platform/upstream/libevdev.git
10 years agotest: unconditionally distribute the valgrind suppressions
Peter Hutterer [Wed, 7 May 2014 08:55:02 +0000 (18:55 +1000)]
test: unconditionally distribute the valgrind suppressions

Reported-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoIgnore NULL as argument in libevdev_uinput_destroy()
Peter Hutterer [Thu, 1 May 2014 21:53:24 +0000 (07:53 +1000)]
Ignore NULL as argument in libevdev_uinput_destroy()

Triggered by the tests when run as non-root. Simply ignore any attempt to
destroy a NULL device, which also matches the behaviour of libevdev_free().

Reported-by: Andreas Radke <a.radke@arcor.de>
Reviewed-by: Daniel Martin <consume.noise@gmail.com>
Tested-by: Daniel Martin <consume.noise@gmail.com>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agolibevdev 1.2 libevdev-1.2
Peter Hutterer [Wed, 30 Apr 2014 05:08:25 +0000 (15:08 +1000)]
libevdev 1.2

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: run all tests through valgrind
Peter Hutterer [Thu, 24 Apr 2014 01:27:14 +0000 (11:27 +1000)]
test: run all tests through valgrind

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agotest: disable parallel testing
Peter Hutterer [Thu, 24 Apr 2014 01:14:34 +0000 (11:14 +1000)]
test: disable parallel testing

The tests create devices on the host system, avoid running them in parallel to
avoid interference between the test devices.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agotest: silence some valgrind warnings
Peter Hutterer [Thu, 24 Apr 2014 01:08:18 +0000 (11:08 +1000)]
test: silence some valgrind warnings

ioctl points to uninitialized bytes - correct but we didn't use those anyway.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agolibevdev 1.2RC1 libevdev-1.1.99.1
Peter Hutterer [Wed, 23 Apr 2014 22:24:06 +0000 (08:24 +1000)]
libevdev 1.2RC1

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDrain all events before synchronizing after SYN_DROPPED
Peter Hutterer [Mon, 7 Apr 2014 05:16:28 +0000 (15:16 +1000)]
Drain all events before synchronizing after SYN_DROPPED

The kernel ring buffer drops all events on SYN_DROPPED, but then continues to
fill up again. So by the time we read the events, the kernel's client buffer is
essentially like this:
  SYN_DROPPED, ev1, ev2, ev3, ...., evN

The kernel's device state represents the device after evN, and that is what
the ioctls return. For EV_KEY, EV_SND, EV_LED and EV_SW the kernel removes
potential duplicates from the client buffer [1], it doesn't do so for EV_ABS.

So we can't actually sync while there are events on the wire because the
events represent an earlier state. So simply discard all events in the kernel
buffer, synchronize, and then start processing again. We lose some granularity
but at least the events are correct.

[1] http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/drivers/input/evdev.c?id=483180281f0ac60d1138710eb21f4b9961901294

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoMove read_more_events() up in the file
Peter Hutterer [Mon, 7 Apr 2014 05:01:44 +0000 (15:01 +1000)]
Move read_more_events() up in the file

No functional change

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoActually make the min queue size a minimum
Peter Hutterer [Mon, 7 Apr 2014 05:00:24 +0000 (15:00 +1000)]
Actually make the min queue size a minimum

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoClarify what signal-safe means
Peter Hutterer [Thu, 10 Apr 2014 01:00:24 +0000 (11:00 +1000)]
Clarify what signal-safe means

Just in case...

