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
28
#include "os/poll_posix.h"
31
#ifdef USBI_TIMERFD_AVAILABLE
32
#include <sys/timerfd.h>
38
* \page io Synchronous and asynchronous device I/O
40
* \section intro Introduction
42
* If you're using libusb in your application, you're probably wanting to
43
* perform I/O with devices - you want to perform USB data transfers.
45
* libusb offers two separate interfaces for device I/O. This page aims to
46
* introduce the two in order to help you decide which one is more suitable
47
* for your application. You can also choose to use both interfaces in your
48
* application by considering each transfer on a case-by-case basis.
50
* Once you have read through the following discussion, you should consult the
51
* detailed API documentation pages for the details:
55
* \section theory Transfers at a logical level
57
* At a logical level, USB transfers typically happen in two parts. For
58
* example, when reading data from a endpoint:
59
* -# A request for data is sent to the device
60
* -# Some time later, the incoming data is received by the host
62
* or when writing data to an endpoint:
64
* -# The data is sent to the device
65
* -# Some time later, the host receives acknowledgement from the device that
66
* the data has been transferred.
68
* There may be an indefinite delay between the two steps. Consider a
69
* fictional USB input device with a button that the user can press. In order
70
* to determine when the button is pressed, you would likely submit a request
71
* to read data on a bulk or interrupt endpoint and wait for data to arrive.
72
* Data will arrive when the button is pressed by the user, which is
73
* potentially hours later.
75
* libusb offers both a synchronous and an asynchronous interface to performing
76
* USB transfers. The main difference is that the synchronous interface
77
* combines both steps indicated above into a single function call, whereas
78
* the asynchronous interface separates them.
80
* \section sync The synchronous interface
82
* The synchronous I/O interface allows you to perform a USB transfer with
83
* a single function call. When the function call returns, the transfer has
84
* completed and you can parse the results.
86
* If you have used the libusb-0.1 before, this I/O style will seem familar to
87
* you. libusb-0.1 only offered a synchronous interface.
89
* In our input device example, to read button presses you might write code
90
* in the following style:
92
unsigned char data[4];
94
int r = libusb_bulk_transfer(handle, EP_IN, data, sizeof(data), &actual_length, 0);
95
if (r == 0 && actual_length == sizeof(data)) {
96
// results of the transaction can now be found in the data buffer
97
// parse them here and report button press
103
* The main advantage of this model is simplicity: you did everything with
104
* a single simple function call.
106
* However, this interface has its limitations. Your application will sleep
107
* inside libusb_bulk_transfer() until the transaction has completed. If it
108
* takes the user 3 hours to press the button, your application will be
109
* sleeping for that long. Execution will be tied up inside the library -
110
* the entire thread will be useless for that duration.
112
* Another issue is that by tieing up the thread with that single transaction
113
* there is no possibility of performing I/O with multiple endpoints and/or
114
* multiple devices simultaneously, unless you resort to creating one thread
117
* Additionally, there is no opportunity to cancel the transfer after the
118
* request has been submitted.
120
* For details on how to use the synchronous API, see the
121
* \ref syncio "synchronous I/O API documentation" pages.
123
* \section async The asynchronous interface
125
* Asynchronous I/O is the most significant new feature in libusb-1.0.
126
* Although it is a more complex interface, it solves all the issues detailed
129
* Instead of providing which functions that block until the I/O has complete,
130
* libusb's asynchronous interface presents non-blocking functions which
131
* begin a transfer and then return immediately. Your application passes a
132
* callback function pointer to this non-blocking function, which libusb will
133
* call with the results of the transaction when it has completed.
135
* Transfers which have been submitted through the non-blocking functions
136
* can be cancelled with a separate function call.
138
* The non-blocking nature of this interface allows you to be simultaneously
139
* performing I/O to multiple endpoints on multiple devices, without having
142
* This added flexibility does come with some complications though:
143
* - In the interest of being a lightweight library, libusb does not create
144
* threads and can only operate when your application is calling into it. Your
145
* application must call into libusb from it's main loop when events are ready
146
* to be handled, or you must use some other scheme to allow libusb to
147
* undertake whatever work needs to be done.
148
* - libusb also needs to be called into at certain fixed points in time in
149
* order to accurately handle transfer timeouts.
150
* - Memory handling becomes more complex. You cannot use stack memory unless
151
* the function with that stack is guaranteed not to return until the transfer
152
* callback has finished executing.
153
* - You generally lose some linearity from your code flow because submitting
154
* the transfer request is done in a separate function from where the transfer
155
* results are handled. This becomes particularly obvious when you want to
156
* submit a second transfer based on the results of an earlier transfer.
158
* Internally, libusb's synchronous interface is expressed in terms of function
159
* calls to the asynchronous interface.
161
* For details on how to use the asynchronous API, see the
162
* \ref asyncio "asynchronous I/O API" documentation pages.
167
* \page packetoverflow Packets and overflows
169
* \section packets Packet abstraction
171
* The USB specifications describe how data is transmitted in packets, with
172
* constraints on packet size defined by endpoint descriptors. The host must
173
* not send data payloads larger than the endpoint's maximum packet size.
175
* libusb and the underlying OS abstract out the packet concept, allowing you
176
* to request transfers of any size. Internally, the request will be divided
177
* up into correctly-sized packets. You do not have to be concerned with
178
* packet sizes, but there is one exception when considering overflows.
180
* \section overflow Bulk/interrupt transfer overflows
182
* When requesting data on a bulk endpoint, libusb requires you to supply a
183
* buffer and the maximum number of bytes of data that libusb can put in that
184
* buffer. However, the size of the buffer is not communicated to the device -
185
* the device is just asked to send any amount of data.
187
* There is no problem if the device sends an amount of data that is less than
188
* or equal to the buffer size. libusb reports this condition to you through
189
* the \ref libusb_transfer::actual_length "libusb_transfer.actual_length"
192
* Problems may occur if the device attempts to send more data than can fit in
193
* the buffer. libusb reports LIBUSB_TRANSFER_OVERFLOW for this condition but
194
* other behaviour is largely undefined: actual_length may or may not be
195
* accurate, the chunk of data that can fit in the buffer (before overflow)
196
* may or may not have been transferred.
198
* Overflows are nasty, but can be avoided. Even though you were told to
199
* ignore packets above, think about the lower level details: each transfer is
200
* split into packets (typically small, with a maximum size of 512 bytes).
201
* Overflows can only happen if the final packet in an incoming data transfer
202
* is smaller than the actual packet that the device wants to transfer.
203
* Therefore, you will never see an overflow if your transfer buffer size is a
204
* multiple of the endpoint's packet size: the final packet will either
205
* fill up completely or will be only partially filled.
209
* @defgroup asyncio Asynchronous device I/O
211
* This page details libusb's asynchronous (non-blocking) API for USB device
212
* I/O. This interface is very powerful but is also quite complex - you will
213
* need to read this page carefully to understand the necessary considerations
214
* and issues surrounding use of this interface. Simplistic applications
215
* may wish to consider the \ref syncio "synchronous I/O API" instead.
217
* The asynchronous interface is built around the idea of separating transfer
218
* submission and handling of transfer completion (the synchronous model
219
* combines both of these into one). There may be a long delay between
220
* submission and completion, however the asynchronous submission function
221
* is non-blocking so will return control to your application during that
222
* potentially long delay.
224
* \section asyncabstraction Transfer abstraction
226
* For the asynchronous I/O, libusb implements the concept of a generic
227
* transfer entity for all types of I/O (control, bulk, interrupt,
228
* isochronous). The generic transfer object must be treated slightly
229
* differently depending on which type of I/O you are performing with it.
231
* This is represented by the public libusb_transfer structure type.
233
* \section asynctrf Asynchronous transfers
235
* We can view asynchronous I/O as a 5 step process:
236
* -# <b>Allocation</b>: allocate a libusb_transfer
237
* -# <b>Filling</b>: populate the libusb_transfer instance with information
238
* about the transfer you wish to perform
239
* -# <b>Submission</b>: ask libusb to submit the transfer
240
* -# <b>Completion handling</b>: examine transfer results in the
241
* libusb_transfer structure
242
* -# <b>Deallocation</b>: clean up resources
245
* \subsection asyncalloc Allocation
247
* This step involves allocating memory for a USB transfer. This is the
248
* generic transfer object mentioned above. At this stage, the transfer
249
* is "blank" with no details about what type of I/O it will be used for.
251
* Allocation is done with the libusb_alloc_transfer() function. You must use
252
* this function rather than allocating your own transfers.
254
* \subsection asyncfill Filling
256
* This step is where you take a previously allocated transfer and fill it
257
* with information to determine the message type and direction, data buffer,
258
* callback function, etc.
260
* You can either fill the required fields yourself or you can use the
261
* helper functions: libusb_fill_control_transfer(), libusb_fill_bulk_transfer()
262
* and libusb_fill_interrupt_transfer().
264
* \subsection asyncsubmit Submission
266
* When you have allocated a transfer and filled it, you can submit it using
267
* libusb_submit_transfer(). This function returns immediately but can be
268
* regarded as firing off the I/O request in the background.
270
* \subsection asynccomplete Completion handling
272
* After a transfer has been submitted, one of four things can happen to it:
274
* - The transfer completes (i.e. some data was transferred)
275
* - The transfer has a timeout and the timeout expires before all data is
277
* - The transfer fails due to an error
278
* - The transfer is cancelled
280
* Each of these will cause the user-specified transfer callback function to
281
* be invoked. It is up to the callback function to determine which of the
282
* above actually happened and to act accordingly.
284
* The user-specified callback is passed a pointer to the libusb_transfer
285
* structure which was used to setup and submit the transfer. At completion
286
* time, libusb has populated this structure with results of the transfer:
287
* success or failure reason, number of bytes of data transferred, etc. See
288
* the libusb_transfer structure documentation for more information.
290
* \subsection Deallocation
292
* When a transfer has completed (i.e. the callback function has been invoked),
293
* you are advised to free the transfer (unless you wish to resubmit it, see
294
* below). Transfers are deallocated with libusb_free_transfer().
296
* It is undefined behaviour to free a transfer which has not completed.
298
* \section asyncresubmit Resubmission
300
* You may be wondering why allocation, filling, and submission are all
301
* separated above where they could reasonably be combined into a single
304
* The reason for separation is to allow you to resubmit transfers without
305
* having to allocate new ones every time. This is especially useful for
306
* common situations dealing with interrupt endpoints - you allocate one
307
* transfer, fill and submit it, and when it returns with results you just
308
* resubmit it for the next interrupt.
310
* \section asynccancel Cancellation
312
* Another advantage of using the asynchronous interface is that you have
313
* the ability to cancel transfers which have not yet completed. This is
314
* done by calling the libusb_cancel_transfer() function.
