Age | Commit message (Collapse) | Author | Files | Lines |
|
Detected by "cppcheck"
Please CC: comments.
Signed-off-by: Thomas Jarosch <thomas.jarosch@intra2net.com>
Signed-off-by: Danny Kukawka <danny.kukawka@web.de>
|
|
Don't close file descriptor twice. Adopted version of a patch from
Thomas Jarosch <thomas.jarosch@intra2net.com>.
|
|
Fixed typo from commit c648362a711.
|
|
Fixed some compiler warnings (strict-aliasing etc.).
|
|
|
|
The hald-probe-video4linux prober supports both v4l1 and v4l2. Support for v4l1
has been removed from Linux kernel 2.6.38. Instead of disabling the prober
altogether, #ifdef the v4l1 parts when building on a newer kernel.
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
|
|
The old v4l1 API is gone from current kernels (2.6.38), and thus
linux/videodev.h does not exist any more. Add a configure check for the include
and do not build hald-probe-video4linux if it is not available.
|
|
|
|
Fixed probing of optical disc volumes by ignoring raid.
Calling "blkid -p -u raid /dev/sr0" blocks the dvd-drive as the
same as hal does. Note that "blkid -p -u filesystem,crypto,other
/dev/sr0" or "blkid -p -u noraid /dev/sr0" don't hang and return
the detected filesystem.
This is a patch that filter out the raid detection on disk volumes
so hal can detect the filesystem and mount the disk.
fd.o#15335
|
|
Remove archives from 'make dist' (hal-*.tar.{bz2,gz}) on make clean.
|
|
Fixed generation of Changelog from git: replaced git-log with
correct commando 'git log'.
|
|
Fixed hal segfault. It crashes because strlen() gets called with a
NULL pointer (ppdev_compute_udi calls hal_util_get_last_element
with a NULL pointer).
bnc#556485
|
|
With new device-mapper udev rules are /dev/mapper/* symlinks
to basic device name /dev/dm-X.
(Change requested by udev upstream.)
This change breaks temporary-cryptsetup workaround inside hal.
With new dm-udev rules (uncluded since device-mapper 1.02.39)
there is DM_UDEV_DISABLE_OTHER_RULES_FLAG variable
which controls that scan should be ignored for this device
(it is set for all internal devices, including temporary cryptsetup,
internal parts of lvm devices etc.)
Ignore device if this flag is set.
See bugs
https://bugzilla.redhat.com/show_bug.cgi?id=613909
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=586286
Signed-off-by: Milan Broz <mbroz@redhat.com>
|
|
Fixed build with automake 1.11.1: whitespace following trailing backslash.
fd.o#29085
|
|
|
|
Commit 476dd36 was incomplete: we need to flush the fd before we sync it, to
really have an effect.
Thanks to Fujii Takafumi for pointing this out!
|
|
This is something I had to do in RHEL6, and so it made sense
to share with upstream, even though another HAL release might
not be likely.
|
|
Some file systems (reported on ubifs) are rather lazy with writing changes to
disk. An automount, followed by a power failure, then results in a stale
/media/.hal-mtab and also a stale mount point directory.
Use fsync() to make hal-mtab changes much more reliable.
Thanks to Fujii Takafumi for debugging this!
|
|
Minor fix for d5f236f3dad5e721f8c2105b49c963839f3ce549.
|
|
Detect also machines with serial (wacom?) tablets and pnp-id FUJ02e7/FUJ02e9
as TabletPCs. Reduce detection of Wacom PNP-id to 'WACf' to prevent future
changes.
|
|
Detect also machines with serial wacom tablets and pnp-id > WACf009
as e.g. WACf010 as TabletPCs.
|
|
Fixed hotplug support for e-SATA from commit dea5997df8966719
(bnc#576506 and bnc#590163).
|
|
Added the new system.lsb.* properties to the SPEC.
|
|
Added new prober to get the results of lsb_release. Parse
the output and add these new keys:
- system.lsb.version
- system.lsb.distributor_id
- system.lsb.description
- system.lsb.release
- system.lsb.codename
|
|
Added new key scsi.transport for SCSI devices to identify e.g.
FiberChannel, iSCSI or SAS. This allow to handle them in FDI
files (e.g. set volume.ignore or storage.automount_enabled_hint).
|
|
Added new scsi.transport property to SPEC.
|
|
Fixed storage.bus handling for SCSI. Prevent fall through to PCI as
storage bus. Current HAL marked SCSI storage devices which are not
usb or ieee1394 get marked with storage.bus=pci and not as scsi.
|
|
Fixed compiler warnings from cgcc.
|
|
Updated SPEC due to new properties from commit dea5997df896.
Added new:
- scsi.hotpluggable
- scsi_host.hotpluggable
|
|
Added *.orig to hald/linux/.gitignore
|
|
Add hotplug support for e-SATA (bnc#576506); hotpluggable flag
is set if ahci_cmd_port is exported in sysfs (available in the
recent kernel)
|
|
Fix drm_compute_udi() to prevent such udi's:
/org/freedesktop/Hal/devices/pci_1002_71c5_drm__null__card0
Use drm.dri_library only if the property is really set.
|
|
Changed event_io() to always use the same exit and cleanup.
|
|
Some BIOSe are broken and create infinite key press events for
the sleep key (FN-F5). So ignore all key repeats for the SLEEP
key. (bnc#525959)
|
|
|
|
The second argument of ioperm() is not the last port to be accessed
but rather length of the port range [port, port + len).
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
|
|
If dbus_error_init() is not called before LIBHAL_FREE_DBUS_ERROR() is
called on that error, then it is uninitialized and may segfault. This
means that dbus_error_init() must be called before any "goto out" if
"out:" calls LIBHAL_FREE_DBUS_ERROR().