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoCalling libevdev_enable_event_code() overwrites EV_ABS/EV_REP values
Peter Hutterer [Fri, 4 Apr 2014 04:11:47 +0000 (14:11 +1000)]
Calling libevdev_enable_event_code() overwrites EV_ABS/EV_REP values

Document this behaviour

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: restore default log priority after tests
Peter Hutterer [Fri, 4 Apr 2014 00:04:56 +0000 (10:04 +1000)]
test: restore default log priority after tests

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoSplit the SYN_DROPPED page into a few sections
Peter Hutterer [Mon, 7 Apr 2014 04:26:31 +0000 (14:26 +1000)]
Split the SYN_DROPPED page into a few sections

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoFix a typo, add a @ref in the documentation
Peter Hutterer [Mon, 7 Apr 2014 04:21:27 +0000 (14:21 +1000)]
Fix a typo, add a @ref in the documentation

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDrop invalid ABS_MT_TRACKING_ID changes
Peter Hutterer [Tue, 1 Apr 2014 07:01:39 +0000 (17:01 +1000)]
Drop invalid ABS_MT_TRACKING_ID changes

Follow-up to
commit 41334b5b40cd5456f5f584b55d8888aaafa1f26e
Author: Peter Hutterer <peter.hutterer@who-t.net>
Date:   Thu Mar 6 11:54:00 2014 +1000

    If the tracking ID changes during SYN_DROPPED, terminate the touch first

In normal mode, we may get double tracking ID events in the same slot, but
only if we either have a user-generated event sequence (uinput) or a malicious
device that tries to send data on a slot > dev->num_slots.
Since the client is unlikely to be able to handle these events, discard the
ABS_MT_TRACKING_ID completely. This is a bug somewhere in the stack, so
complain and hobble on along.

Note: the kernel doesn't allow that, but we cap to num_slots anyway, see
66fee1bec4c4b021e1b54adcd775cf6e2aa84869.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoMove slot-related initialization down
Peter Hutterer [Tue, 1 Apr 2014 05:40:09 +0000 (15:40 +1000)]
Move slot-related initialization down

No real effects, but improves readability

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoDrop hardcoded MAX_SLOTS in favour of pre-allocated memory
Peter Hutterer [Tue, 1 Apr 2014 05:32:46 +0000 (15:32 +1000)]
Drop hardcoded MAX_SLOTS in favour of pre-allocated memory

We can't allocate in sync_mt_state since it may be called in the signal
handler. So pre-allocate based on the device's number of slots, store that in
the libevdev struct and use it for the sync process.

This fixes a remaining bug with the handling of ABS_MT_TRACKING_ID. If a
device had > MAX_SLOTS and a slot above that limit would start or stop during
a SYN_DROPPED event, the slot would not be synced, and a subsequent touch in
that slot may double-terminate or double-open a touchpoint in the client.
For the effects of that see

commit 41334b5b40cd5456f5f584b55d8888aaafa1f26e
Author: Peter Hutterer <peter.hutterer@who-t.net>
Date:   Thu Mar 6 11:54:00 2014 +1000

    If the tracking ID changes during SYN_DROPPED, terminate the touch first

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoOnly sync the initial MT state for Protocol B devices
Peter Hutterer [Tue, 1 Apr 2014 05:12:56 +0000 (15:12 +1000)]
Only sync the initial MT state for Protocol B devices

For protocol A devices we won't get the information from the kernel anyway and
we expect all axes to be updated in the next event.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoDrop unnecessary memset
Peter Hutterer [Wed, 2 Apr 2014 01:57:41 +0000 (11:57 +1000)]
Drop unnecessary memset

The EVICOCGMTSLOTS ioctl returns all slot values for the requested code or an
error code, it doesn't return the number of bytes successfully transferred.
Thus all values in the input array are always defined (on success), we don't
need to memset it.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoUse libevdev_reset from libevdev_free
Peter Hutterer [Wed, 2 Apr 2014 01:48:14 +0000 (11:48 +1000)]
Use libevdev_reset from libevdev_free

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoAdd quote around device name during error message
Peter Hutterer [Tue, 1 Apr 2014 06:59:46 +0000 (16:59 +1000)]
Add quote around device name during error message

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: remove checks for test device creation failure
Peter Hutterer [Wed, 2 Apr 2014 01:12:42 +0000 (11:12 +1000)]
test: remove checks for test device creation failure

test_create_... fails on a non-zero return code anyway

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: the kernel test build doesn't need to link to libevdev
Peter Hutterer [Wed, 2 Apr 2014 01:24:27 +0000 (11:24 +1000)]
test: the kernel test build doesn't need to link to libevdev