316
* libusb_cancel_transfer() is asynchronous/non-blocking in itself. When the
317
* cancellation actually completes, the transfer's callback function will
318
* be invoked, and the callback function should check the transfer status to
319
* determine that it was cancelled.
321
* Freeing the transfer after it has been cancelled but before cancellation
322
* has completed will result in undefined behaviour.
324
* When a transfer is cancelled, some of the data may have been transferred.
325
* libusb will communicate this to you in the transfer callback. Do not assume
326
* that no data was transferred.
328
* \section bulk_overflows Overflows on device-to-host bulk/interrupt endpoints
330
* If your device does not have predictable transfer sizes (or it misbehaves),
331
* your application may submit a request for data on an IN endpoint which is
332
* smaller than the data that the device wishes to send. In some circumstances
333
* this will cause an overflow, which is a nasty condition to deal with. See
334
* the \ref packetoverflow page for discussion.
336
* \section asyncctrl Considerations for control transfers
338
* The <tt>libusb_transfer</tt> structure is generic and hence does not
339
* include specific fields for the control-specific setup packet structure.
341
* In order to perform a control transfer, you must place the 8-byte setup
342
* packet at the start of the data buffer. To simplify this, you could
343
* cast the buffer pointer to type struct libusb_control_setup, or you can
344
* use the helper function libusb_fill_control_setup().
346
* The wLength field placed in the setup packet must be the length you would
347
* expect to be sent in the setup packet: the length of the payload that
348
* follows (or the expected maximum number of bytes to receive). However,
349
* the length field of the libusb_transfer object must be the length of
350
* the data buffer - i.e. it should be wLength <em>plus</em> the size of
351
* the setup packet (LIBUSB_CONTROL_SETUP_SIZE).
353
* If you use the helper functions, this is simplified for you:
354
* -# Allocate a buffer of size LIBUSB_CONTROL_SETUP_SIZE plus the size of the
355
* data you are sending/requesting.
356
* -# Call libusb_fill_control_setup() on the data buffer, using the transfer
357
* request size as the wLength value (i.e. do not include the extra space you
358
* allocated for the control setup).
359
* -# If this is a host-to-device transfer, place the data to be transferred
360
* in the data buffer, starting at offset LIBUSB_CONTROL_SETUP_SIZE.
361
* -# Call libusb_fill_control_transfer() to associate the data buffer with
362
* the transfer (and to set the remaining details such as callback and timeout).
363
* - Note that there is no parameter to set the length field of the transfer.
364
* The length is automatically inferred from the wLength field of the setup
366
* -# Submit the transfer.
368
* The multi-byte control setup fields (wValue, wIndex and wLength) must
369
* be given in little-endian byte order (the endianness of the USB bus).
370
* Endianness conversion is transparently handled by
371
* libusb_fill_control_setup() which is documented to accept host-endian
374
* Further considerations are needed when handling transfer completion in
375
* your callback function:
376
* - As you might expect, the setup packet will still be sitting at the start
377
* of the data buffer.
378
* - If this was a device-to-host transfer, the received data will be sitting
379
* at offset LIBUSB_CONTROL_SETUP_SIZE into the buffer.
380
* - The actual_length field of the transfer structure is relative to the
381
* wLength of the setup packet, rather than the size of the data buffer. So,
382
* if your wLength was 4, your transfer's <tt>length</tt> was 12, then you
383
* should expect an <tt>actual_length</tt> of 4 to indicate that the data was
384
* transferred in entirity.
386
* To simplify parsing of setup packets and obtaining the data from the
387
* correct offset, you may wish to use the libusb_control_transfer_get_data()
388
* and libusb_control_transfer_get_setup() functions within your transfer
391
* Even though control endpoints do not halt, a completed control transfer
392
* may have a LIBUSB_TRANSFER_STALL status code. This indicates the control
393
* request was not supported.
395
* \section asyncintr Considerations for interrupt transfers
397
* All interrupt transfers are performed using the polling interval presented
398
* by the bInterval value of the endpoint descriptor.
400
* \section asynciso Considerations for isochronous transfers
402
* Isochronous transfers are more complicated than transfers to
403
* non-isochronous endpoints.
405
* To perform I/O to an isochronous endpoint, allocate the transfer by calling
406
* libusb_alloc_transfer() with an appropriate number of isochronous packets.
408
* During filling, set \ref libusb_transfer::type "type" to
409
* \ref libusb_transfer_type::LIBUSB_TRANSFER_TYPE_ISOCHRONOUS
410
* "LIBUSB_TRANSFER_TYPE_ISOCHRONOUS", and set
411
* \ref libusb_transfer::num_iso_packets "num_iso_packets" to a value less than
412
* or equal to the number of packets you requested during allocation.
413
* libusb_alloc_transfer() does not set either of these fields for you, given
414
* that you might not even use the transfer on an isochronous endpoint.
416
* Next, populate the length field for the first num_iso_packets entries in
417
* the \ref libusb_transfer::iso_packet_desc "iso_packet_desc" array. Section
418
* 5.6.3 of the USB2 specifications describe how the maximum isochronous
419
* packet length is determined by the wMaxPacketSize field in the endpoint
421
* Two functions can help you here:
423
* - libusb_get_max_iso_packet_size() is an easy way to determine the max
424
* packet size for an isochronous endpoint. Note that the maximum packet
425
* size is actually the maximum number of bytes that can be transmitted in
426
* a single microframe, therefore this function multiplies the maximum number
427
* of bytes per transaction by the number of transaction opportunities per
429
* - libusb_set_iso_packet_lengths() assigns the same length to all packets
430
* within a transfer, which is usually what you want.
432
* For outgoing transfers, you'll obviously fill the buffer and populate the
433
* packet descriptors in hope that all the data gets transferred. For incoming
434
* transfers, you must ensure the buffer has sufficient capacity for
435
* the situation where all packets transfer the full amount of requested data.
437
* Completion handling requires some extra consideration. The
438
* \ref libusb_transfer::actual_length "actual_length" field of the transfer
439
* is meaningless and should not be examined; instead you must refer to the
440
* \ref libusb_iso_packet_descriptor::actual_length "actual_length" field of
441
* each individual packet.
443
* The \ref libusb_transfer::status "status" field of the transfer is also a
445
* - If the packets were submitted and the isochronous data microframes
446
* completed normally, status will have value
447
* \ref libusb_transfer_status::LIBUSB_TRANSFER_COMPLETED
448
* "LIBUSB_TRANSFER_COMPLETED". Note that bus errors and software-incurred
449
* delays are not counted as transfer errors; the transfer.status field may
450
* indicate COMPLETED even if some or all of the packets failed. Refer to
451
* the \ref libusb_iso_packet_descriptor::status "status" field of each
452
* individual packet to determine packet failures.
453
* - The status field will have value
454
* \ref libusb_transfer_status::LIBUSB_TRANSFER_ERROR
455
* "LIBUSB_TRANSFER_ERROR" only when serious errors were encountered.
456
* - Other transfer status codes occur with normal behaviour.
458
* The data for each packet will be found at an offset into the buffer that
459
* can be calculated as if each prior packet completed in full. The
460
* libusb_get_iso_packet_buffer() and libusb_get_iso_packet_buffer_simple()
461
* functions may help you here.
463
* \section asyncmem Memory caveats
465
* In most circumstances, it is not safe to use stack memory for transfer
466
* buffers. This is because the function that fired off the asynchronous
467
* transfer may return before libusb has finished using the buffer, and when
468
* the function returns it's stack gets destroyed. This is true for both
469
* host-to-device and device-to-host transfers.
471
* The only case in which it is safe to use stack memory is where you can
472
* guarantee that the function owning the stack space for the buffer does not
473
* return until after the transfer's callback function has completed. In every
474
* other case, you need to use heap memory instead.
476
* \section asyncflags Fine control
478
* Through using this asynchronous interface, you may find yourself repeating
479
* a few simple operations many times. You can apply a bitwise OR of certain
480
* flags to a transfer to simplify certain things:
481
* - \ref libusb_transfer_flags::LIBUSB_TRANSFER_SHORT_NOT_OK
482
* "LIBUSB_TRANSFER_SHORT_NOT_OK" results in transfers which transferred
483
* less than the requested amount of data being marked with status
484
* \ref libusb_transfer_status::LIBUSB_TRANSFER_ERROR "LIBUSB_TRANSFER_ERROR"
485
* (they would normally be regarded as COMPLETED)
486
* - \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_BUFFER
487
* "LIBUSB_TRANSFER_FREE_BUFFER" allows you to ask libusb to free the transfer
488
* buffer when freeing the transfer.
489
* - \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_TRANSFER
490
* "LIBUSB_TRANSFER_FREE_TRANSFER" causes libusb to automatically free the
491
* transfer after the transfer callback returns.
493
* \section asyncevent Event handling
495
* In accordance of the aim of being a lightweight library, libusb does not
496
* create threads internally. This means that libusb code does not execute
497
* at any time other than when your application is calling a libusb function.
498
* However, an asynchronous model requires that libusb perform work at various
499
* points in time - namely processing the results of previously-submitted
500
* transfers and invoking the user-supplied callback function.
502
* This gives rise to the libusb_handle_events() function which your
503
* application must call into when libusb has work do to. This gives libusb
504
* the opportunity to reap pending transfers, invoke callbacks, etc.
506
* The first issue to discuss here is how your application can figure out
507
* when libusb has work to do. In fact, there are two naive options which
508
* do not actually require your application to know this:
509
* -# Periodically call libusb_handle_events() in non-blocking mode at fixed
510
* short intervals from your main loop
511
* -# Repeatedly call libusb_handle_events() in blocking mode from a dedicated
514
* The first option is plainly not very nice, and will cause unnecessary
515
* CPU wakeups leading to increased power usage and decreased battery life.
516
* The second option is not very nice either, but may be the nicest option
517
* available to you if the "proper" approach can not be applied to your
518
* application (read on...).
520
* The recommended option is to integrate libusb with your application main
521
* event loop. libusb exposes a set of file descriptors which allow you to do
522
* this. Your main loop is probably already calling poll() or select() or a
523
* variant on a set of file descriptors for other event sources (e.g. keyboard
524
* button presses, mouse movements, network sockets, etc). You then add
525
* libusb's file descriptors to your poll()/select() calls, and when activity
526
* is detected on such descriptors you know it is time to call
527
* libusb_handle_events().
529
* There is one final event handling complication. libusb supports
530
* asynchronous transfers which time out after a specified time period, and
531
* this requires that libusb is called into at or after the timeout so that
532
* the timeout can be handled. So, in addition to considering libusb's file
533
* descriptors in your main event loop, you must also consider that libusb
534
* sometimes needs to be called into at fixed points in time even when there
535
* is no file descriptor activity.
