Tormod Volden [Sun, 10 Apr 2022 08:57:16 +0000 (10:57 +0200)]
libusb 1.0.26
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Martin Ling [Sat, 9 Apr 2022 09:22:01 +0000 (10:22 +0100)]
darwin: Document OS-specific behaviour of libusb_cancel_transfer
The behaviour of libusb_cancel_transfer is different on Darwin:
- Cancellation cancels all transfers on the same endpoint.
- A ClearFeature(ENDPOINT_HALT) request is sent after cancellation.
This documents both differences and their implications.
Fixes #1110
Closes #1121
Benjamin Berg [Thu, 7 Apr 2022 10:43:08 +0000 (12:43 +0200)]
darwin: Fix crash in log handler when stopping event thread
The darwin event thread (in contrast to other OS implementations) tries
to log to the context that created it. However, this context is only
guaranteed to be valid until the thread has started. It may be that
another context is the last one to be destroyed, causing the event
thread to log using an already destroyed context.
Fix this by only passing on ctx where it is acceptable.
Fixes #1108
Shawn Hoffman [Sun, 27 Feb 2022 02:17:22 +0000 (18:17 -0800)]
windows: Silence zero-sized array warning in libusb.h for msvc
Silence "nonstandard extension used : zero-sized array in struct/union"
(C4200) warning caused by use of ZERO_SIZED_ARRAY in structures defined
in libusb.h when compiling with msvc.
A libusb user including the header from a c++ file is more likely to be
annoyed by the warning because it is activated at level 2 as opposed to
being included from a C file, where it's not activated until level 4.
(on msvc, level 4 is the last step before -Wall).
This causes problems when you also enable -Werror for your own source,
since the libusb header just gets merged into the compilation unit.
9>hotplugtest.c
9>C:\src\libusb\libusb\libusb.h(786,11): error C2220: the following warning is treated as an error
9>C:\src\libusb\libusb\libusb.h(786,11): error C2220: uint8_t dev_capability_data[ZERO_SIZED_ARRAY];
9>C:\src\libusb\libusb\libusb.h(786,11): error C2220: ^
9>C:\src\libusb\libusb\libusb.h(786,11): warning C4200: nonstandard extension used: zero-sized array in struct/union
9>C:\src\libusb\libusb\libusb.h(786,11): warning C4200: uint8_t dev_capability_data[ZERO_SIZED_ARRAY];
9>C:\src\libusb\libusb\libusb.h(786,11): warning C4200: ^
9>C:\src\libusb\libusb\libusb.h(811,47): warning C4200: nonstandard extension used: zero-sized array in struct/union
5>C:\src\libusb\libusb\libusb.h(811,47): warning C4200: struct libusb_bos_dev_capability_descriptor *dev_capability[ZERO_SIZED_ARRAY];
5>C:\src\libusb\libusb\libusb.h(811,47): warning C4200: ^
9>C:\src\libusb\libusb\libusb.h(1284,38): warning C4200: nonstandard extension used: zero-sized array in struct/union
9>C:\src\libusb\libusb\libusb.h(1284,38): warning C4200: struct libusb_iso_packet_descriptor iso_packet_desc[ZERO_SIZED_ARRAY];
9>C:\src\libusb\libusb\libusb.h(1284,38): warning C4200: ^
9>Done building project "hotplugtest.vcxproj" -- FAILED.
Closes #1112
Tormod Volden [Sun, 3 Apr 2022 13:35:00 +0000 (15:35 +0200)]
windows: Update isochronous OUT packet actual length
On Windows, unlike other platforms, the isochronous packet actual_length
value is not set on completion of OUT transfers. However, our API
requires the user to check this value for isochronous transfers instead
of the transfer actual_length, if the transferred length is of interest.
The usbd Length field is not used for isochronous OUT transfers:
https://docs.microsoft.com/en-us/windows-hardware/drivers/ddi/usb/ns-usb-_usbd_iso_packet_descriptor
To make it consistent with other platforms, just return the requested
length.
Fixes #1105
Closes #1107
Tested-by: Xiaofan Chen <xiaofanc@gmail.com>
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Mon, 28 Mar 2022 11:23:37 +0000 (13:23 +0200)]
windows: Fix type mismatch in winusbx_copy_transfer_data
We cannot use the CHECK_WINUSBX_AVAILABLE macro since it
may return with the wrong error type/value.
Fixes #1100
Closes #1101
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Mon, 28 Mar 2022 13:37:28 +0000 (15:37 +0200)]
CI: Add MinGW build on MSYS2 (Windows)
Closes #1102
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Tue, 22 Mar 2022 22:13:35 +0000 (23:13 +0100)]
libusb 1.0.26-rc1
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
osy [Sun, 16 May 2021 06:29:54 +0000 (23:29 -0700)]
darwin: Reset by re-enumerate on non-macOS platforms
On non-macOS platforms, ResetDevice() does nothing so we use the "old" way of
calling re-enumerate. If the device is captured, we have to re-enumerate, then
re-capture, re-authorize, and finally restore the state.