We statically compile libevdev in anyway through $(common_sources)

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: fix kernel test for revoke after
Peter Hutterer [Wed, 2 Apr 2014 01:31:59 +0000 (11:31 +1000)]
test: fix kernel test for revoke after

dev2 by definition doesn't initialize, we expect it to fail. Freeing it after
is a bad idea. Also initialize it to NULL so this is a bit more obvious now.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: fix the tracking_id sync test
Peter Hutterer [Tue, 1 Apr 2014 06:31:41 +0000 (16:31 +1000)]
test: fix the tracking_id sync test

A max of num_slots -1 caused the first MT_SLOT event to be skipped, leading to
wrong tracking IDs in the slots.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoFix indentation
Peter Hutterer [Tue, 1 Apr 2014 04:02:41 +0000 (14:02 +1000)]
Fix indentation

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agolibevdev 1.1 libevdev-1.1
Peter Hutterer [Tue, 25 Mar 2014 06:20:24 +0000 (16:20 +1000)]
libevdev 1.1

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoAdd documentation for axis updates without a tracking ID
Peter Hutterer [Wed, 19 Mar 2014 04:04:59 +0000 (14:04 +1000)]
Add documentation for axis updates without a tracking ID

If a touch starts and terminates while in SYN_DROPPED, the tracking ID appears
to stay at -1, but the other axes may update. We need to pass these on to the
client since the kernel may buffer the next event with the same value
otherwise. Note this in the documentation so that client's don't create touch
points based on out-of-touchpoint updates.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agolibevdev 1.1RC2 libevdev-1.0.99.2
Peter Hutterer [Wed, 19 Mar 2014 00:30:02 +0000 (10:30 +1000)]
libevdev 1.1RC2

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoBump library revision
Peter Hutterer [Wed, 19 Mar 2014 00:29:51 +0000 (10:29 +1000)]
Bump library revision

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agouinput: check for asprintf failure
Peter Hutterer [Tue, 18 Mar 2014 03:55:21 +0000 (13:55 +1000)]
uinput: check for asprintf failure

This doesn't really do much here, but strictly speaking: if asprintf returns
-1, devnode is undefined. So reset it to NULL to avoid weird pointers. And
also free the rest of the names if we ever have more than one device - which
also shouldn't happen.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agotest: shut up compiler warning
Peter Hutterer [Tue, 18 Mar 2014 03:58:18 +0000 (13:58 +1000)]
test: shut up compiler warning

test-libevdev-events.c: In function ‘test_double_syn_dropped_event’:
test-libevdev-events.c:187:2: warning: ignoring return value of ‘read’,
declared with attribute warn_unused_result [-Wunused-result]

This read was there to drain events even when there shouldn't be any on the
pipe anyway. So let's add an assert.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoShut up another warning
Peter Hutterer [Tue, 18 Mar 2014 04:00:07 +0000 (14:00 +1000)]
Shut up another warning

Triggered with -O3

../libevdev/libevdev.c: In function ‘libevdev_get_event_value’:
../libevdev/libevdev.c:1112:6: warning: ‘value’ may be used uninitialized in
this function [-Wmaybe-uninitialized]

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoShut up compiler warning
Peter Hutterer [Tue, 18 Mar 2014 03:44:39 +0000 (13:44 +1000)]
Shut up compiler warning

Triggered with -O

../libevdev/libevdev.c: In function ‘libevdev_has_event_code’:
../libevdev/libevdev-util.h:58:20: warning: ‘mask’ may be used uninitialized
in this function [-Wmaybe-uninitialized]

Optimisation clearly shuffles things around here: in the code, if no max is
found, we return -1 and bail out before we access mask.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoSanitize events when we're skipping sync as well
Peter Hutterer [Mon, 17 Mar 2014 06:08:36 +0000 (16:08 +1000)]
Sanitize events when we're skipping sync as well