537
* For the details on retrieving the set of file descriptors and determining
538
* the next timeout, see the \ref poll "polling and timing" API documentation.
542
* @defgroup poll Polling and timing
544
* This page documents libusb's functions for polling events and timing.
545
* These functions are only necessary for users of the
546
* \ref asyncio "asynchronous API". If you are only using the simpler
547
* \ref syncio "synchronous API" then you do not need to ever call these
550
* The justification for the functionality described here has already been
551
* discussed in the \ref asyncevent "event handling" section of the
552
* asynchronous API documentation. In summary, libusb does not create internal
553
* threads for event processing and hence relies on your application calling
554
* into libusb at certain points in time so that pending events can be handled.
555
* In order to know precisely when libusb needs to be called into, libusb
556
* offers you a set of pollable file descriptors and information about when
557
* the next timeout expires.
559
* If you are using the asynchronous I/O API, you must take one of the two
560
* following options, otherwise your I/O will not complete.
562
* \section pollsimple The simple option
564
* If your application revolves solely around libusb and does not need to
565
* handle other event sources, you can have a program structure as follows:
568
// find and open device
569
// maybe fire off some initial async I/O
571
while (user_has_not_requested_exit)
572
libusb_handle_events(ctx);
577
* With such a simple main loop, you do not have to worry about managing
578
* sets of file descriptors or handling timeouts. libusb_handle_events() will
579
* handle those details internally.
581
* \section pollmain The more advanced option
583
* In more advanced applications, you will already have a main loop which
584
* is monitoring other event sources: network sockets, X11 events, mouse
585
* movements, etc. Through exposing a set of file descriptors, libusb is
586
* designed to cleanly integrate into such main loops.
588
* In addition to polling file descriptors for the other event sources, you
589
* take a set of file descriptors from libusb and monitor those too. When you
590
* detect activity on libusb's file descriptors, you call
591
* libusb_handle_events_timeout() in non-blocking mode.
593
* What's more, libusb may also need to handle events at specific moments in
594
* time. No file descriptor activity is generated at these times, so your
595
* own application needs to be continually aware of when the next one of these
596
* moments occurs (through calling libusb_get_next_timeout()), and then it
597
* needs to call libusb_handle_events_timeout() in non-blocking mode when
598
* these moments occur. This means that you need to adjust your
599
* poll()/select() timeout accordingly.
601
* In pseudo-code, you want something that looks like:
605
libusb_get_pollfds(ctx)
606
while (user has not requested application exit) {
607
libusb_get_next_timeout(ctx);
608
poll(on libusb file descriptors plus any other event sources of interest,
609
using a timeout no larger than the value libusb just suggested)
610
if (poll() indicated activity on libusb file descriptors)
611
libusb_handle_events_timeout(ctx, 0);
612
if (time has elapsed to or beyond the libusb timeout)
613
libusb_handle_events_timeout(ctx, 0);
614
// handle events from other sources here
620
* \subsection polltime Notes on time-based events
622
* The above complication with having to track time and call into libusb at
623
* specific moments is a bit of a headache. For maximum compatibility, you do
624
* need to write your main loop as above, but you may decide that you can
625
* restrict the supported platforms of your application and get away with
626
* a more simplistic scheme.
628
* These time-based event complications are \b not required on the following
631
* - Linux, provided that the following version requirements are satisfied:
632
* - Linux v2.6.27 or newer, compiled with timerfd support
633
* - glibc v2.9 or newer
634
* - libusb v1.0.5 or newer
636
* Under these configurations, libusb_get_next_timeout() will \em always return
637
* 0, so your main loop can be simplified to:
641
libusb_get_pollfds(ctx)
642
while (user has not requested application exit) {
643
poll(on libusb file descriptors plus any other event sources of interest,
644
using any timeout that you like)
645
if (poll() indicated activity on libusb file descriptors)
646
libusb_handle_events_timeout(ctx, 0);
647
// handle events from other sources here
653
* Do remember that if you simplify your main loop to the above, you will
654
* lose compatibility with some platforms (including legacy Linux platforms,
655
* and <em>any future platforms supported by libusb which may have time-based
656
* event requirements</em>). The resultant problems will likely appear as
657
* strange bugs in your application.
659
* You can use the libusb_pollfds_handle_timeouts() function to do a runtime
660
* check to see if it is safe to ignore the time-based event complications.
661
* If your application has taken the shortcut of ignoring libusb's next timeout
662
* in your main loop, then you are advised to check the return value of
663
* libusb_pollfds_handle_timeouts() during application startup, and to abort
664
* if the platform does suffer from these timing complications.
666
* \subsection fdsetchange Changes in the file descriptor set
668
* The set of file descriptors that libusb uses as event sources may change
669
* during the life of your application. Rather than having to repeatedly
670
* call libusb_get_pollfds(), you can set up notification functions for when
671
* the file descriptor set changes using libusb_set_pollfd_notifiers().
673
* \subsection mtissues Multi-threaded considerations
675
* Unfortunately, the situation is complicated further when multiple threads
676
* come into play. If two threads are monitoring the same file descriptors,
677
* the fact that only one thread will be woken up when an event occurs causes
680
* The events lock, event waiters lock, and libusb_handle_events_locked()
681
* entities are added to solve these problems. You do not need to be concerned
682
* with these entities otherwise.
684
* See the extra documentation: \ref mtasync
687
/** \page mtasync Multi-threaded applications and asynchronous I/O
689
* libusb is a thread-safe library, but extra considerations must be applied
690
* to applications which interact with libusb from multiple threads.
692
* The underlying issue that must be addressed is that all libusb I/O
693
* revolves around monitoring file descriptors through the poll()/select()
694
* system calls. This is directly exposed at the
695
* \ref asyncio "asynchronous interface" but it is important to note that the
696
* \ref syncio "synchronous interface" is implemented on top of the
697
* asynchonrous interface, therefore the same considerations apply.
699
* The issue is that if two or more threads are concurrently calling poll()
700
* or select() on libusb's file descriptors then only one of those threads
701
* will be woken up when an event arrives. The others will be completely
702
* oblivious that anything has happened.
704
* Consider the following pseudo-code, which submits an asynchronous transfer
705
* then waits for its completion. This style is one way you could implement a
706
* synchronous interface on top of the asynchronous interface (and libusb
707
* does something similar, albeit more advanced due to the complications
708
* explained on this page).
711
void cb(struct libusb_transfer *transfer)
713
int *completed = transfer->user_data;
718
struct libusb_transfer *transfer;
719
unsigned char buffer[LIBUSB_CONTROL_SETUP_SIZE];
722
transfer = libusb_alloc_transfer(0);
723
libusb_fill_control_setup(buffer,
724
LIBUSB_REQUEST_TYPE_VENDOR | LIBUSB_ENDPOINT_OUT, 0x04, 0x01, 0, 0);
725
libusb_fill_control_transfer(transfer, dev, buffer, cb, &completed, 1000);
726
libusb_submit_transfer(transfer);
729
poll(libusb file descriptors, 120*1000);
730
if (poll indicates activity)
731
libusb_handle_events_timeout(ctx, 0);
733
printf("completed!");
738
* Here we are <em>serializing</em> completion of an asynchronous event
739
* against a condition - the condition being completion of a specific transfer.
740
* The poll() loop has a long timeout to minimize CPU usage during situations
741
* when nothing is happening (it could reasonably be unlimited).
743
* If this is the only thread that is polling libusb's file descriptors, there
744
* is no problem: there is no danger that another thread will swallow up the
745
* event that we are interested in. On the other hand, if there is another
746
* thread polling the same descriptors, there is a chance that it will receive
747
* the event that we were interested in. In this situation, <tt>myfunc()</tt>
748
* will only realise that the transfer has completed on the next iteration of
749
* the loop, <em>up to 120 seconds later.</em> Clearly a two-minute delay is
750
* undesirable, and don't even think about using short timeouts to circumvent
753
* The solution here is to ensure that no two threads are ever polling the
754
* file descriptors at the same time. A naive implementation of this would
755
* impact the capabilities of the library, so libusb offers the scheme
756
* documented below to ensure no loss of functionality.
758
* Before we go any further, it is worth mentioning that all libusb-wrapped
759
* event handling procedures fully adhere to the scheme documented below.
760
* This includes libusb_handle_events() and all the synchronous I/O functions -
761
* libusb hides this headache from you. You do not need to worry about any
762
* of these issues if you stick to that level.
764
* The problem is when we consider the fact that libusb exposes file
765
* descriptors to allow for you to integrate asynchronous USB I/O into
766
* existing main loops, effectively allowing you to do some work behind
767
* libusb's back. If you do take libusb's file descriptors and pass them to
768
* poll()/select() yourself, you need to be aware of the associated issues.
770
* \section eventlock The events lock
772
* The first concept to be introduced is the events lock. The events lock
773
* is used to serialize threads that want to handle events, such that only
774
* one thread is handling events at any one time.
776
* You must take the events lock before polling libusb file descriptors,
777
* using libusb_lock_events(). You must release the lock as soon as you have
778
* aborted your poll()/select() loop, using libusb_unlock_events().
780
* \section threadwait Letting other threads do the work for you
782
* Although the events lock is a critical part of the solution, it is not
783
* enough on it's own. You might wonder if the following is sufficient...
785
libusb_lock_events(ctx);
787
poll(libusb file descriptors, 120*1000);
788
if (poll indicates activity)
789
libusb_handle_events_timeout(ctx, 0);
791
libusb_unlock_events(ctx);
793
* ...and the answer is that it is not. This is because the transfer in the
794
* code shown above may take a long time (say 30 seconds) to complete, and
795
* the lock is not released until the transfer is completed.
797
* Another thread with similar code that wants to do event handling may be
798
* working with a transfer that completes after a few milliseconds. Despite
799
* having such a quick completion time, the other thread cannot check that
800
* status of its transfer until the code above has finished (30 seconds later)
801
* due to contention on the lock.
803
* To solve this, libusb offers you a mechanism to determine when another
804
* thread is handling events. It also offers a mechanism to block your thread
805
* until the event handling thread has completed an event (and this mechanism
806
* does not involve polling of file descriptors).
808
* After determining that another thread is currently handling events, you
809
* obtain the <em>event waiters</em> lock using libusb_lock_event_waiters().
810
* You then re-check that some other thread is still handling events, and if
811
* so, you call libusb_wait_for_event().