Closes #1088
Tormod Volden [Wed, 23 Mar 2022 08:34:00 +0000 (09:34 +0100)]
darwin: Fix Makefile build on macOS 10.11 and older
TARGET_OS_OSX is not defined when building with the clang 8 provided by
Xcode 8.2.1 on 10.11.5 (unless the Xcode project is used).
If it is not defined we'll just assume we have a "OSX" target.
Closes #1097
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Seneral [Thu, 2 Jul 2020 00:23:36 +0000 (02:23 +0200)]
windows: Correct isochronous frame calculation for full-speed devices
The WinUSB isochronous transfer implementation always assumed HighSpeed
microframes.
https://github.com/libusb/libusb/issues/747#issuecomment-
649491619
References #747
References #749
Keith Ahluwalia [Mon, 21 Mar 2022 10:51:55 +0000 (10:51 +0000)]
windows: Fix WinUSB isochronous transfer regression
Put back missing CHECK_WINUSBX_AVAILABLE macro.
The subapi is generally set in the backend functions themselves, using
the macro CHECK_WINUSBX_AVAILABLE. However, this macro was (likely
inadvertently) removed from winusbx_copy_transfer_data() in commit
9c28ad2 causing the subapi in this function to always be SUB_API_NOTSET.
This results in the error:
libusb: debug [winusbx_copy_transfer_data] unsupported API call for 'copy_transfer_data' (unrecognized device driver)
https://github.com/libusb/libusb/issues/747#issuecomment-
649619615
References #749
Closes #1095
Benjamin Berg [Tue, 22 Feb 2022 10:45:38 +0000 (11:45 +0100)]
core: Install first context as implicit default (and warn about its use)
There was a behaviour change in libusb 1.0.25 which triggers issues when
the API is misused. This caused for instance gutenprint to crash, see
https://bugzilla.redhat.com/show_bug.cgi?id=2055504
This seems to affect several applications "out in the wild".
For now, work around this by installing an implicit default. But, change
the code to log an error in case this "feature" is being used.
This will allow some grace time for developers to fix their
applications, before we at a later point revert to the stricter
behaviour.
Fixes #1089
Mathias Hjärtström [Thu, 17 Mar 2022 11:43:18 +0000 (12:43 +0100)]
windows: Fix product string retrieval on HID composite devices
A HID composite device with three interfaces (e.g. keyboard and touch
screen built into one), with all three interfaces referencing their own
names in the interface descriptor, was reported to have an iProduct
string equal to the name of the last interface instead of the actual
product name (e.g. "TOUCH" repeated twice instead of "PRODUCT" and
"TOUCH").
This behavior differ from what for instance Microsoft USB Device Viewer
will report for the same device. This fix will make them report the same
thing.
Use HidD_GetIndexedString() instead of HidD_GetProductString(), as the
latter would otherwise return the name of the interface instead of the
iProduct string whenever the iInterface member of the
USB_INTERFACE_DESCRIPTOR structure for the interface is nonzero (see
Remarks section in the Microsoft documentation of the HID API routines).
Closes #1091
Tormod Volden [Wed, 2 Feb 2022 15:12:25 +0000 (16:12 +0100)]
darwin: Always use GetPipePropertiesV3 when available
The V3 API goes back to macOS 10.8.2 (2012), and we have already been
using it when targeting 10.9 or higher for a while. However in a few
places we were still using the older API.
Affects get_endpoints() and submit_iso_transfer().
In the latter also:
- Check success of pipe properties retrieval.
- Remove a duplicate call to GetPipeProperties().
References #919
Closes #1056
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Ryan Hileman [Fri, 18 Mar 2022 15:54:52 +0000 (08:54 -0700)]
Add interface bounds checks in darwin and windows endpoint functions
A broken device with invalid interface numbering could cause
an out-of-bounds array access.
Fixes #1039
Closes #1093
Yegor Yefremov [Sat, 12 Feb 2022 14:07:36 +0000 (15:07 +0100)]
android: Fix typo and syntax in comments
Fix a typo and add missing commas.
Also, delete a blank line at the end of the file and
use /* */ for comments consistently.
Closes #1071
Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Tormod Volden [Wed, 9 Feb 2022 10:41:26 +0000 (11:41 +0100)]
appveyor: Retain MinGW and cygwin build artifacts
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Thu, 3 Feb 2022 22:57:48 +0000 (23:57 +0100)]
windows: Add "or later" to debug print of Windows version 8
Windows 10 and 11 is detected as Windows 8 unless there is an
application manifest. We simply cannot tell by appropriate means.
Note that we currently only use the detected version to check for the
minimal version (Vista), so it doesn't matter much.
