A manager is not connected to a device directly any more. It first connects
to an output, and then to the display. Update overlay and manager get_device ops
to return the device via the connected output.
Signed-off-by: Archit Taneja <archit@ti.com>
The display sysfs attribute's store function needs to be changed with the
introduction of outputs.
The DSS driver ensures that there is one display per output, and that a
registered omap_dss_device will have an output connected to it. The display
sysfs store function unsets the current output connected to the manager, and
sets it with the output connected to the new display. If the new display doesn't
have an output for some reason, we just bail out. The function doesn't set/unset
output->device links. These remain the same as when the omap_dss_device was
registered.
Signed-off-by: Archit Taneja <archit@ti.com>
To retrieve the manager pointer via a device, we need to now access it via the
output to which the device is connected. Make this change in omapfb_ioctl()
where the WAITFORVSYNC ioctl tries to access the manager's device.
Signed-off-by: Archit Taneja <archit@ti.com>
With addition of output entities, a device connects to an output, and an output
connects to overlay manager. Replace the dssdev->manager references with
dssdev->output->manager to access the manager correctly.
When enabling the HDMI output, check whether the output entity connected to
display is not NULL.
Signed-off-by: Archit Taneja <archit@ti.com>
With addition of output entities, a device connects to an output, and an output
connects to overlay manager. Replace the dssdev->manager references with
dssdev->output->manager to access the manager correctly.
When enabling the VENC output, check whether the output entity connected to
display is not NULL.
Signed-off-by: Archit Taneja <archit@ti.com>
With addition of output entities, a device connects to an output, and an output
connects to overlay manager. Replace the dssdev->manager references with
dssdev->output->manager to access the manager correctly.
When enabling the RFBI output, check whether the output entity connected to
display is not NULL.
Signed-off-by: Archit Taneja <archit@ti.com>
With addition of output entities, a device connects to an output, and an output
connects to overlay manager. Replace the dssdev->manager references with
dssdev->output->manager to access the manager correctly.
When enabling the SDI output, check whether the output entity connected to
display is not NULL.
Signed-off-by: Archit Taneja <archit@ti.com>
With addition of output entities, a device connects to an output, and an output
connects to overlay manager. Replace the dssdev->manager references with
dssdev->output->manager to access the manager correctly.
When enabling the DSI output, check whether the output entity connected to
display is not NULL.
In dsi_init_display(), the display won't be connected to the DSI output yet,
that happens later in dss_recheck_connections() in the panel driver's probe. Get
the dsidev platform device pointer using the DSI moudle number provided in the
omap_dss_device struct.
Signed-off-by: Archit Taneja <archit@ti.com>
dsi_pdev_map is a struct visible globally in the DSI driver to get the platform
device pointer of the DSI device corresponding to it's module ID. This was
required because there was no clean way to derive the platform device from
the DSI module instance number or from the connected panel.
With the new output entity, it is possible to retrieve the platform device
pointer if the omap_dss_output pointer is available. Modify the functions
dsi_get_dsidev_from_dssdev() dsi_get_dsidev_from_id() so that they use output
instead of dsi_pdev_map to retrieve the dsi platform device pointer.
Signed-off-by: Archit Taneja <archit@ti.com>
With addition of output entities, a device connects to an output, and an output
connects to overlay manager. Replace the dssdev->manager references with
dssdev->output->manager to access the manager correctly.
When enabling the DPI output, check whether the output entity connected to
display is not NULL.
Signed-off-by: Archit Taneja <archit@ti.com>
Links between DSS entities are made in dss_init_connections() when a panel
device is registered, and are removed in dss_uninit_connections() when the
device is unregistered. Modify these functions to incorporate the addition of
outputs.
The fields in omap_dss_device struct gives information on which output and
manager to connect to. The desired manager and output pointers are retrieved and
prepared to form the desired links. The output is linked to the device, and then
the manager to the output.
A helper function omapdss_get_output_from_device() is created to retrieve the
output from the display by checking it's type, and the module id in case of DSI.
Signed-off-by: Archit Taneja <archit@ti.com>
With the introduction of output entities, managers will now connect to outputs.
Use the helper op for managers named get_device. This will abstract away the
information on how to get the device from an overlay manager.
Using the helper function will reduce the number of pointer dereferences a user
of OMAPDSS needs to do and reduce risk of a NULL dereference.
Signed-off-by: Archit Taneja <archit@ti.com>
With the introduction of output entities, managers will now connect to outputs.
Use the helper op for overlays named get_device. This will abstract away the
information on how to get the device from an overlay.
Using the helper function will reduce the number of pointer dereferences a user
of OMAPDSS needs to do and reduce risk of a NULL dereference.
Signed-off-by: Archit Taneja <archit@ti.com>
With the introduction of output entities, managers will now connect to outputs.