If a client doesn't sync expliciltly, make sure we sanitize the events when we
update the internal library state.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoadd a linebreak for correct rendering of the documentation
Peter Hutterer [Tue, 18 Mar 2014 05:11:39 +0000 (15:11 +1000)]
add a linebreak for correct rendering of the documentation

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agodoc: move the section for the internal test suite down
Peter Hutterer [Tue, 11 Mar 2014 00:57:15 +0000 (10:57 +1000)]
doc: move the section for the internal test suite down

"Related Pages" in doxygen are ordered in the order they appear in the source
file. The internal test suite is least likely to be of interest to the
reader, so move it to the bottom.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agodoc: include linux/input.h verbatim in the documentation
Peter Hutterer [Tue, 11 Mar 2014 00:56:10 +0000 (10:56 +1000)]
doc: include linux/input.h verbatim in the documentation

This makes it easier to look up what specific version libevdev was compiled
against.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agodoc: Add documentation on how SYN_DROPPED is handled
Peter Hutterer [Tue, 11 Mar 2014 00:44:10 +0000 (10:44 +1000)]
doc: Add documentation on how SYN_DROPPED is handled

This is getting a bit complex, so add some high-level documentation that we at
least know what we're trying to do.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoconfigure: add link to our wiki page
Peter Hutterer [Mon, 10 Mar 2014 23:19:18 +0000 (09:19 +1000)]
configure: add link to our wiki page

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agolibevdev 1.1RC1 libevdev-1.0.99.1
Peter Hutterer [Mon, 10 Mar 2014 22:21:49 +0000 (08:21 +1000)]
libevdev 1.1RC1

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: fix distcheck
Peter Hutterer [Mon, 10 Mar 2014 22:30:29 +0000 (08:30 +1000)]
test: fix distcheck

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoAdd a documentation note to update the slot after syncing
Peter Hutterer [Thu, 6 Mar 2014 02:38:47 +0000 (12:38 +1000)]
Add a documentation note to update the slot after syncing

Clients may not care about the events generated during SYN_DROPPED, but the
current slot must be updated to avoid a client being out-of-date.
Same with tracking IDs, if they changed, the caller will likely have to update
some internal states.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoIf the tracking ID changes during SYN_DROPPED, terminate the touch first
Peter Hutterer [Thu, 6 Mar 2014 01:54:00 +0000 (11:54 +1000)]
If the tracking ID changes during SYN_DROPPED, terminate the touch first

Most clients can't deal with tracking ID changes unless a -1 is sent first. So
if we notice that the tracking ID has changed during the sync process, send a
set of ABS_MT_TRACKING_ID -1 events for each of those, then send the rest of
the events.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoSend an extra ABS_MT_SLOT event to sync the client up with the current slot
Benjamin Tissoires [Wed, 5 Mar 2014 02:03:57 +0000 (12:03 +1000)]
Send an extra ABS_MT_SLOT event to sync the client up with the current slot

If multiple slots have changed during the sync handling, the client must be
re-set to the current slot before continuing with normal events.

Signed-off-by: Benjamin Tissoires <btissoir@redhat.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoDon't sync the MT state for fake MT devices
Peter Hutterer [Wed, 5 Mar 2014 03:15:20 +0000 (13:15 +1000)]
Don't sync the MT state for fake MT devices

Devices with ABS_MT_SLOT-1 are fake MT devices, they merely overlap the
axis range but don't actually provide slots. The EVIOCGABS ioctl won't work to
retrieve the current value - the kernel does not store values for those axes
and the return value is always 0.

Thus, simply ignore those axes for fake MT devices and instead rely on the
next event to update the caller with the correct state for each axis.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoMove some functions/macros between libevdev-int.h and libevdev-util.h
Peter Hutterer [Thu, 6 Mar 2014 00:22:20 +0000 (10:22 +1000)]
Move some functions/macros between libevdev-int.h and libevdev-util.h

This allows libevdev-util.h to be used by tests, it no longer relies on
libevdev internal structs.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoCap slot values to the announced maximum
Peter Hutterer [Wed, 5 Mar 2014 00:57:01 +0000 (10:57 +1000)]
Cap slot values to the announced maximum

A malicious device may announce N slots but then send a slot index >= N. The
slot state is almost always allocated (definitely the case in libevdev and
true for most callers), so providing a slot number higher than the announced
maximum is likely to lead to invalid dereferences. Don't allow that.
Likewise, don't allow negative slot numbers.