Closes #1021
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Wed, 9 Feb 2022 11:35:17 +0000 (12:35 +0100)]
windows: Discern Windows 11 (if manifest allows)
Pre-releases of Windows 11 had build numbers from 20000.
Thanks to Pete Batard for extensive explanation.
Fixes #1021
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Thu, 3 Feb 2022 22:52:55 +0000 (23:52 +0100)]
windows: Add Windows 11 to list of defined releases
And guess "12" for whatever comes after. Although with macOS already at
12 this is not given :)
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Wed, 2 Feb 2022 17:15:59 +0000 (18:15 +0100)]
darwin: Avoid cast warnings in time calculation
Also use the USEC_PER_SEC macro for readability.
References #1035
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Benjamin Berg [Fri, 18 Feb 2022 16:00:45 +0000 (17:00 +0100)]
Add container build for umockdev
Benjamin Berg [Fri, 18 Feb 2022 15:26:29 +0000 (16:26 +0100)]
tests: Add some umockdev based tests
Add some basic umockdev tests. The setup in this case is all in one
process, but umockdev in principle could also work between processes.
This is just a start, more tests would make sense.
Tormod Volden [Mon, 21 Feb 2022 22:45:15 +0000 (23:45 +0100)]
darwin: Avoid possibly undefined version macro in configure.ac
References #1056
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Andrew Goodney [Wed, 16 Feb 2022 21:28:14 +0000 (13:28 -0800)]
darwin: Use usbi_get_monotonic_time() wrapper
On macOS < 10.12 we cannot use clock_gettime().
Closes #1070
Tormod Volden [Thu, 10 Feb 2022 17:43:23 +0000 (18:43 +0100)]
windows: Link gcc helpers statically on MinGW
Fixes #1049
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Yegor Yefremov [Sun, 6 Feb 2022 06:12:14 +0000 (07:12 +0100)]
linux_usbfs: Fix maybe-uninitialized error with -O3
Initialize active_config to an invalid value to avoid the following
compilation error:
os/linux_usbfs.c: In function ‘op_get_configuration’:
os/linux_usbfs.c:1452:12: error: ‘active_config’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
1452 | *config = (uint8_t)active_config;
Closes #1062
Closes #1063
Signed-off-by: Yegor Yefremov <yegorslists@googlemail.com>
Fabrice Fontaine [Mon, 7 Feb 2022 10:55:54 +0000 (11:55 +0100)]
configure.ac: link with -latomic if needed
libusb unconditionally uses atomic_fetch_add since version 1.0.25 and
commit 1a08aa8 and commit eed8a37 but some architectures (e.g. sparc)
needs to link with -latomic to be able to use it. So check if -latomic
is needed and update libusb-1.0.pc accordingly to avoid the following
build failure with openocd:
/home/buildroot/autobuild/instance-0/output-1/host/opt/ext-toolchain/bin/../lib/gcc/sparc-buildroot-linux-uclibc/10.3.0/../../../../sparc-buildroot-linux-uclibc/bin/ld: /home/buildroot/autobuild/instance-0/output-1/host/sparc-buildroot-linux-uclibc/sysroot/usr/lib/libusb-1.0.so: undefined reference to `__atomic_fetch_add_4'
collect2: error: ld returned 1 exit status
Full build log:
http://autobuild.buildroot.org/results/
4a27a769bb3cdf78643c3049b87d792178d6512c
Closes #1064
Signed-off-by: Fabrice Fontaine <fontaine.fabrice@gmail.com>
Benjamin Berg [Tue, 1 Mar 2022 14:08:38 +0000 (15:08 +0100)]
core: Suppress hotplug events during initial enumeration
The initial enumeration should not result in hotplug events to be fired.
This is just a convenience though, API users still need to be prepared
to be notified a second time for a device that was plugged in between
libusb_init and libusb_hotplug_register_callback.
This regressed with commit 6929b82 ("Fix segmentation fault in
libusb_init() if usbi_backend.init() fails"). This commit avoids the
mentioned segmentation fault by avoiding to clean up the hotplug code if
it was not yet initialised.
Fixes #1082
Closes #1090
References #989
Benjamin Berg [Tue, 15 Feb 2022 10:13:41 +0000 (11:13 +0100)]
io: Track device in usbi_transfer
transfer->dev_handle currently has the behaviour that it will be unset
if the device is closed. The sync API uses this fact to catch an error
case.
In other cases, transfer->dev_handle will keep its value, which means
that if the transfer lives longer than the device handle, the pointer
becomes invalid.
The transfer does however keep a reference to the device, which owns the
pointer to the context. As such, we can track this reference internal to
the transfer, and it is set while the transfer is in-flight.
With this, switch the logging infrastructure to use itransfer->dev->ctx
while checking that itransfer->dev is non-NULL.