813
* libusb_wait_for_event() puts your application to sleep until an event
814
* occurs, or until a thread releases the events lock. When either of these
815
* things happen, your thread is woken up, and should re-check the condition
816
* it was waiting on. It should also re-check that another thread is handling
817
* events, and if not, it should start handling events itself.
819
* This looks like the following, as pseudo-code:
822
if (libusb_try_lock_events(ctx) == 0) {
823
// we obtained the event lock: do our own event handling
825
if (!libusb_event_handling_ok(ctx)) {
826
libusb_unlock_events(ctx);
829
poll(libusb file descriptors, 120*1000);
830
if (poll indicates activity)
831
libusb_handle_events_locked(ctx, 0);
833
libusb_unlock_events(ctx);
835
// another thread is doing event handling. wait for it to signal us that
836
// an event has completed
837
libusb_lock_event_waiters(ctx);
840
// now that we have the event waiters lock, double check that another
841
// thread is still handling events for us. (it may have ceased handling
842
// events in the time it took us to reach this point)
843
if (!libusb_event_handler_active(ctx)) {
844
// whoever was handling events is no longer doing so, try again
845
libusb_unlock_event_waiters(ctx);
849
libusb_wait_for_event(ctx);
851
libusb_unlock_event_waiters(ctx);
853
printf("completed!\n");
856
* A naive look at the above code may suggest that this can only support
857
* one event waiter (hence a total of 2 competing threads, the other doing
858
* event handling), because the event waiter seems to have taken the event
859
* waiters lock while waiting for an event. However, the system does support
860
* multiple event waiters, because libusb_wait_for_event() actually drops
861
* the lock while waiting, and reaquires it before continuing.
863
* We have now implemented code which can dynamically handle situations where
864
* nobody is handling events (so we should do it ourselves), and it can also
865
* handle situations where another thread is doing event handling (so we can
866
* piggyback onto them). It is also equipped to handle a combination of
867
* the two, for example, another thread is doing event handling, but for
868
* whatever reason it stops doing so before our condition is met, so we take
869
* over the event handling.
871
* Four functions were introduced in the above pseudo-code. Their importance
872
* should be apparent from the code shown above.
873
* -# libusb_try_lock_events() is a non-blocking function which attempts
874
* to acquire the events lock but returns a failure code if it is contended.
875
* -# libusb_event_handling_ok() checks that libusb is still happy for your
876
* thread to be performing event handling. Sometimes, libusb needs to
877
* interrupt the event handler, and this is how you can check if you have
878
* been interrupted. If this function returns 0, the correct behaviour is
879
* for you to give up the event handling lock, and then to repeat the cycle.
880
* The following libusb_try_lock_events() will fail, so you will become an
881
* events waiter. For more information on this, read \ref fullstory below.
882
* -# libusb_handle_events_locked() is a variant of
883
* libusb_handle_events_timeout() that you can call while holding the
884
* events lock. libusb_handle_events_timeout() itself implements similar
885
* logic to the above, so be sure not to call it when you are
886
* "working behind libusb's back", as is the case here.
887
* -# libusb_event_handler_active() determines if someone is currently
888
* holding the events lock
890
* You might be wondering why there is no function to wake up all threads
891
* blocked on libusb_wait_for_event(). This is because libusb can do this
892
* internally: it will wake up all such threads when someone calls
893
* libusb_unlock_events() or when a transfer completes (at the point after its
894
* callback has returned).
896
* \subsection fullstory The full story
898
* The above explanation should be enough to get you going, but if you're
899
* really thinking through the issues then you may be left with some more
900
* questions regarding libusb's internals. If you're curious, read on, and if
901
* not, skip to the next section to avoid confusing yourself!
903
* The immediate question that may spring to mind is: what if one thread
904
* modifies the set of file descriptors that need to be polled while another
905
* thread is doing event handling?
907
* There are 2 situations in which this may happen.
908
* -# libusb_open() will add another file descriptor to the poll set,
909
* therefore it is desirable to interrupt the event handler so that it
910
* restarts, picking up the new descriptor.
911
* -# libusb_close() will remove a file descriptor from the poll set. There
912
* are all kinds of race conditions that could arise here, so it is
913
* important that nobody is doing event handling at this time.
915
* libusb handles these issues internally, so application developers do not
916
* have to stop their event handlers while opening/closing devices. Here's how
917
* it works, focusing on the libusb_close() situation first:
919
* -# During initialization, libusb opens an internal pipe, and it adds the read
920
* end of this pipe to the set of file descriptors to be polled.
921
* -# During libusb_close(), libusb writes some dummy data on this control pipe.
922
* This immediately interrupts the event handler. libusb also records
923
* internally that it is trying to interrupt event handlers for this
924
* high-priority event.
925
* -# At this point, some of the functions described above start behaving
927
* - libusb_event_handling_ok() starts returning 1, indicating that it is NOT
928
* OK for event handling to continue.
929
* - libusb_try_lock_events() starts returning 1, indicating that another
930
* thread holds the event handling lock, even if the lock is uncontended.
931
* - libusb_event_handler_active() starts returning 1, indicating that
932
* another thread is doing event handling, even if that is not true.
933
* -# The above changes in behaviour result in the event handler stopping and
934
* giving up the events lock very quickly, giving the high-priority
935
* libusb_close() operation a "free ride" to acquire the events lock. All
936
* threads that are competing to do event handling become event waiters.
937
* -# With the events lock held inside libusb_close(), libusb can safely remove
938
* a file descriptor from the poll set, in the safety of knowledge that
939
* nobody is polling those descriptors or trying to access the poll set.
940
* -# After obtaining the events lock, the close operation completes very
941
* quickly (usually a matter of milliseconds) and then immediately releases
943
* -# At the same time, the behaviour of libusb_event_handling_ok() and friends
944
* reverts to the original, documented behaviour.
945
* -# The release of the events lock causes the threads that are waiting for
946
* events to be woken up and to start competing to become event handlers
947
* again. One of them will succeed; it will then re-obtain the list of poll
948
* descriptors, and USB I/O will then continue as normal.
950
* libusb_open() is similar, and is actually a more simplistic case. Upon a
951
* call to libusb_open():
953
* -# The device is opened and a file descriptor is added to the poll set.
954
* -# libusb sends some dummy data on the control pipe, and records that it
955
* is trying to modify the poll descriptor set.
956
* -# The event handler is interrupted, and the same behaviour change as for
957
* libusb_close() takes effect, causing all event handling threads to become
959
* -# The libusb_open() implementation takes its free ride to the events lock.
960
* -# Happy that it has successfully paused the events handler, libusb_open()
961
* releases the events lock.
962
* -# The event waiter threads are all woken up and compete to become event
963
* handlers again. The one that succeeds will obtain the list of poll
964
* descriptors again, which will include the addition of the new device.
966
* \subsection concl Closing remarks
968
* The above may seem a little complicated, but hopefully I have made it clear
969
* why such complications are necessary. Also, do not forget that this only
970
* applies to applications that take libusb's file descriptors and integrate
971
* them into their own polling loops.
973
* You may decide that it is OK for your multi-threaded application to ignore
974
* some of the rules and locks detailed above, because you don't think that
975
* two threads can ever be polling the descriptors at the same time. If that
976
* is the case, then that's good news for you because you don't have to worry.
977
* But be careful here; remember that the synchronous I/O functions do event
978
* handling internally. If you have one thread doing event handling in a loop
979
* (without implementing the rules and locking semantics documented above)
980
* and another trying to send a synchronous USB transfer, you will end up with
981
* two threads monitoring the same descriptors, and the above-described
982
* undesirable behaviour occuring. The solution is for your polling thread to
983
* play by the rules; the synchronous I/O functions do so, and this will result
984
* in them getting along in perfect harmony.
986
* If you do have a dedicated thread doing event handling, it is perfectly
987
* legal for it to take the event handling lock for long periods of time. Any
988
* synchronous I/O functions you call from other threads will transparently
989
* fall back to the "event waiters" mechanism detailed above. The only
990
* consideration that your event handling thread must apply is the one related
991
* to libusb_event_handling_ok(): you must call this before every poll(), and
992
* give up the events lock if instructed.
995
int usbi_io_init(struct libusb_context *ctx)
999
usbi_mutex_init(&ctx->flying_transfers_lock, NULL);
1000
usbi_mutex_init(&ctx->pollfds_lock, NULL);
1001
usbi_mutex_init(&ctx->pollfd_modify_lock, NULL);
1002
usbi_mutex_init(&ctx->events_lock, NULL);
1003
usbi_mutex_init(&ctx->event_waiters_lock, NULL);
1004
usbi_cond_init(&ctx->event_waiters_cond, NULL);
1005
list_init(&ctx->flying_transfers);
1006
list_init(&ctx->pollfds);
1008
/* FIXME should use an eventfd on kernels that support it */
1009
p = usbi_pipe(ctx->ctrl_pipe);
1011
r = LIBUSB_ERROR_OTHER;
1015
r = usbi_add_pollfd(ctx, ctx->ctrl_pipe[0], POLLIN);
1019
#ifdef USBI_TIMERFD_AVAILABLE
1020
ctx->timerfd = timerfd_create(usbi_backend->get_timerfd_clockid(),
1022
if (ctx->timerfd >= 0) {
1023
usbi_dbg("using timerfd for timeouts");
1024
r = usbi_add_pollfd(ctx, ctx->timerfd, POLLIN);
1028
usbi_dbg("timerfd not available (code %d error %d)", ctx->timerfd, errno);
1036
#ifdef USBI_TIMERFD_AVAILABLE
1037
if (ctx->timerfd != -1)
1038
close(ctx->timerfd);
1041
usbi_remove_pollfd(ctx, ctx->ctrl_pipe[0]);
1042
usbi_close(ctx->ctrl_pipe[0]);
1043
usbi_close(ctx->ctrl_pipe[1]);
1045
usbi_mutex_destroy(&ctx->flying_transfers_lock);
1046
usbi_mutex_destroy(&ctx->pollfds_lock);
1047
usbi_mutex_destroy(&ctx->pollfd_modify_lock);
1048
usbi_mutex_destroy(&ctx->events_lock);
1049
usbi_mutex_destroy(&ctx->event_waiters_lock);
1050
usbi_cond_destroy(&ctx->event_waiters_cond);
1054
void usbi_io_exit(struct libusb_context *ctx)
1056
usbi_remove_pollfd(ctx, ctx->ctrl_pipe[0]);
1057
usbi_close(ctx->ctrl_pipe[0]);
1058
usbi_close(ctx->ctrl_pipe[1]);
1059
#ifdef USBI_TIMERFD_AVAILABLE
1060
if (usbi_using_timerfd(ctx)) {
1061
usbi_remove_pollfd(ctx, ctx->timerfd);
1062
close(ctx->timerfd);
1065
usbi_mutex_destroy(&ctx->flying_transfers_lock);
1066
usbi_mutex_destroy(&ctx->pollfds_lock);
1067
usbi_mutex_destroy(&ctx->pollfd_modify_lock);
1068
usbi_mutex_destroy(&ctx->events_lock);
1069
usbi_mutex_destroy(&ctx->event_waiters_lock);
1070
usbi_cond_destroy(&ctx->event_waiters_cond);
1073
static int calculate_timeout(struct usbi_transfer *transfer)
1076
struct timespec current_time;
1077
unsigned int timeout =
1078
__USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout;
1083
r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, ¤t_time);
1085
usbi_err(ITRANSFER_CTX(transfer),
1086
"failed to read monotonic clock, errno=%d", errno);
1090
current_time.tv_sec += timeout / 1000;
1091
current_time.tv_nsec += (timeout % 1000) * 1000000;
1093
if (current_time.tv_nsec > 1000000000) {
1094
current_time.tv_nsec -= 1000000000;
1095
current_time.tv_sec++;
1098
TIMESPEC_TO_TIMEVAL(&transfer->timeout, ¤t_time);
1102
/* add a transfer to the (timeout-sorted) active transfers list.