Note that the kernel filters these events anyway, the only way to trigger this
is to change the device fd to something outside the kernel's control.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agotest: add two more checks for the current slot
Peter Hutterer [Thu, 27 Feb 2014 06:52:51 +0000 (16:52 +1000)]
test: add two more checks for the current slot

When syncing, we expect the slot to stay the same until the client has
processed the events. This already worked, just add a check to make sure.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoIncrease MAX_SLOTS to 60
Peter Hutterer [Thu, 27 Feb 2014 04:35:35 +0000 (14:35 +1000)]
Increase MAX_SLOTS to 60

As seen on 3M devices, which seems to be the maximum seen so far. Some Stantum
devices report 255 touches but are only capable of 10, so the are not affected
by our limits.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Acked-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agotest: add a test for > MAX_SLOTS mt values
Peter Hutterer [Thu, 27 Feb 2014 04:27:59 +0000 (14:27 +1000)]
test: add a test for > MAX_SLOTS mt values

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoReduce memory requirement for MT syncing
Peter Hutterer [Thu, 27 Feb 2014 03:10:35 +0000 (13:10 +1000)]
Reduce memory requirement for MT syncing

Changes the algorithm: before we'd ioctl all axes for all slots, then generate
events for all slots one-by-one.

Now we ioctl the slot state for each axis, copy the new event value into
the device and mark a bitfield that we've updated the value. Then loop through
the slots and generate events where changed.

Side-effect: this makes it easy to check if anything in the slot has updated,
so we can skip empty slot events during sync.

Min memory requirement for the state storage was:
  MAX_SLOTS  * (ABS_MT_CNT + 1) * sizeof(int) = 1980
Min memory requirement now:
  (ABS_MT_CNT + 1) * sizeof(int) + NLONGS((MAX_SLOTS * ABS_MT_CNT) bits) = 544

This is sigsafe code, so this was stack memory. Reducing the requirement
allows us to up MAX_SLOTS in the future if we need to.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoDynamically allocate the slot values
Peter Hutterer [Thu, 27 Feb 2014 01:34:47 +0000 (11:34 +1000)]
Dynamically allocate the slot values

Instead of relying on a static MAX_SLOTS array, allocated it based on the
number of slots we have on the device. The previous checks for MAX_SLOTS were
incomplete, causing out-of-bound reads.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDon't sync past MAX_SLOTS slots
Peter Hutterer [Thu, 27 Feb 2014 01:36:04 +0000 (11:36 +1000)]
Don't sync past MAX_SLOTS slots

If a device has more than MAX_SLOTS slots, we'd run out-of-bounds on the sync
array. This function is sig-safe, so we can't alloc here, merely limit the
access.

Reported-by: Jonas Ådahl <jadahl@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoAdd unlikely() macro
Peter Hutterer [Tue, 4 Mar 2014 06:59:50 +0000 (16:59 +1000)]
Add unlikely() macro

Taken from systemd.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
10 years agoFix a wrong reference in the libevdev_next_event documentation
Peter Hutterer [Tue, 4 Mar 2014 03:31:22 +0000 (13:31 +1000)]
Fix a wrong reference in the libevdev_next_event documentation

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDocument the FORCE_SYNC behavior a bit better
Peter Hutterer [Tue, 4 Mar 2014 03:29:27 +0000 (13:29 +1000)]
Document the FORCE_SYNC behavior a bit better

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: whitespace fixes (newlines)
Peter Hutterer [Wed, 5 Mar 2014 02:57:18 +0000 (12:57 +1000)]
test: whitespace fixes (newlines)

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoFix memory leaks when failing to create a uinput device
Peter Hutterer [Thu, 27 Feb 2014 01:29:19 +0000 (11:29 +1000)]
Fix memory leaks when failing to create a uinput device

For an invalid fd, or a failure to open the device, the pre-allocated uinput
device struct would leak.