Note that this was a regression caused by 6cae9c6 ("core: update
usbi_dbg to take the context as an argument"), specifically when
resolving the context while freeing a transfer after closing a device.
Note that the transfer will now keep a reference to the device until it
is free'ed. This allows it to use the correct context for logging even
in libusb_free_transfer.
The alternative to all this would be to just explicitly pass NULL to the
log handler in libusb_free_transfer.
Fixes #1038
Closes #1073
Benjamin Berg [Fri, 4 Feb 2022 21:50:28 +0000 (22:50 +0100)]
core: Unset device ctx if it has been destroyed
Devices can outlive their context in some cases (in particular with
python garbage collection). Guard against this by clearing the ctx
pointer so that it is not pointing to uninitialized memory.
Closes #1058
Tormod Volden [Mon, 31 Jan 2022 12:56:37 +0000 (13:56 +0100)]
libusb 1.0.25
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sat, 29 Jan 2022 21:04:08 +0000 (22:04 +0100)]
appveyor: Retain selected build artifacts
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sat, 22 Jan 2022 13:39:45 +0000 (14:39 +0100)]
examples: Do not assume positive errno macros
Some functions were returning e.g. -ENOMEM and the caller would check
for negative return values. However, on Haiku, contrary to modern
standards, these macros are negative, so this logic would fail. In any
case, change the function to return -1 instead. For good measure also
set errno to the appropriate value, although it is not used in the
current code.
This was discovered on Haiku builds because the value for ENOMEM is
INT_MIN which cannot be negated without overflow.
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sat, 22 Jan 2022 11:53:57 +0000 (12:53 +0100)]
netbsd: Fix typo preventing build
Fixup of commit 6cae9c6
Tested-by: Xiaofan Chen <xiaofanc@gmail.com>
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Fri, 21 Jan 2022 20:10:16 +0000 (21:10 +0100)]
libusb 1.0.25-rc1
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Fri, 21 Jan 2022 19:20:59 +0000 (20:20 +0100)]
Add libusb_set_option to libusb-1.0.def manually
Since libusb_set_option() is variadic, there is no symbol
for it with @ appended in the DLL, and the automatic
generation of libusb-1.0.def misses it.
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Fri, 21 Jan 2022 17:20:51 +0000 (18:20 +0100)]
Update AUTHORS
Based on `git shortlog -s | cut -c8-` and a lot of manual fixups.
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Fri, 21 Jan 2022 16:58:59 +0000 (17:58 +0100)]
Update libusb-1.0.def
Updated by running the following on a libusb-1.0 DLL generated by MinGW32:
echo -e "LIBRARY \"libusb-1.0.dll\"\nEXPORTS" > libusb-1.0.def
strings libusb-1.0.dll | grep ^libusb | grep @ | LC_ALL=C sort | uniq | \
sed -e "s/\(.*\)@\([0-9]*\)/ \1\n \1@\2 = \1/" | LC_ALL=C sort >> libusb-1.0.def
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Ludovic Rousseau [Fri, 31 Dec 2021 17:29:28 +0000 (18:29 +0100)]
Add github action for macOS
Signed-off-by: Ludovic Rousseau <ludovic.rousseau@free.fr>
Ludovic Rousseau [Fri, 31 Dec 2021 17:29:04 +0000 (18:29 +0100)]
Add github action for GNU/Linux
Signed-off-by: Ludovic Rousseau <ludovic.rousseau@free.fr>
Andrew Goodney [Tue, 28 Dec 2021 04:34:05 +0000 (20:34 -0800)]
The current code for calculating the timeout in darwin_reenumerate_device_timeout() doesn't calculate elapsed microseconds, it counts the number of times the loop runs.
This results in very long timeouts.
This PR uses clock_gettime(CLOCK_MONOTONIC, ...) to calculate the elapsed time
Closes #1035
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Wed, 1 Dec 2021 10:38:49 +0000 (11:38 +0100)]
darwin: Minor tidy up in darwin_set_interface_altsetting()
Closes #1032
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Nathan Hjelm [Tue, 30 Nov 2021 22:10:01 +0000 (15:10 -0700)]
darwin: fix behavior of libusb_set_interface_alt_setting when it stalls the pipe
In some versions of macOS a pipe stall returned when setting the alternate interface
causes the interface to become unusable. To handle this case the backend was always
re-claiming the interface before clearing the pipe stall. In macOS Monterey
unconditionally re-claiming the interface leads to an error due to the process already
having exclusive access. To resolve this issue we attempt to clear the halt and only
re-claim the interface if clearing the pipe stall returns kIOUSBUnknownPipeErr. Tested
with 12.0.1 and 10.13 and get the expected results in both cases with a custom USB
device that has this behavior.
Fixes #1011
Closes #1031
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Fri, 1 Oct 2021 21:57:14 +0000 (23:57 +0200)]
Bump LIBUSB_API_VERSION
LIBUSB_OPTION_NO_DEVICE_DISCOVERY is new.