1103
* returns 1 if the transfer has a timeout and it is the timeout next to
1105
static int add_to_flying_list(struct usbi_transfer *transfer)
1107
struct usbi_transfer *cur;
1108
struct timeval *timeout = &transfer->timeout;
1109
struct libusb_context *ctx = ITRANSFER_CTX(transfer);
1113
usbi_mutex_lock(&ctx->flying_transfers_lock);
1115
/* if we have no other flying transfers, start the list with this one */
1116
if (list_empty(&ctx->flying_transfers)) {
1117
list_add(&transfer->list, &ctx->flying_transfers);
1118
if (timerisset(timeout))
1123
/* if we have infinite timeout, append to end of list */
1124
if (!timerisset(timeout)) {
1125
list_add_tail(&transfer->list, &ctx->flying_transfers);
1129
/* otherwise, find appropriate place in list */
1130
list_for_each_entry(cur, &ctx->flying_transfers, list, struct usbi_transfer) {
1131
/* find first timeout that occurs after the transfer in question */
1132
struct timeval *cur_tv = &cur->timeout;
1134
if (!timerisset(cur_tv) || (cur_tv->tv_sec > timeout->tv_sec) ||
1135
(cur_tv->tv_sec == timeout->tv_sec &&
1136
cur_tv->tv_usec > timeout->tv_usec)) {
1137
list_add_tail(&transfer->list, &cur->list);
1144
/* otherwise we need to be inserted at the end */
1145
list_add_tail(&transfer->list, &ctx->flying_transfers);
1147
usbi_mutex_unlock(&ctx->flying_transfers_lock);
1151
/** \ingroup asyncio
1152
* Allocate a libusb transfer with a specified number of isochronous packet
1153
* descriptors. The returned transfer is pre-initialized for you. When the new
1154
* transfer is no longer needed, it should be freed with
1155
* libusb_free_transfer().
1157
* Transfers intended for non-isochronous endpoints (e.g. control, bulk,
1158
* interrupt) should specify an iso_packets count of zero.
1160
* For transfers intended for isochronous endpoints, specify an appropriate
1161
* number of packet descriptors to be allocated as part of the transfer.
1162
* The returned transfer is not specially initialized for isochronous I/O;
1163
* you are still required to set the
1164
* \ref libusb_transfer::num_iso_packets "num_iso_packets" and
1165
* \ref libusb_transfer::type "type" fields accordingly.
1167
* It is safe to allocate a transfer with some isochronous packets and then
1168
* use it on a non-isochronous endpoint. If you do this, ensure that at time
1169
* of submission, num_iso_packets is 0 and that type is set appropriately.
1171
* \param iso_packets number of isochronous packet descriptors to allocate
1172
* \returns a newly allocated transfer, or NULL on error
1174
API_EXPORTED struct libusb_transfer *libusb_alloc_transfer(int iso_packets)
1176
size_t os_alloc_size = usbi_backend->transfer_priv_size
1177
+ (usbi_backend->add_iso_packet_size * iso_packets);
1178
size_t alloc_size = sizeof(struct usbi_transfer)
1179
+ sizeof(struct libusb_transfer)
1180
+ (sizeof(struct libusb_iso_packet_descriptor) * iso_packets)
1182
struct usbi_transfer *itransfer = malloc(alloc_size);
1186
memset(itransfer, 0, alloc_size);
1187
itransfer->num_iso_packets = iso_packets;
1188
usbi_mutex_init(&itransfer->lock, NULL);
1189
return __USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1192
/** \ingroup asyncio
1193
* Free a transfer structure. This should be called for all transfers
1194
* allocated with libusb_alloc_transfer().
1196
* If the \ref libusb_transfer_flags::LIBUSB_TRANSFER_FREE_BUFFER
1197
* "LIBUSB_TRANSFER_FREE_BUFFER" flag is set and the transfer buffer is
1198
* non-NULL, this function will also free the transfer buffer using the
1199
* standard system memory allocator (e.g. free()).
1201
* It is legal to call this function with a NULL transfer. In this case,
1202
* the function will simply return safely.
1204
* It is not legal to free an active transfer (one which has been submitted
1205
* and has not yet completed).
1207
* \param transfer the transfer to free
1209
API_EXPORTED void libusb_free_transfer(struct libusb_transfer *transfer)
1211
struct usbi_transfer *itransfer;
1215
if (transfer->flags & LIBUSB_TRANSFER_FREE_BUFFER && transfer->buffer)
1216
free(transfer->buffer);
1218
itransfer = __LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1219
usbi_mutex_destroy(&itransfer->lock);
1223
/** \ingroup asyncio
1224
* Submit a transfer. This function will fire off the USB transfer and then
1225
* return immediately.
1227
* \param transfer the transfer to submit
1228
* \returns 0 on success
1229
* \returns LIBUSB_ERROR_NO_DEVICE if the device has been disconnected
1230
* \returns LIBUSB_ERROR_BUSY if the transfer has already been submitted.
1231
* \returns another LIBUSB_ERROR code on other failure
1233
API_EXPORTED int libusb_submit_transfer(struct libusb_transfer *transfer)
1235
struct libusb_context *ctx = TRANSFER_CTX(transfer);
1236
struct usbi_transfer *itransfer =
1237
__LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1241
usbi_mutex_lock(&itransfer->lock);
1242
itransfer->transferred = 0;
1243
itransfer->flags = 0;
1244
r = calculate_timeout(itransfer);
1246
r = LIBUSB_ERROR_OTHER;
1250
first = add_to_flying_list(itransfer);
1251
r = usbi_backend->submit_transfer(itransfer);
1253
usbi_mutex_lock(&ctx->flying_transfers_lock);
1254
list_del(&itransfer->list);
1255
usbi_mutex_unlock(&ctx->flying_transfers_lock);
1257
#ifdef USBI_TIMERFD_AVAILABLE
1258
else if (first && usbi_using_timerfd(ctx)) {
1259
/* if this transfer has the lowest timeout of all active transfers,
1260
* rearm the timerfd with this transfer's timeout */
1261
const struct itimerspec it = { {0, 0},
1262
{ itransfer->timeout.tv_sec, itransfer->timeout.tv_usec * 1000 } };
1263
usbi_dbg("arm timerfd for timeout in %dms (first in line)", transfer->timeout);
1264
r = timerfd_settime(ctx->timerfd, TFD_TIMER_ABSTIME, &it, NULL);
1266
r = LIBUSB_ERROR_OTHER;
1271
usbi_mutex_unlock(&itransfer->lock);
1275
/** \ingroup asyncio
1276
* Asynchronously cancel a previously submitted transfer.
1277
* This function returns immediately, but this does not indicate cancellation
1278
* is complete. Your callback function will be invoked at some later time
1279
* with a transfer status of
1280
* \ref libusb_transfer_status::LIBUSB_TRANSFER_CANCELLED
1281
* "LIBUSB_TRANSFER_CANCELLED."
1283
* \param transfer the transfer to cancel
1284
* \returns 0 on success
1285
* \returns LIBUSB_ERROR_NOT_FOUND if the transfer is already complete or
1287
* \returns a LIBUSB_ERROR code on failure
1289
API_EXPORTED int libusb_cancel_transfer(struct libusb_transfer *transfer)
1291
struct usbi_transfer *itransfer =
1292
__LIBUSB_TRANSFER_TO_USBI_TRANSFER(transfer);
1296
usbi_mutex_lock(&itransfer->lock);
1297
r = usbi_backend->cancel_transfer(itransfer);
1299
usbi_err(TRANSFER_CTX(transfer),
1300
"cancel transfer failed error %d", r);
1301
usbi_mutex_unlock(&itransfer->lock);
1305
#ifdef USBI_TIMERFD_AVAILABLE
1306
static int disarm_timerfd(struct libusb_context *ctx)
1308
const struct itimerspec disarm_timer = { { 0, 0 }, { 0, 0 } };
1312
r = timerfd_settime(ctx->timerfd, 0, &disarm_timer, NULL);
1314
return LIBUSB_ERROR_OTHER;
1319
/* iterates through the flying transfers, and rearms the timerfd based on the
1320
* next upcoming timeout.
1321
* must be called with flying_list locked.
1322
* returns 0 if there was no timeout to arm, 1 if the next timeout was armed,
1323
* or a LIBUSB_ERROR code on failure.
1325
static int arm_timerfd_for_next_timeout(struct libusb_context *ctx)
1327
struct usbi_transfer *transfer;
1329
list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
1330
struct timeval *cur_tv = &transfer->timeout;
1332
/* if we've reached transfers of infinite timeout, then we have no
1334
if (!timerisset(cur_tv))
1337
/* act on first transfer that is not already cancelled */
1338
if (!(transfer->flags & USBI_TRANSFER_TIMED_OUT)) {
1340
const struct itimerspec it = { {0, 0},
1341
{ cur_tv->tv_sec, cur_tv->tv_usec * 1000 } };
1342
usbi_dbg("next timeout originally %dms", __USBI_TRANSFER_TO_LIBUSB_TRANSFER(transfer)->timeout);
1343
r = timerfd_settime(ctx->timerfd, TFD_TIMER_ABSTIME, &it, NULL);
1345
return LIBUSB_ERROR_OTHER;
1353
static int disarm_timerfd(struct libusb_context *ctx)
1357
static int arm_timerfd_for_next_timeout(struct libusb_context *ctx)
1363
/* Handle completion of a transfer (completion might be an error condition).