We can drop the open_uinput() function now, since skipping to the error
handling means we'll return -errno anyway.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoPlug a memory leak for name, uniq, phys
Peter Hutterer [Thu, 27 Feb 2014 01:22:27 +0000 (11:22 +1000)]
Plug a memory leak for name, uniq, phys

If a device is assigned a name, uniq and/or phys before calling
libevdev_set_fd(), those values would leak.

Change the default alloc to calloc, so name, uniq, and phys are initialized to
zero before we call libevdev_reset

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: plug more memory leaks
Peter Hutterer [Thu, 27 Feb 2014 01:13:53 +0000 (11:13 +1000)]
test: plug more memory leaks

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: fix memleak when using the test-uinput device
Peter Hutterer [Thu, 27 Feb 2014 01:08:51 +0000 (11:08 +1000)]
test: fix memleak when using the test-uinput device

The libevdev-created uinput device was never freed.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotools: Fix compiler warning
Emmanuele Bassi [Tue, 25 Feb 2014 10:48:49 +0000 (10:48 +0000)]
tools: Fix compiler warning

Signed-off-by: Emmanuele Bassi <ebassi@gnome.org>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoFix the include directives when building in tools/
Emmanuele Bassi [Tue, 25 Feb 2014 10:48:48 +0000 (10:48 +0000)]
Fix the include directives when building in tools/

We need to add top_srcdir to fix the build when srcdir != builddir.

Signed-off-by: Emmanuele Bassi <ebassi@gnome.org>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoAdd a test for EVIOCREVOKE
Peter Hutterer [Tue, 8 Oct 2013 05:43:50 +0000 (15:43 +1000)]
Add a test for EVIOCREVOKE

New in 3.12, EVIOCREVOKE revokes access to an evdev device. This is unlikely
to be used by a libevdev user, see.
http://lists.freedesktop.org/archives/input-tools/2014-January/000688.html

This patch adds a new test-kernel binary that tests the kernel API directly.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
10 years agoDocument that the return value of get_name|uniq|phys is not permanent
Peter Hutterer [Mon, 24 Feb 2014 00:55:53 +0000 (10:55 +1000)]
Document that the return value of get_name|uniq|phys is not permanent

It's a pointer to the copy libevdev has, but that copy may be freed by the
caller if someone calls e.g. libevdev_set_name()

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotools: add a tool to print the actual range provided by a touchpad
Peter Hutterer [Thu, 20 Feb 2014 04:40:32 +0000 (14:40 +1000)]
tools: add a tool to print the actual range provided by a touchpad

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoReplace hardcoded 0 with the enum for it
Peter Hutterer [Mon, 24 Feb 2014 23:44:35 +0000 (09:44 +1000)]
Replace hardcoded 0 with the enum for it

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: remove a bunch of superfluous empty lines
Peter Hutterer [Tue, 25 Feb 2014 00:06:21 +0000 (10:06 +1000)]
test: remove a bunch of superfluous empty lines

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agolibevdev 1.0 libevdev-1.0
Peter Hutterer [Tue, 18 Feb 2014 04:46:40 +0000 (14:46 +1000)]
libevdev 1.0

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoActually filter by log priority
Peter Hutterer [Wed, 12 Feb 2014 04:54:13 +0000 (14:54 +1000)]
Actually filter by log priority

If a message is higher than the current priority, filter it. And add a few
tests that the priority is handled the way it should.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
10 years agoUse 0 instead of "zero"
Peter Hutterer [Wed, 12 Feb 2014 05:44:33 +0000 (15:44 +1000)]
Use 0 instead of "zero"

We're a bit inconsistent here anyway because of the ample use of "non-zero"
which can't really be converted to non-0. But let's at least be consistent
with the use of 0 instead of zero.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoMore documentation fixes and improvements
Peter Hutterer [Wed, 12 Feb 2014 05:28:34 +0000 (15:28 +1000)]
More documentation fixes and improvements

