDLE_CTX(handle)->flying_transfers_lock);
2324 usbi_backend->clear_transfer_priv(to_cancel);
2325 usbi_handle_transfer_completion(to_cancel, LIBUSB_TRANSFER_NO_DEVICE);