1364
* This will invoke the user-supplied callback function, which may end up
1365
* freeing the transfer. Therefore you cannot use the transfer structure
1366
* after calling this function, and you should free all backend-specific
1367
* data before calling it.
1368
* Do not call this function with the usbi_transfer lock held. User-specified
1369
* callback functions may attempt to directly resubmit the transfer, which
1370
* will attempt to take the lock. */
1371
int usbi_handle_transfer_completion(struct usbi_transfer *itransfer,
1372
enum libusb_transfer_status status)
1374
struct libusb_transfer *transfer =
1375
__USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1376
struct libusb_context *ctx = TRANSFER_CTX(transfer);
1380
/* FIXME: could be more intelligent with the timerfd here. we don't need
1381
* to disarm the timerfd if there was no timer running, and we only need
1382
* to rearm the timerfd if the transfer that expired was the one with
1383
* the shortest timeout. */
1385
usbi_mutex_lock(&ctx->flying_transfers_lock);
1386
list_del(&itransfer->list);
1387
r = arm_timerfd_for_next_timeout(ctx);
1388
usbi_mutex_unlock(&ctx->flying_transfers_lock);
1392
} else if (r == 0) {
1393
r = disarm_timerfd(ctx);
1398
if (status == LIBUSB_TRANSFER_COMPLETED
1399
&& transfer->flags & LIBUSB_TRANSFER_SHORT_NOT_OK) {
1400
int rqlen = transfer->length;
1401
if (transfer->type == LIBUSB_TRANSFER_TYPE_CONTROL)
1402
rqlen -= LIBUSB_CONTROL_SETUP_SIZE;
1403
if (rqlen != itransfer->transferred) {
1404
usbi_dbg("interpreting short transfer as error");
1405
status = LIBUSB_TRANSFER_ERROR;
1409
flags = transfer->flags;
1410
transfer->status = status;
1411
transfer->actual_length = itransfer->transferred;
1412
if (transfer->callback)
1413
transfer->callback(transfer);
1414
/* transfer might have been freed by the above call, do not use from
1416
if (flags & LIBUSB_TRANSFER_FREE_TRANSFER)
1417
libusb_free_transfer(transfer);
1418
usbi_mutex_lock(&ctx->event_waiters_lock);
1419
usbi_cond_broadcast(&ctx->event_waiters_cond);
1420
usbi_mutex_unlock(&ctx->event_waiters_lock);
1424
/* Similar to usbi_handle_transfer_completion() but exclusively for transfers
1425
* that were asynchronously cancelled. The same concerns w.r.t. freeing of
1426
* transfers exist here.
1427
* Do not call this function with the usbi_transfer lock held. User-specified
1428
* callback functions may attempt to directly resubmit the transfer, which
1429
* will attempt to take the lock. */
1430
int usbi_handle_transfer_cancellation(struct usbi_transfer *transfer)
1432
/* if the URB was cancelled due to timeout, report timeout to the user */
1433
if (transfer->flags & USBI_TRANSFER_TIMED_OUT) {
1434
usbi_dbg("detected timeout cancellation");
1435
return usbi_handle_transfer_completion(transfer, LIBUSB_TRANSFER_TIMED_OUT);
1438
/* otherwise its a normal async cancel */
1439
return usbi_handle_transfer_completion(transfer, LIBUSB_TRANSFER_CANCELLED);
1443
* Attempt to acquire the event handling lock. This lock is used to ensure that
1444
* only one thread is monitoring libusb event sources at any one time.
1446
* You only need to use this lock if you are developing an application
1447
* which calls poll() or select() on libusb's file descriptors directly.
1448
* If you stick to libusb's event handling loop functions (e.g.
1449
* libusb_handle_events()) then you do not need to be concerned with this
1452
* While holding this lock, you are trusted to actually be handling events.
1453
* If you are no longer handling events, you must call libusb_unlock_events()
1454
* as soon as possible.
1456
* \param ctx the context to operate on, or NULL for the default context
1457
* \returns 0 if the lock was obtained successfully
1458
* \returns 1 if the lock was not obtained (i.e. another thread holds the lock)
1461
API_EXPORTED int libusb_try_lock_events(libusb_context *ctx)
1464
USBI_GET_CONTEXT(ctx);
1466
/* is someone else waiting to modify poll fds? if so, don't let this thread
1467
* start event handling */
1468
usbi_mutex_lock(&ctx->pollfd_modify_lock);
1469
r = ctx->pollfd_modify;
1470
usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1472
usbi_dbg("someone else is modifying poll fds");
1476
r = usbi_mutex_trylock(&ctx->events_lock);
1480
ctx->event_handler_active = 1;
1485
* Acquire the event handling lock, blocking until successful acquisition if
1486
* it is contended. This lock is used to ensure that only one thread is
1487
* monitoring libusb event sources at any one time.
1489
* You only need to use this lock if you are developing an application
1490
* which calls poll() or select() on libusb's file descriptors directly.
1491
* If you stick to libusb's event handling loop functions (e.g.
1492
* libusb_handle_events()) then you do not need to be concerned with this
1495
* While holding this lock, you are trusted to actually be handling events.
1496
* If you are no longer handling events, you must call libusb_unlock_events()
1497
* as soon as possible.
1499
* \param ctx the context to operate on, or NULL for the default context
1502
API_EXPORTED void libusb_lock_events(libusb_context *ctx)
1504
USBI_GET_CONTEXT(ctx);
1505
usbi_mutex_lock(&ctx->events_lock);
1506
ctx->event_handler_active = 1;
1510
* Release the lock previously acquired with libusb_try_lock_events() or
1511
* libusb_lock_events(). Releasing this lock will wake up any threads blocked
1512
* on libusb_wait_for_event().
1514
* \param ctx the context to operate on, or NULL for the default context
1517
API_EXPORTED void libusb_unlock_events(libusb_context *ctx)
1519
USBI_GET_CONTEXT(ctx);
1520
ctx->event_handler_active = 0;
1521
usbi_mutex_unlock(&ctx->events_lock);
1523
/* FIXME: perhaps we should be a bit more efficient by not broadcasting
1524
* the availability of the events lock when we are modifying pollfds
1525
* (check ctx->pollfd_modify)? */
1526
usbi_mutex_lock(&ctx->event_waiters_lock);
1527
usbi_cond_broadcast(&ctx->event_waiters_cond);
1528
usbi_mutex_unlock(&ctx->event_waiters_lock);
1532
* Determine if it is still OK for this thread to be doing event handling.
1534
* Sometimes, libusb needs to temporarily pause all event handlers, and this
1535
* is the function you should use before polling file descriptors to see if
1538
* If this function instructs your thread to give up the events lock, you
1539
* should just continue the usual logic that is documented in \ref mtasync.
1540
* On the next iteration, your thread will fail to obtain the events lock,
1541
* and will hence become an event waiter.
1543
* This function should be called while the events lock is held: you don't
1544
* need to worry about the results of this function if your thread is not
1545
* the current event handler.
1547
* \param ctx the context to operate on, or NULL for the default context
1548
* \returns 1 if event handling can start or continue
1549
* \returns 0 if this thread must give up the events lock
1550
* \see \ref fullstory "Multi-threaded I/O: the full story"
1552
API_EXPORTED int libusb_event_handling_ok(libusb_context *ctx)
1555
USBI_GET_CONTEXT(ctx);
1557
/* is someone else waiting to modify poll fds? if so, don't let this thread
1558
* continue event handling */
1559
usbi_mutex_lock(&ctx->pollfd_modify_lock);
1560
r = ctx->pollfd_modify;
1561
usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1563
usbi_dbg("someone else is modifying poll fds");
1572
* Determine if an active thread is handling events (i.e. if anyone is holding
1573
* the event handling lock).
1575
* \param ctx the context to operate on, or NULL for the default context
1576
* \returns 1 if a thread is handling events
1577
* \returns 0 if there are no threads currently handling events
1580
API_EXPORTED int libusb_event_handler_active(libusb_context *ctx)
1583
USBI_GET_CONTEXT(ctx);
1585
/* is someone else waiting to modify poll fds? if so, don't let this thread
1586
* start event handling -- indicate that event handling is happening */
1587
usbi_mutex_lock(&ctx->pollfd_modify_lock);
1588
r = ctx->pollfd_modify;
1589
usbi_mutex_unlock(&ctx->pollfd_modify_lock);
1591
usbi_dbg("someone else is modifying poll fds");
1595
return ctx->event_handler_active;
1599
* Acquire the event waiters lock. This lock is designed to be obtained under
1600
* the situation where you want to be aware when events are completed, but
1601
* some other thread is event handling so calling libusb_handle_events() is not
1604
* You then obtain this lock, re-check that another thread is still handling
1605
* events, then call libusb_wait_for_event().
1607
* You only need to use this lock if you are developing an application
1608
* which calls poll() or select() on libusb's file descriptors directly,
1609
* <b>and</b> may potentially be handling events from 2 threads simultaenously.
1610
* If you stick to libusb's event handling loop functions (e.g.
1611
* libusb_handle_events()) then you do not need to be concerned with this
1614
* \param ctx the context to operate on, or NULL for the default context
1617
API_EXPORTED void libusb_lock_event_waiters(libusb_context *ctx)
1619
USBI_GET_CONTEXT(ctx);
1620
usbi_mutex_lock(&ctx->event_waiters_lock);
1624
* Release the event waiters lock.
1625
* \param ctx the context to operate on, or NULL for the default context
1628
API_EXPORTED void libusb_unlock_event_waiters(libusb_context *ctx)
1630
USBI_GET_CONTEXT(ctx);
1631
usbi_mutex_unlock(&ctx->event_waiters_lock);
1635
* Wait for another thread to signal completion of an event. Must be called
1636
* with the event waiters lock held, see libusb_lock_event_waiters().
1638
* This function will block until any of the following conditions are met:
1639
* -# The timeout expires
1640
* -# A transfer completes
1641
* -# A thread releases the event handling lock through libusb_unlock_events()
1643
* Condition 1 is obvious. Condition 2 unblocks your thread <em>after</em>
1644
* the callback for the transfer has completed. Condition 3 is important
1645
* because it means that the thread that was previously handling events is no
1646
* longer doing so, so if any events are to complete, another thread needs to
1647
* step up and start event handling.
1649
* This function releases the event waiters lock before putting your thread
1650
* to sleep, and reacquires the lock as it is being woken up.
1652
* \param ctx the context to operate on, or NULL for the default context
1653
* \param tv maximum timeout for this blocking function. A NULL value
1654
* indicates unlimited timeout.