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoMark a few more functions as signal-safe
Peter Hutterer [Wed, 12 Feb 2014 05:21:42 +0000 (15:21 +1000)]
Mark a few more functions as signal-safe

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoImprove the documentation a bit
Peter Hutterer [Tue, 11 Feb 2014 06:31:35 +0000 (16:31 +1000)]
Improve the documentation a bit

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDocument that we don't support EVIOCREVOKE
Peter Hutterer [Tue, 11 Feb 2014 06:37:14 +0000 (16:37 +1000)]
Document that we don't support EVIOCREVOKE

This is intentional, see
http://lists.freedesktop.org/archives/input-tools/2014-January/000688.html

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoREADME: remove the warning, it's stable now
Peter Hutterer [Tue, 11 Feb 2014 06:27:50 +0000 (16:27 +1000)]
README: remove the warning, it's stable now

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agolibevdev 0.9.1 libevdev-0.9.1
Peter Hutterer [Wed, 29 Jan 2014 05:09:05 +0000 (15:09 +1000)]
libevdev 0.9.1

libevdev 1.0-rc1

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: remove NULL test for SW_MAX, 3.13 defines SW_MUTE_DEVICE
Peter Hutterer [Wed, 29 Jan 2014 05:08:12 +0000 (15:08 +1000)]
test: remove NULL test for SW_MAX, 3.13 defines SW_MUTE_DEVICE

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoBump soname
Peter Hutterer [Thu, 2 Jan 2014 23:44:13 +0000 (09:44 +1000)]
Bump soname

Bump once to the stable interface we have now. This isn't really required as
we're compatible with everything but functions deprecated with 0.4 but we
might as well do it now to force a rebuild of everyone. Should've done that
for 0.6 when we actually dropped the function, but ...

Now that we have proper library versioning we shouldn't need any
backwards-incompatible soname bumps for a while.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
10 years agoAdd ld version-script
David Herrmann [Sat, 17 Aug 2013 11:58:15 +0000 (13:58 +0200)]
Add ld version-script

Explicit symbol versioning allows us to provide multiple versions of
incompatible API changes. It is a very common practice in GNU world and
avoids the problems occuring if distributions try to ship multiple version
of a single DSO.

Background information available at:
  http://people.redhat.com/drepper/dsohowto.pdf

To see the symbol-versions, use objdump:
  objdump -T libevdev/.libs/libevdev.so.1.0.0
This can also be used to verify that all symbols are correctly
exported (typos in linker-scripts are silently ignored by ld).

Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Updated for evdev 0.6's API.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoinclude: update linux/input.h for kernel 3.13
Peter Hutterer [Tue, 21 Jan 2014 22:27:31 +0000 (08:27 +1000)]
include: update linux/input.h for kernel 3.13

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoWarn about a SYN_DROPPED right after finishing a sync
Peter Hutterer [Fri, 17 Jan 2014 00:15:33 +0000 (10:15 +1000)]
Warn about a SYN_DROPPED right after finishing a sync

If the first event after a completed device sync is a SYN_DROPPED, warn the
user that they're not fast enough handling this device.

The test for this is rather complicated since we can't write SYN_DROPPED
through uinput so we have to juggle the device fd and a pipe and switch
between the two at the right time (taking into account that libevdev will read
events from the fd whenever it can).

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: abort on unexpected log messages
Peter Hutterer [Fri, 17 Jan 2014 00:21:48 +0000 (10:21 +1000)]
test: abort on unexpected log messages

Add two log functions, one that aborts on a received message. We know when we
expect to receive an error, so anytime this happens unexpectedly should
terminate the test.