Closes #1004
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Nathan Hjelm <hjelmn@me.com>
Tormod Volden [Fri, 1 Oct 2021 22:05:32 +0000 (00:05 +0200)]
Keep LIBUSB_OPTION_WEAK_AUTHORITY as a macro with same value
If the value is always the same we retain binary compatibility.
References #935
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Nathan Hjelm <hjelmn@me.com>
AdamKorcz [Thu, 25 Nov 2021 11:03:30 +0000 (11:03 +0000)]
Fuzzing: Add CIFuzz
Matthias Bolte [Wed, 22 Sep 2021 11:40:23 +0000 (13:40 +0200)]
core: Unlock and clear default ctx in all error paths in libusb_init
Commit
32a22069428cda9d63aa666e92fb8882a83d4515 reordered and
refactored libusb_init. This resulted in moving code outside the
default ctx lock that was unrelated to it. But this resulted in
broken error path cleanup logic that leaves the default ctx as
locked, half initialized and freed in case a libusb_set_option or
the usbi_io_init call fails.
Undo part of the previous reordering to unlock and clear the default
ctx in all error paths in libusb_init.
Closes #995
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Tue, 9 Nov 2021 08:57:57 +0000 (09:57 +0100)]
core: Minor code cleanup in libusb_init (NOP)
After commit f7084fe the if clause would always be true, so remove it.
Thanks to "xloem" for reporting.
References #942
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Thu, 28 Oct 2021 12:48:55 +0000 (14:48 +0200)]
Windows: Allow synchronous control transfers (for libusb0)
Some of the changes in commit 9c28ad2 rely on all transfers having (or
appearing to have) asynchronous completion.
However, the libusb0.sys backend of libusbk.dll performs all control
transfers synchronously and ignores any "overlapped" structure handed to
it. Our asynchronous handling will in this case be pending and
eventually time out although the USB request itself was successful.
Therefore restore the possibility of synchronous completion of control
transfers, by forcing the completion handling on a successful return
from request submission. This brings the code closer to how it was
established in commit ce95f65.
Fixes #94
Tested-by: Xiaofan Chen <xiaofanc@gmail.com>
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Julian Scheel [Mon, 8 Jun 2020 12:32:16 +0000 (14:32 +0200)]
Windows: Fix timeouts if only endpoint 0 exists
The default timeouts of 5 seconds used by WinUSB are cleared in
configure_endpoints(). This was not called if not at least one extra
endpoint apart from endpoint 0 existed, so that in those cases the
default timeout of 5 seconds remained in place.
Fix this by calling configure_endpoints, even if no interface specific
endpoints were found, so that it can configure endpoint 0.
Closes #963
Signed-off-by: Julian Scheel <julian@jusst.de>
[Tormod: Refactored to reduce duplicated code]
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sat, 16 Oct 2021 22:13:44 +0000 (00:13 +0200)]
Windows: Ignore any device interface GUIDs beyond the first
The NZXT Kraken Z73 device has been reported to brokenly have two
identical device interface GUIDs registered. Accept such cases by
treating the ERROR_MORE_DATA return value as success and processing the
first returned GUID as if it would be the only one.
If one day we see legitimate cases of multiple GUIDs we will need to
deal with them properly.
References jdkvision/libusb#1
Closes #1010
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Evan Miller [Sat, 6 Nov 2021 23:42:53 +0000 (19:42 -0400)]
darwin: Fix compilation on macOS 10.9 and earlier
kUSBReEnumerateCaptureDeviceMask was introduced in the 10.10 SDK, so
compiling on 10.9 and earlier will fail with:
os/darwin_usb.c: In function 'darwin_reenumerate_device':
os/darwin_usb.c:1749:18: error: 'kUSBReEnumerateCaptureDeviceMask' undeclared (first use in this function); did you mean 'USBReEnumerateOptions'?
options |= kUSBReEnumerateCaptureDeviceMask;
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
USBReEnumerateOptions
os/darwin_usb.c:1749:18: note: each undeclared identifier is reported only once for each function it appears in
Downstream report: https://trac.macports.org/ticket/63668
Closes #1023
Peter-St [Wed, 30 Dec 2020 21:07:23 +0000 (22:07 +0100)]
examples: Add example for unrooted Android
Closes #830
Tormod Volden [Mon, 1 Nov 2021 22:40:44 +0000 (23:40 +0100)]
descriptor: Provide more detailed warning if suspicious bLength
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sat, 2 Oct 2021 14:34:46 +0000 (16:34 +0200)]
doc: Explain default options in libusb_set_option documentation
References #942
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Bruno Harbulot [Thu, 12 Aug 2021 19:26:12 +0000 (20:26 +0100)]
core: Apply default options to all new contexts
The default options configured with libusb_set_option(NULL, ...)
were only applied when the default context was created, and were
ignored when calling libusb_init() with a non-null context pointer.