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
|
|
Do not crash on unhandled rules (bnc#537452)
|
|
|
|
retval 1 for LSB compliance
|
|
Now that the property names for USB interfaces have been fixed, correct
the environment variable names.
|
|
USB interfaces in hal should have usb.X for properties, compared to
usb_device for the parent.
|
|
Use the correct macros for kFreeBSD.
|
|
Check for availability of libufs on (k)freebsd and compile the code
conditionally.
|
|
Link against -lrt as we need that for clock_gettime on kfreebsd.
AC_SEARCH_LIBS should handle the case nicely that on freebsd this
function is already part of libc.
|
|
This commit essentially duplicates the policy for:
<match key="info.capabilities" contains="input.mouse">
and then changes `input.mouse' to `input.touchpad'. This
is necessary because in Linus Torvalds's Linux repo:
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
the following commit:
commit 7105d2ea73e1391b681d0e1212c42f561c64d429
Author: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Date: Fri Dec 11 23:54:54 2009 -0800
Input: ALPS - do not set REL_X/REL_Y capabilities on the touchpad
Relative events are only reported via secondary device therefore device
associated with the touchpad should not advertise these capabilities.
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
made these few changes:
diff --git a/drivers/input/mouse/alps.c b/drivers/input/mouse/alps.c
index a3f492a..b03e7e0 100644
--- a/drivers/input/mouse/alps.c
+++ b/drivers/input/mouse/alps.c
@@ -487,6 +487,17 @@ int alps_init(struct psmouse *psmouse)
if (alps_hw_init(psmouse))
goto init_fail;
+ /*
+ * Undo part of setup done for us by psmouse core since touchpad
+ * is not a relative device.
+ */
+ __clear_bit(EV_REL, dev1->evbit);
+ __clear_bit(REL_X, dev1->relbit);
+ __clear_bit(REL_Y, dev1->relbit);
+
+ /*
+ * Now set up our capabilities.
+ */
dev1->evbit[BIT_WORD(EV_KEY)] |= BIT_MASK(EV_KEY);
dev1->keybit[BIT_WORD(BTN_TOUCH)] |= BIT_MASK(BTN_TOUCH);
dev1->keybit[BIT_WORD(BTN_TOOL_FINGER)] |= BIT_MASK(BTN_TOOL_FINGER);
so that HAL no longer adds:
input.mouse
to an ALPS touchpad's:
info.capabilities
so that HAL no longer marks the ALPS touchpad with:
input.x11_driver = 'evdev'
because the policy file:
fdi/policy/10osvendor/10-x11-input.fdi
doesn't define the policy for:
<match key="info.capabilities" contains="input.touchpad">
which was previous unnecessary because everything used to
be caught by the policy for:
<match key="info.capabilities" contains="input.mouse">
Signed-off-by: Michael Witten <mfwitten@gmail.com>
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
|
|
Some bits in the code use realpath() with destination paths of size
HAL_PATH_MAX. This potentially breaks on systems where PATH_MAX is bigger than
HAL_PATH_MAX (which was 512 until now).
Since we can't use PATH_MAX directly (due to platforms like GNU/Hurd, which
apparently don't have it), just bump HAL_PATH_MAX to 4096 (as PATH_MAX is on
Linux), to avoid potential buffer overflows and also fix hal on Linux when
enabling FORTIFY in gcc.
https://bugs.freedesktop.org/show_bug.cgi?id=25888
|
|
it may happen in the out: case that LIBHAL_FREE_DBUS_ERROR (&error) is
called before the error structure is initialized via
dbus_error_init (&error). This could lead to a segfault during startup
as seen in dmesg:
|Intel AES-NI instructions are not detected.
|padlock: VIA PadLock not detected.
|hald-probe-inpu[1793]: segfault at 2 ip 00007f656fb68969 sp 00007fff39eeb950 error 4 in libc-2.10.2.so[7f656faf3000+14a000]
|hald-probe-inpu[1796]: segfault at 2 ip 00007fa2c3293969 sp 00007fffd92a5dd0 error 4 in libc-2.10.2.so[7fa2c321e000+14a000]
|hald-probe-inpu[1797]: segfault at 2 ip 00007f1d08ba2969 sp 00007fff34244e30 error 4 in libc-2.10.2.so[7f1d08b2d000+14a000]
|hald-probe-inpu[1799]: segfault at 2 ip 00007f35c0e3d969 sp 00007fffe5ec7ee0 error 4 in libc-2.10.2.so[7f35c0dc8000+14a000]
|hald-probe-inpu[1800]: segfault at 2 ip 00007f931c556969 sp 00007fffe1825b60 error 4 in libc-2.10.2.so[7f931c4e1000+14a000]
|hald-probe-inpu[1801]: segfault at 2 ip 00007f5156d9a969 sp 00007fff4e620af0 error 4 in libc-2.10.2.so[7f5156d25000+14a000]
|RPC: Registered udp transport module.
|RPC: Registered tcp transport module.
What I run into seems to be reported as Debian #562068 [0]. This patch
makes the segfault go away on my machine.
[0] http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=562068
Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Signed-off-by: Martin Pitt <martin.pitt@ubuntu.com>
|
|
* Add support for file systems with spaces in their label names.
* Ignore ufsid labels outright. These labels are special since they are
created and destroyed based on whether or not the underlying device is
mounted or not. This should fix some weird remounting problems with UFS
volumes as well as some mount contention.
* Attempt to workaround the GEOM lock race by sleeping half of a second
when a DESTROY event is seen by the storage subsystem. Additionally, protect
against kern.geom.conftxt returning NULL. When this happens, all disks
would get removed, and that's bad.
|
|
* Fix support for the dynamic UFSID labels.
* Do not crash on an assertion due to a "dirty" DBusError object.
|