1655
* \returns 0 after a transfer completes or another thread stops event handling
1656
* \returns 1 if the timeout expired
1659
API_EXPORTED int libusb_wait_for_event(libusb_context *ctx, struct timeval *tv)
1661
struct timespec timeout;
1664
USBI_GET_CONTEXT(ctx);
1666
usbi_cond_wait(&ctx->event_waiters_cond, &ctx->event_waiters_lock);
1667
usbi_dbg("tv was NULL");
1671
r = usbi_backend->clock_gettime(USBI_CLOCK_REALTIME, &timeout);
1673
usbi_err(ctx, "failed to read realtime clock, error %d", errno);
1674
return LIBUSB_ERROR_OTHER;
1677
timeout.tv_sec += tv->tv_sec;
1678
timeout.tv_nsec += tv->tv_usec * 1000;
1679
if (timeout.tv_nsec > 1000000000) {
1680
timeout.tv_nsec -= 1000000000;
1684
r = usbi_cond_timedwait(&ctx->event_waiters_cond,
1685
&ctx->event_waiters_lock, &timeout);
1686
usbi_dbg("usbi_cond_timedwait returned %d", r);
1687
return (r == ETIMEDOUT);
1690
static void handle_timeout(struct usbi_transfer *itransfer)
1692
struct libusb_transfer *transfer =
1693
__USBI_TRANSFER_TO_LIBUSB_TRANSFER(itransfer);
1696
itransfer->flags |= USBI_TRANSFER_TIMED_OUT;
1697
r = libusb_cancel_transfer(transfer);
1699
usbi_warn(TRANSFER_CTX(transfer),
1700
"async cancel failed %d errno=%d", r, errno);
1703
#ifdef USBI_OS_HANDLES_TIMEOUT
1704
static int handle_timeouts_locked(struct libusb_context *ctx)
1708
static int handle_timeouts(struct libusb_context *ctx)
1713
static int handle_timeouts_locked(struct libusb_context *ctx)
1716
struct timespec systime_ts;
1717
struct timeval systime;
1718
struct usbi_transfer *transfer;
1720
if (list_empty(&ctx->flying_transfers))
1723
/* get current time */
1724
r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, &systime_ts);
1728
TIMESPEC_TO_TIMEVAL(&systime, &systime_ts);
1730
/* iterate through flying transfers list, finding all transfers that
1731
* have expired timeouts */
1732
list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
1733
struct timeval *cur_tv = &transfer->timeout;
1735
/* if we've reached transfers of infinite timeout, we're all done */
1736
if (!timerisset(cur_tv))
1739
/* ignore timeouts we've already handled */
1740
if (transfer->flags & USBI_TRANSFER_TIMED_OUT)
1743
/* if transfer has non-expired timeout, nothing more to do */
1744
if ((cur_tv->tv_sec > systime.tv_sec) ||
1745
(cur_tv->tv_sec == systime.tv_sec &&
1746
cur_tv->tv_usec > systime.tv_usec))
1749
/* otherwise, we've got an expired timeout to handle */
1750
handle_timeout(transfer);
1755
static int handle_timeouts(struct libusb_context *ctx)
1758
USBI_GET_CONTEXT(ctx);
1759
usbi_mutex_lock(&ctx->flying_transfers_lock);
1760
r = handle_timeouts_locked(ctx);
1761
usbi_mutex_unlock(&ctx->flying_transfers_lock);
1766
#ifdef USBI_TIMERFD_AVAILABLE
1767
static int handle_timerfd_trigger(struct libusb_context *ctx)
1771
r = disarm_timerfd(ctx);
1775
usbi_mutex_lock(&ctx->flying_transfers_lock);
1777
/* process the timeout that just happened */
1778
r = handle_timeouts_locked(ctx);
1782
/* arm for next timeout*/
1783
r = arm_timerfd_for_next_timeout(ctx);
1786
usbi_mutex_unlock(&ctx->flying_transfers_lock);
1791
/* do the actual event handling. assumes that no other thread is concurrently
1792
* doing the same thing. */
1793
static int handle_events(struct libusb_context *ctx, struct timeval *tv)
1796
struct usbi_pollfd *ipollfd;
1802
usbi_mutex_lock(&ctx->pollfds_lock);
1803
list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
1806
/* TODO: malloc when number of fd's changes, not on every poll */
1807
fds = malloc(sizeof(*fds) * nfds);
1809
usbi_mutex_unlock(&ctx->pollfds_lock);
1810
return LIBUSB_ERROR_NO_MEM;
1813
list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd) {
1814
struct libusb_pollfd *pollfd = &ipollfd->pollfd;
1815
int fd = pollfd->fd;
1818
fds[i].events = pollfd->events;
1821
usbi_mutex_unlock(&ctx->pollfds_lock);
1823
timeout_ms = (tv->tv_sec * 1000) + (tv->tv_usec / 1000);
1825
/* round up to next millisecond */
1826
if (tv->tv_usec % 1000)
1829
usbi_dbg("poll() %d fds with timeout in %dms", nfds, timeout_ms);
1830
r = usbi_poll(fds, nfds, timeout_ms);
1831
usbi_dbg("poll() returned %d", r);
1834
return handle_timeouts(ctx);
1835
} else if (r == -1 && errno == EINTR) {
1837
return LIBUSB_ERROR_INTERRUPTED;
1840
usbi_err(ctx, "poll failed %d err=%d\n", r, errno);
1841
return LIBUSB_ERROR_IO;
1844
/* fd[0] is always the ctrl pipe */
1845
if (fds[0].revents) {
1846
/* another thread wanted to interrupt event handling, and it succeeded!
1847
* handle any other events that cropped up at the same time, and
1849
usbi_dbg("caught a fish on the control pipe");
1855
/* prevent OS backend from trying to handle events on ctrl pipe */
1861
#ifdef USBI_TIMERFD_AVAILABLE
1862
/* on timerfd configurations, fds[1] is the timerfd */
1863
if (usbi_using_timerfd(ctx) && fds[1].revents) {
1864
/* timerfd indicates that a timeout has expired */
1866
usbi_dbg("timerfd triggered");
1868
ret = handle_timerfd_trigger(ctx);
1870
/* return error code */
1873
} else if (r == 1) {
1874
/* no more active file descriptors, nothing more to do */
1878
/* more events pending...
1879
* prevent OS backend from trying to handle events on timerfd */
1886
r = usbi_backend->handle_events(ctx, fds, nfds, r);
1888
usbi_err(ctx, "backend handle_events failed with error %d", r);
1895
/* returns the smallest of:
1896
* 1. timeout of next URB
1897
* 2. user-supplied timeout
1898
* returns 1 if there is an already-expired timeout, otherwise returns 0
1901
static int get_next_timeout(libusb_context *ctx, struct timeval *tv,
1902
struct timeval *out)
1904
struct timeval timeout;
1905
int r = libusb_get_next_timeout(ctx, &timeout);
1907
/* timeout already expired? */
1908
if (!timerisset(&timeout))
1911
/* choose the smallest of next URB timeout or user specified timeout */
1912
if (timercmp(&timeout, tv, <))
1923
* Handle any pending events.
1925
* libusb determines "pending events" by checking if any timeouts have expired
1926
* and by checking the set of file descriptors for activity.
1928
* If a zero timeval is passed, this function will handle any already-pending
1929
* events and then immediately return in non-blocking style.
1931
* If a non-zero timeval is passed and no events are currently pending, this
1932
* function will block waiting for events to handle up until the specified
1933
* timeout. If an event arrives or a signal is raised, this function will
1936
* \param ctx the context to operate on, or NULL for the default context
1937
* \param tv the maximum time to block waiting for events, or zero for
1939
* \returns 0 on success, or a LIBUSB_ERROR code on failure
1941
API_EXPORTED int libusb_handle_events_timeout_check(libusb_context *ctx,
1942
struct timeval *tv, int *completed)
1945
struct timeval poll_timeout;
1947
USBI_GET_CONTEXT(ctx);
1948
r = get_next_timeout(ctx, tv, &poll_timeout);
1950
/* timeout already expired */
1951
return handle_timeouts(ctx);
1955
if (libusb_try_lock_events(ctx) == 0) {
1957
if (completed == NULL || !*completed) {
1958
/* we obtained the event lock: do our own event handling */
1959
usbi_dbg("doing our own event handling");
1960
r = handle_events(ctx, &poll_timeout);
1962
libusb_unlock_events(ctx);
1966
/* another thread is doing event handling. wait for thread events that
1967
* notify event completion. */
1968
libusb_lock_event_waiters(ctx);
1970
if (completed == NULL || !*completed) {
1971
if (!libusb_event_handler_active(ctx)) {
1972
/* we hit a race: whoever was event handling earlier finished in the
1973
* time it took us to reach this point. try the cycle again. */
1974
libusb_unlock_event_waiters(ctx);
1975
usbi_dbg("event handler was active but went away, retrying");
1979
usbi_dbg("another thread is doing event handling, wait for notification");
1980
r = libusb_wait_for_event(ctx, &poll_timeout);
1982
libusb_unlock_event_waiters(ctx);
1987
return handle_timeouts(ctx);
1992
API_EXPORTED int libusb_handle_events_timeout(libusb_context *ctx,
1995
return libusb_handle_events_timeout_check(ctx, tv, NULL);
1999
* Handle any pending events in blocking mode. There is currently a timeout
2000
* hardcoded at 60 seconds but we plan to make it unlimited in future. For
2001
* finer control over whether this function is blocking or non-blocking, or
2002
* for control over the timeout, use libusb_handle_events_timeout() instead.
2004
* \param ctx the context to operate on, or NULL for the default context
2005
* \returns 0 on success, or a LIBUSB_ERROR code on failure
2007
API_EXPORTED int libusb_handle_events_check(libusb_context *ctx,
2013
return libusb_handle_events_timeout_check(ctx, &tv, completed);
2016
API_EXPORTED int libusb_handle_events(libusb_context *ctx)
2021
return libusb_handle_events_timeout_check(ctx, &tv, NULL);
2025
* Handle any pending events by polling file descriptors, without checking if
2026
* any other threads are already doing so. Must be called with the event lock
2027
* held, see libusb_lock_events().
2029
* This function is designed to be called under the situation where you have
2030
* taken the event lock and are calling poll()/select() directly on libusb's
2031
* file descriptors (as opposed to using libusb_handle_events() or similar).
2032
* You detect events on libusb's descriptors, so you then call this function
2033
* with a zero timeout value (while still holding the event lock).
2035
* \param ctx the context to operate on, or NULL for the default context
2036
* \param tv the maximum time to block waiting for events, or zero for
2038
* \returns 0 on success, or a LIBUSB_ERROR code on failure
2041
API_EXPORTED int libusb_handle_events_locked(libusb_context *ctx,
2045
struct timeval poll_timeout;
2047
USBI_GET_CONTEXT(ctx);
2048
r = get_next_timeout(ctx, tv, &poll_timeout);
2050
/* timeout already expired */
2051
return handle_timeouts(ctx);
2054
return handle_events(ctx, &poll_timeout);
2058
* Determines whether your application must apply special timing considerations
2059
* when monitoring libusb's file descriptors.