Make sure the default options will be applied to all new contexts. This
is important when using LIBUSB_OPTION_NO_DEVICE_DISCOVERY which must
be respected during the context initialization in order to work.
Closes #942
[Tormod: amend comments]
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Tormod Volden [Sun, 31 Oct 2021 17:32:25 +0000 (18:32 +0100)]
xusb: Print total length of configuration descriptor
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
luz paz [Sat, 30 Oct 2021 21:15:04 +0000 (17:15 -0400)]
Fix various typos in docs/comments
Found via `codespell -q 3`
Closes #1015
Tormod Volden [Sun, 31 Oct 2021 08:12:07 +0000 (09:12 +0100)]
xusb: Print configuration descriptor length
Also inform about reading OS string descriptor.
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Sean McBride [Fri, 20 Aug 2021 19:57:52 +0000 (15:57 -0400)]
darwin: After last CFRelease of struct field, set it to NULL
Other code checks this field for being NULL or not, so NULLing it upon
last release seems the right thing to do. I noticed this during code
review, I never actually reproduced any problem.
Closes #981
Sean McBride [Fri, 29 Oct 2021 13:47:01 +0000 (09:47 -0400)]
doc: Improve comments related to device discovery and hotplug
Closes #1013
Tormod Volden [Sat, 2 Oct 2021 09:03:29 +0000 (11:03 +0200)]
doc: Update note about LIBUSB_TRANSFER_ADD_ZERO_PACKET availability
Closes #1005
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Paul Cercueil [Wed, 6 Oct 2021 13:31:24 +0000 (14:31 +0100)]
Windows: Fix ssize_t unconditionally defined
The standard procedure to define ssize_t on Windows is to wrap it around
a check for the _SSIZE_T_DEFINED macro.
If not done, it makes it impossible to use the libusb.h header along
with other headers (from other libraries) that also attempt to define
ssize_t.
Closes #1007
Signed-off-by: Paul Cercueil <paul@crapouillou.net>
Bohdan Tymkiv [Wed, 28 Jul 2021 14:46:04 +0000 (17:46 +0300)]
Windows: Partial fix for data race in composite_copy_transfer_data
This only fixes issue (A):
libusb: error [composite_copy_transfer_data] program assertion failed - no function to copy transfer data
References #966
Closes #967
Nathan Hjelm [Thu, 28 Oct 2021 16:36:30 +0000 (10:36 -0600)]
darwin: fix deprecation warning about kIOMasterPortDefault
The kIOMasterPortDefault constant has been replaced by kIOMainPortDefault. Both
are synonmys for 0. To avoid extra checks on the target macOS release just go
ahead and use 0.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Thu, 28 Oct 2021 16:25:03 +0000 (10:25 -0600)]
darwin: add support for handling new error code (kUSBHostReturnPipeStalled)
With macOS Monterey Apple started using a new error code to indicate a pipe
stall. This new error code needs to be handled both in translation to libusb
error codes and when handling errors setting the alt setting.
References #1011
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Wed, 22 Sep 2021 13:55:33 +0000 (15:55 +0200)]
git hooks: Remove "source" bashism
"source" is not supported in Bourne shell
Closes #997
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Peter-St [Mon, 21 Dec 2020 20:47:24 +0000 (21:47 +0100)]
Update Android Readme
Add another approach for connecting USB devices on Android.
References #830
Closes #996
Ankur Verma [Tue, 14 Sep 2021 01:25:30 +0000 (18:25 -0700)]
Fix segmentation fault in libusb_init() if usbi_backend.init() fails
If the backend init fails, the control goes to err_io_exit which tries
to clean up hotplug related lists that aren't initialized. Moving
hotplug_init before makes sure the lists are valid so if backend init
fails they get cleaned up without errors.
Closes #989
Sean McBride [Fri, 20 Aug 2021 19:57:52 +0000 (15:57 -0400)]
Fix comment typos
References #981
xloem [Wed, 4 Aug 2021 10:38:07 +0000 (10:38 +0000)]
Android build name correction
Fixes #31
See #143 and #596 for solutions to the whole build system.
Closes #971
mcuee [Fri, 27 Aug 2021 09:49:30 +0000 (17:49 +0800)]
Fix Windows HID backend missing byte
When the report id is zero, we should indeed skip the first byte of
transfer_priv->hid_buffer, since it is not part of the actual report.
But the report size should not change.
Thanks to Jonas Malaco who has done the investigation in this issue.