And for those tests do issue a log message, let them ignore it and don't
print anything.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoCount the number of events needed for a full sync
Peter Hutterer [Thu, 16 Jan 2014 23:24:20 +0000 (09:24 +1000)]
Count the number of events needed for a full sync

Make sure we have a queue that is at least large enough to do a full sync
after a SYN_DROPPED, plus store a few extra events in case some came in after
the sync.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
10 years agotest: fix typo
Peter Hutterer [Thu, 16 Jan 2014 23:47:39 +0000 (09:47 +1000)]
test: fix typo

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: rename a test to match what it's testing better
Peter Hutterer [Thu, 16 Jan 2014 23:46:09 +0000 (09:46 +1000)]
test: rename a test to match what it's testing better

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDrop the argparse requirement for make-event-names.py
Peter Hutterer [Tue, 14 Jan 2014 04:18:06 +0000 (14:18 +1000)]
Drop the argparse requirement for make-event-names.py

We only take one option (path to linux/input.h) anyway so drop the argparse
requirement. This way libevdev builds on RHEL6 too which doesn't ship
argparse.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoDrop --output flag and python generation code from make-event-names.py
Peter Hutterer [Tue, 14 Jan 2014 04:08:29 +0000 (14:08 +1000)]
Drop --output flag and python generation code from make-event-names.py

This is a leftover from when the file was in evemu and used to generate python
headers too.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: Benjamin Tissoires <benjamin.tissoires@gmail.com>
10 years agoinclude: update linux/input.h for kernel 3.12
Peter Hutterer [Wed, 8 Jan 2014 00:52:31 +0000 (10:52 +1000)]
include: update linux/input.h for kernel 3.12

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>
10 years agotest: detect if we're running inside gdb and disable forking
Peter Hutterer [Mon, 23 Dec 2013 22:19:16 +0000 (08:19 +1000)]
test: detect if we're running inside gdb and disable forking

The Check test framework forks by default which is annoying when running gdb.
Try to detect whether we're inside gdb by ptracing ourselves. If that works,
we're not inside a debugger. If it doesn't, then assume we're inside a
debugger and set CK_FORK to "no".

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDocumentation rewordings and additions
Peter Hutterer [Mon, 23 Dec 2013 23:12:13 +0000 (09:12 +1000)]
Documentation rewordings and additions

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDocumentation typo fixes
Peter Hutterer [Mon, 23 Dec 2013 23:11:55 +0000 (09:11 +1000)]
Documentation typo fixes

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoAdd more cross-references
Peter Hutterer [Mon, 23 Dec 2013 23:10:47 +0000 (09:10 +1000)]
Add more cross-references

doxygen doesn't detect enum cross-references automatically, prefix them with
@ref

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agodoc: remove initial * from example code lines
Peter Hutterer [Mon, 23 Dec 2013 22:07:00 +0000 (08:07 +1000)]
doc: remove initial * from example code lines

doxygen actually copies that over into the resulting output.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agolibevdev 0.6 libevdev-0.6
Peter Hutterer [Mon, 23 Dec 2013 07:38:07 +0000 (17:38 +1000)]
libevdev 0.6

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: add basic link test
Peter Hutterer [Thu, 12 Dec 2013 03:13:54 +0000 (13:13 +1000)]
test: add basic link test

This test doesn't do anything but compile and link against libevdev. It's a
simple protection to avoid linker errors. If we ever have libs we depend on
and they don't get resolved properly, this test should warn us in time.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agotest: split tests into run-time and build-time tests
Peter Hutterer [Thu, 12 Dec 2013 02:57:21 +0000 (12:57 +1000)]
test: split tests into run-time and build-time tests

The build-time tests don't do anything, we don't need to run them.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
10 years agoDon't treat devices with (ABS_MT_SLOT - 1) as multitouch devices
Peter Hutterer [Wed, 20 Nov 2013 01:58:08 +0000 (11:58 +1000)]
Don't treat devices with (ABS_MT_SLOT - 1) as multitouch devices

Some devices (PS3 sixaxis controller) merely have a bunch of axes, without the
semantic information that linux/input.h requires. For those, the ABS_MT range
may be merely another axis, not the special range that we need to treat it
with.

Use a simple heuristic: if ABS_MT_SLOT - 1 is enabled, don't treat ABS_MT as
multitouch axes. The ABS_MT_SLOT - 1 axis is not used for a real axis.

Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reviewed-by: David Herrmann <dh.herrmann@gmail.com>