2061
* This function is only useful for applications which retrieve and poll
2062
* libusb's file descriptors in their own main loop (\ref pollmain).
2064
* Ordinarily, libusb's event handler needs to be called into at specific
2065
* moments in time (in addition to times when there is activity on the file
2066
* descriptor set). The usual approach is to use libusb_get_next_timeout()
2067
* to learn about when the next timeout occurs, and to adjust your
2068
* poll()/select() timeout accordingly so that you can make a call into the
2069
* library at that time.
2071
* Some platforms supported by libusb do not come with this baggage - any
2072
* events relevant to timing will be represented by activity on the file
2073
* descriptor set, and libusb_get_next_timeout() will always return 0.
2074
* This function allows you to detect whether you are running on such a
2079
* \param ctx the context to operate on, or NULL for the default context
2080
* \returns 0 if you must call into libusb at times determined by
2081
* libusb_get_next_timeout(), or 1 if all timeout events are handled internally
2082
* or through regular activity on the file descriptors.
2083
* \see \ref pollmain "Polling libusb file descriptors for event handling"
2085
API_EXPORTED int libusb_pollfds_handle_timeouts(libusb_context *ctx)
2087
#if defined(USBI_OS_HANDLES_TIMEOUT)
2089
#elif defined(USBI_TIMERFD_AVAILABLE)
2090
USBI_GET_CONTEXT(ctx);
2091
return usbi_using_timerfd(ctx);
2098
* Determine the next internal timeout that libusb needs to handle. You only
2099
* need to use this function if you are calling poll() or select() or similar
2100
* on libusb's file descriptors yourself - you do not need to use it if you
2101
* are calling libusb_handle_events() or a variant directly.
2103
* You should call this function in your main loop in order to determine how
2104
* long to wait for select() or poll() to return results. libusb needs to be
2105
* called into at this timeout, so you should use it as an upper bound on
2106
* your select() or poll() call.
2108
* When the timeout has expired, call into libusb_handle_events_timeout()
2109
* (perhaps in non-blocking mode) so that libusb can handle the timeout.
2111
* This function may return 1 (success) and an all-zero timeval. If this is
2112
* the case, it indicates that libusb has a timeout that has already expired
2113
* so you should call libusb_handle_events_timeout() or similar immediately.
2114
* A return code of 0 indicates that there are no pending timeouts.
2116
* On some platforms, this function will always returns 0 (no pending
2117
* timeouts). See \ref polltime.
2119
* \param ctx the context to operate on, or NULL for the default context
2120
* \param tv output location for a relative time against the current
2121
* clock in which libusb must be called into in order to process timeout events
2122
* \returns 0 if there are no pending timeouts, 1 if a timeout was returned,
2123
* or LIBUSB_ERROR_OTHER on failure
2125
API_EXPORTED int libusb_get_next_timeout(libusb_context *ctx,
2128
#ifndef USBI_OS_HANDLES_TIMEOUT
2129
struct usbi_transfer *transfer;
2130
struct timespec cur_ts;
2131
struct timeval cur_tv;
2132
struct timeval *next_timeout;
2136
USBI_GET_CONTEXT(ctx);
2137
if (usbi_using_timerfd(ctx))
2140
usbi_mutex_lock(&ctx->flying_transfers_lock);
2141
if (list_empty(&ctx->flying_transfers)) {
2142
usbi_mutex_unlock(&ctx->flying_transfers_lock);
2143
usbi_dbg("no URBs, no timeout!");
2147
/* find next transfer which hasn't already been processed as timed out */
2148
list_for_each_entry(transfer, &ctx->flying_transfers, list, struct usbi_transfer) {
2149
if (!(transfer->flags & USBI_TRANSFER_TIMED_OUT)) {
2154
usbi_mutex_unlock(&ctx->flying_transfers_lock);
2157
usbi_dbg("all URBs have already been processed for timeouts");
2161
next_timeout = &transfer->timeout;
2163
/* no timeout for next transfer */
2164
if (!timerisset(next_timeout)) {
2165
usbi_dbg("no URBs with timeouts, no timeout!");
2169
r = usbi_backend->clock_gettime(USBI_CLOCK_MONOTONIC, &cur_ts);
2171
usbi_err(ctx, "failed to read monotonic clock, errno=%d", errno);
2172
return LIBUSB_ERROR_OTHER;
2174
TIMESPEC_TO_TIMEVAL(&cur_tv, &cur_ts);
2176
if (!timercmp(&cur_tv, next_timeout, <)) {
2177
usbi_dbg("first timeout already expired");
2180
timersub(next_timeout, &cur_tv, tv);
2181
usbi_dbg("next timeout in %d.%06ds", tv->tv_sec, tv->tv_usec);
2191
* Register notification functions for file descriptor additions/removals.
2192
* These functions will be invoked for every new or removed file descriptor
2193
* that libusb uses as an event source.
2195
* To remove notifiers, pass NULL values for the function pointers.
2197
* Note that file descriptors may have been added even before you register
2198
* these notifiers (e.g. at libusb_init() time).
2200
* Additionally, note that the removal notifier may be called during
2201
* libusb_exit() (e.g. when it is closing file descriptors that were opened
2202
* and added to the poll set at libusb_init() time). If you don't want this,
2203
* remove the notifiers immediately before calling libusb_exit().
2205
* \param ctx the context to operate on, or NULL for the default context
2206
* \param added_cb pointer to function for addition notifications
2207
* \param removed_cb pointer to function for removal notifications
2208
* \param user_data User data to be passed back to callbacks (useful for
2209
* passing context information)
2211
API_EXPORTED void libusb_set_pollfd_notifiers(libusb_context *ctx,
2212
libusb_pollfd_added_cb added_cb, libusb_pollfd_removed_cb removed_cb,
2215
USBI_GET_CONTEXT(ctx);
2216
ctx->fd_added_cb = added_cb;
2217
ctx->fd_removed_cb = removed_cb;
2218
ctx->fd_cb_user_data = user_data;
2221
/* Add a file descriptor to the list of file descriptors to be monitored.
2222
* events should be specified as a bitmask of events passed to poll(), e.g.
2223
* POLLIN and/or POLLOUT. */
2224
int usbi_add_pollfd(struct libusb_context *ctx, int fd, short events)
2226
struct usbi_pollfd *ipollfd = malloc(sizeof(*ipollfd));
2228
return LIBUSB_ERROR_NO_MEM;
2230
usbi_dbg("add fd %d events %d", fd, events);
2231
ipollfd->pollfd.fd = fd;
2232
ipollfd->pollfd.events = events;
2233
usbi_mutex_lock(&ctx->pollfds_lock);
2234
list_add_tail(&ipollfd->list, &ctx->pollfds);
2235
usbi_mutex_unlock(&ctx->pollfds_lock);
2237
if (ctx->fd_added_cb)
2238
ctx->fd_added_cb(fd, events, ctx->fd_cb_user_data);
2242
/* Remove a file descriptor from the list of file descriptors to be polled. */
2243
void usbi_remove_pollfd(struct libusb_context *ctx, int fd)
2245
struct usbi_pollfd *ipollfd;
2248
usbi_dbg("remove fd %d", fd);
2249
usbi_mutex_lock(&ctx->pollfds_lock);
2250
list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2251
if (ipollfd->pollfd.fd == fd) {
2257
usbi_dbg("couldn't find fd %d to remove", fd);
2258
usbi_mutex_unlock(&ctx->pollfds_lock);
2262
list_del(&ipollfd->list);
2263
usbi_mutex_unlock(&ctx->pollfds_lock);
2265
if (ctx->fd_removed_cb)
2266
ctx->fd_removed_cb(fd, ctx->fd_cb_user_data);
2270
* Retrieve a list of file descriptors that should be polled by your main loop
2271
* as libusb event sources.
2273
* The returned list is NULL-terminated and should be freed with free() when
2274
* done. The actual list contents must not be touched.
2276
* \param ctx the context to operate on, or NULL for the default context
2277
* \returns a NULL-terminated list of libusb_pollfd structures, or NULL on
2280
API_EXPORTED const struct libusb_pollfd **libusb_get_pollfds(
2281
libusb_context *ctx)
2284
struct libusb_pollfd **ret = NULL;
2285
struct usbi_pollfd *ipollfd;
2288
USBI_GET_CONTEXT(ctx);
2290
usbi_mutex_lock(&ctx->pollfds_lock);
2291
list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2294
ret = calloc(cnt + 1, sizeof(struct libusb_pollfd *));
2298
list_for_each_entry(ipollfd, &ctx->pollfds, list, struct usbi_pollfd)
2299
ret[i++] = (struct libusb_pollfd *) ipollfd;
2303
usbi_mutex_unlock(&ctx->pollfds_lock);
2304
return (const struct libusb_pollfd **) ret;
2306
usbi_err(ctx, "external polling of libusb's internal descriptors "\
2307
"is not yet supported on Windows platforms");
2312
/* Backends call this from handle_events to report disconnection of a device.
2313
* The transfers get cancelled appropriately.
2315
void usbi_handle_disconnect(struct libusb_device_handle *handle)
2317
struct usbi_transfer *cur;
2318
struct usbi_transfer *to_cancel;
2320
usbi_dbg("device %d.%d",
2321
handle->dev->bus_number, handle->dev->device_address);
2323
/* terminate all pending transfers with the LIBUSB_TRANSFER_NO_DEVICE
2326
* this is a bit tricky because:
2327
* 1. we can't do transfer completion while holding flying_transfers_lock
2328
* 2. the transfers list can change underneath us - if we were to build a
2329
* list of transfers to complete (while holding look), the situation
2330
* might be different by the time we come to free them
2332
* so we resort to a loop-based approach as below
2333
* FIXME: is this still potentially racy?
2337
usbi_mutex_lock(&HANDLE_CTX(handle)->flying_transfers_lock);
2339
list_for_each_entry(cur, &HANDLE_CTX(handle)->flying_transfers, list, struct usbi_transfer)
2340
if (__USBI_TRANSFER_TO_LIBUSB_TRANSFER(cur)->dev_handle == handle) {
2344
usbi_mutex_unlock(&HANDLE_CTX(handle)->flying_transfers_lock);
2349
usbi_backend->clear_transfer_priv(to_cancel);
2350
usbi_handle_transfer_completion(to_cancel, LIBUSB_TRANSFER_NO_DEVICE);
added
removed
libusb1/libusb/io.c