Reference: Jonas Malaco's comments
https://github.com/libusb/libusb/issues/902#issuecomment-
810726897
Fixes #902
Fixes #977
Closes #986
Nathan Hjelm [Sun, 8 Aug 2021 15:22:31 +0000 (09:22 -0600)]
darwin: fix typo
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Sun, 8 Aug 2021 15:16:38 +0000 (09:16 -0600)]
darwin: improve the error message on kernel driver detach failure
Detaching a kernel driver in macOS requires that either the binary has the
com.apple.vm.device-access entitlement or is run with effective UID 0 (root). This
commit adds an effective UID check before attempting to capture a device and prints
a warning that neither is the user root nor does the binary have the correct
entitlement.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Sun, 8 Aug 2021 05:02:16 +0000 (23:02 -0600)]
darwin: fix USB capture for root
This commit restores the ability to capture-open a USB device when running as root.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Tue, 27 Jul 2021 05:29:01 +0000 (23:29 -0600)]
configure: remove usage of deprecated AC_HELP_STRING
This macro was replaced by AS_HELP_STRING. This commit updates the lingering deprecated call
with the new macro as used elsewhere in configure.ac.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Tue, 27 Jul 2021 05:23:12 +0000 (23:23 -0600)]
core: set default backend options before calling backend init
This commit restores the previous behavior with regards to setting some backend options and
restores the ability to set default context options before libusb_init. None of the backends
use anything set in init in their set_option function so this should be safe with all
backends and options.
References #942
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
mcuee [Sun, 25 Jul 2021 08:26:37 +0000 (16:26 +0800)]
Update libusb-1.0.def
This fixes issue #944 C++ symbol for libusb_set_option not exported correctly.
As per the report in issue #944, original commit is correct but subsequent edits cause the regression.
https://github.com/libusb/libusb/blob/v1.0.22-rc1/libusb/libusb-1.0.def#L156
Closes #964
Signed-off-by: Xiaofan Chen xiaofanc@gmail.com
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Thu, 22 Jul 2021 05:12:19 +0000 (23:12 -0600)]
darwin: only attempt to auto-detach driver is one is active
This commit improves the auto-detach code so that it doesn't attempt to detach a kernel
driver if none exists. This will avoid extra info/warning output on the failed detach.
Closes #960
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Wed, 21 Jul 2021 17:49:37 +0000 (11:49 -0600)]
darwin: use the IO registry to detect if a kernel driver is attached to an interface
The implementation of libusb_kernel_driver_active was attempting to open the interface to
check if a driver is attached. This may have side effects (like configuring the device)
that may be unexpected to the user. This commit updates the code to find the interface's
IO registry entry (either IOUSBHostInterface or the legacy IOUSBInterface) and check if
the entry has a child entry. A child entry indicates that a driver is currently attached.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Wed, 21 Jul 2021 17:47:25 +0000 (11:47 -0600)]
xusb: add check for interface kernel driver
This commit adds a call to libusb_kernel_driver_active and prints out the result. This
provides a way to quickly check the result when testing.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Tue, 20 Jul 2021 15:31:06 +0000 (09:31 -0600)]
core: update usbi_dbg to take the context as an argument
This commit fixes a performance issue caused by the disconnection of the first context
allocated from the default context. usbi_dbg now takes the explicit context instead of
relying on the default context (which may not exist) in most cases. All call sites have
been updated to pass the context or explicitly pass NULL if the context is not
available. We should actively discourage using NULL as the context in the future and
patch all call sites to always pass the context.
Fixes #951
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Tue, 20 Jul 2021 21:48:27 +0000 (15:48 -0600)]
darwin: downgrade error message on failure to open interface to info
This resolves an issue introduced by the implementation of the kernel detach code. This
code used claim_interface to detect if a kernel driver is attached. The issue is that if
a driver is attached or the interface is not available and an error message was printed
out. Since this is not an error the info level is more appropriate.
Also downgrading a number of warnings to clean up the output.
Closes #955
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Tue, 20 Jul 2021 21:44:59 +0000 (15:44 -0600)]
darwin: add a comment on how the bus number is calculated
This commit adds a comment about the bus number calculation to indicate how the
location ID is constructed and why libusb is using the top byte for the bus
number. It also indicates that the bus number starts at 0.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Dmitry Zakablukov [Tue, 13 Jul 2021 14:50:28 +0000 (17:50 +0300)]
darwin: update the list of endpoints after successfull SetAlternateInterface()
Closes #950
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Sun, 11 Jul 2021 12:40:01 +0000 (14:40 +0200)]
winusb: Ignore missing DeviceInterfaceGUID
Closes #948
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Fri, 9 Jul 2021 15:17:33 +0000 (09:17 -0600)]
darwin: improve support for auto-detaching a kernel driver
This commit improves the code around auto attach/detach of a kernel driver:
- Log a warning when a driver can not be auto-detached and continue when claiming an
interface. It could be the case that the interface has no driver attached or that
the claim will still succeed.
- Log a warning if the driver can not be auto-attached and return success when
releasing an interface. The operation did complete successfully but the
re-enumeration may have failed. This can be discovered on the next device
access.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Fri, 9 Jul 2021 04:43:28 +0000 (22:43 -0600)]
darwin: fix SEGV on libusb_close after failed re-enumeration
This commit fixes an issue that can occur if the user calls libusb_close after a failed attempt
of libusb_reset (re-enumerate). This can lead to an attempt to dereference dpriv->device which
was set to NULL during the re-enumeration.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Dmitry Zakablukov [Wed, 30 Jun 2021 16:45:53 +0000 (19:45 +0300)]
[Issue-912] Lock open_devs_lock in context to avoid access violation
Closes #943
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Tue, 15 Jun 2021 20:33:45 +0000 (22:33 +0200)]
New NO_DEVICE_DISCOVERY option to replace WEAK_AUTHORITY option
The option disables device enumeration (and hotplug, which depends on
enumeration). Since this has a much broader usage scope than "weak
authority", give the option a name reflecting what it does.
For now, keep the old option name working as an alias.
Closes #935
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Craig Hutchinson [Tue, 8 Jun 2021 18:03:43 +0000 (20:03 +0200)]
windows: Allow GUID with and without trailing zeroes
Improve the GUID length check to accommodate the cases where the GUID
string fetched from registry does not have a trailing zero, or if a
multi-string entry, one or two trailing zeroes.
Closes #927
Signed-off-by: Craig Hutchinson <54269136+CraigHutchinson@users.noreply.github.com>
[Tormod: Use correct types and simplify comparisons]
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Ryan Schmidt [Sat, 12 Jun 2021 03:19:49 +0000 (22:19 -0500)]
Fix build failure on Mac OS X 10.7
Apple documentation says kUSBDeviceSpeedSuper is available in 10.7 and
later but in fact it is only defined in the 10.8 SDK and later.
Closes #410
Closes #929
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Sat, 12 Jun 2021 04:27:19 +0000 (22:27 -0600)]
darwin: release device parent reference when re-enumerating device
This commit fixes a dangling device reference that can occur when a device is
being re-enumerated. The problem is the code was unconditionally re-caching the
parent device which then takes a reference on the parent device. If the field
was already populated this reference was not needed. To ensure the code works
even if the parent device also changes during the re-enumeration the code
release the parent reference and grabs a fresh reference.
Fixes #924
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Mon, 7 Jun 2021 16:19:41 +0000 (18:19 +0200)]
Do not restrict "weak authority" option to Android
libusb_wrap_sys_device() is an often needed feature on Android, but can
also be very convenient on other Linux platforms.
In many cases the initial device scanning is not needed and
LIBUSB_OPTION_WEAK_AUTHORITY thus provides a performance enhancement.
Closes #925
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Fri, 11 Jun 2021 13:59:23 +0000 (07:59 -0600)]
core: really fix dangling devices
The prior fix made an incorrect assumption about the reference counting of
parent devices. This CL fixes the issue by checking if the parent device
should be removed from the list as well. This is needed as when the child
device is released the parent may also be released.
Closes #924
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Thu, 14 Jan 2021 19:59:56 +0000 (20:59 +0100)]
darwin_set_interface_altsetting: Avoid device reset on pipe error
If the device doesn't support the SET_INTERFACE request (which is
allowed if there are no alternate interfaces) it should stall the pipe.
A successful device request would clear all endpoints of the interface.
We therefore plaster over the failed request by clearing the endpoints
and returning success, effectively replacing the missing device request
support.
Up to now this was done by resetting the device, but this would affect
other interfaces as well. So only clear the endpoints of the selected
interface.
Only do this in the case the request tried setting the already
selected setting. Otherwise return the pipe error to the caller.
For any other error than pipe error return with failure.
Note that we need to reclaim the interface after the pipe error.
This is maybe papering over a bug in libusb or darwin: The interface
becomes inaccessible after the failed SetAlternateInterface() call. The
reclaim recovers from it, without device side-effects. The need to
recover was maybe the reason for the original reset "hammer".
Closes #852
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Tormod Volden [Thu, 25 Mar 2021 22:47:27 +0000 (23:47 +0100)]
winusb: Check for ClassGuid changes on re-enumeration
Previously only the dev_id (VID/PID/serial) was checked, but if the
bcdDevice has changed, Windows will have re-enumerated the device and
possibly attributed another ClassGuid to it.
Fixes issue #897.
Closes #900
Thanks to Craig Hutchinson for reporting, suggestion and testing.
Signed-off-by: Tormod Volden <debian.tormod@gmail.com>
Signed-off-by: Nathan Hjelm <hjelmn@google.com>
Nathan Hjelm [Fri, 11 Jun 2021 03:02:05 +0000 (21:02 -0600)]
core: only increment devices_released when a device is destroyed
Fix a typo in the last commit to ensure the loop does exit when there is
a dangling reference.
Signed-off-by: Nathan Hjelm <hjelmn@google.com>