Use the helper op for overlays named get_device. This will abstract away the
information on how to get the omap_dss_device pointer from an overlay.
Using the helper function will reduce the number of pointer dereferences a user
of OMAPDSS needs to do and reduce risk of a NULL dereference.
Cc: Vaibhav Hiremath <hvaibhav@ti.com>
Signed-off-by: Archit Taneja <archit@ti.com>
With the introduction of output entities, managers will now connect to outputs.
Create helper ops for overlays and managers named get_device. This will abstract
away the information on how to get the device from an overlay or an overlay
manager. The get_device ops currently retrieve the output via a
ovl->manager->device reference. This will be later replaced by
ovl->manager->output->device references.
Signed-off-by: Archit Taneja <archit@ti.com>
Add set_output/unset_output ops for overlay managers, these form links between
managers and outputs. Create a function in dss features which tell all the
output instances that connect to a manager, use it when a manager tries to set
an output. Add a constraint of not unsetting an output when the manager is
enabled.
Keep the omap_dss_device pointer and set/unset_device ops in overlay_manager for
now to not break things. Keep the dss feature function get_supported_displays
as it's used in some places. These will be removed later.
Signed-off-by: Archit Taneja <archit@ti.com>
Cleanup the label maze in this function. Having a
seperate function to first handle the traps that don't
generate a signal makes it easier to convert into
more readable conditional paths.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Cc: Oleg Nesterov <oleg@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Peter Zijlstra <a.p.zijlstra@chello.nl>
Link: http://lkml.kernel.org/r/1348577479-2564-1-git-send-email-fweisbec@gmail.com
[ Fixed 32-bit build failure. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
GAS in binutils(2.16.91) could not parse parentheses within
macro parameters unless fully parenthesized, and this is a
workaround to make old gas work without generating below errors:
arch/x86/kernel/entry_64.S: Assembler messages:
arch/x86/kernel/entry_64.S:387: Error: too many positional arguments
arch/x86/kernel/entry_64.S:389: Error: too many positional arguments
[...]
Signed-off-by: Tao Guo <glorioustao@gmail.com>
Reluctantly-Acked-by: Jan Beulich <jbeulich@novell.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Andrew Morton <akpm@linux-foundation.org>
Link: http://lkml.kernel.org/r/1348648102-12653-1-git-send-email-glorioustao@gmail.com
[ Jan argues that these old GAS versions are fragile - which is so, but lets give them a chance. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Allow regulators managed via DAPM to make use of the bypass support that
has recently been added to the regulator API by setting a flag
SND_SOC_DAPM_REGULATOR_BYPASS. When this flag is set the regulator will
be put into bypass mode before being disabled, allowing the regulator to
fall into bypass mode if it can't be disabled due to other users.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Commit 31d2092eb0 ("x86: move
mp_register_lapic_address to boot.c") renamed mp_register_lapic
to acpi_register_lapic. But mp_register_lapic remains in a
comment. So the patch rename it.
Signed-off-by: Yasuaki Ishimatsu <isimatu.yasuaki@jp.fujitsu.com>
Cc: Len Brown <lenb@kernel.org>
Link: http://lkml.kernel.org/r/50625239.3050403@jp.fujitsu.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Allow regulators to be put into a non-regulating mode bypassing the
input straight to the output, mostly used by low power retention modes.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
iQIcBAABAgAGBQJQYuORAAoJEOoSHmUN5Tg4xRUQAL1OwYJsLfbwNEjeVnxDjpWU
La4HS3xLYZ/CHHSXJRJnithc/fqrjuTryEMUYVDaXW3MHt53Yc98ENJmnieurR4Y
LJ6lMI2+DKPmgiR9G2HlqOUPbqBRMytAQS7e8u+xLV4MFB+xSPO/y/ntHPfQCzS3
OqEU1xXkmxRMUPBugY5fgoI92tpYAmTJwP+EziiBV5mWjL85DA7oARGU1V3opgrh
EKJzMFdiajtCaTj8JTjMDaGLAd6+R8S1TrT+dEF85VWZgVCcMG+RY37z3oknWfHR
2yDC7jBRl9LiPaP/QvBegbauU862uQlZo0Kz3xDPdU3a1fV7ugpuIusuQCalh0nD
RHLaiqK4Y8xDvvh0P6WwBKY1HTTs6Y0vgN/OqYoLETnavkOV8LM4JnkoHb9BfeiP
LhGKBX5u/AiffDL30gVhy5adSXLJQBPyFYjVXCXZQB6EA4XsDxS+bVq31E0XXCY+
JsNOOu391sEsOPrtpnGDsztvG+AMKBWDe594xGUTWJKnlScW3vd3wEWS1JNw90TG
TkiA6GadKRADPV1R49iEJdKRcgqGnbY3+msP3rK9Pe/0gxF4QSfSwmvE9eWv+UFA
Jq0899StOIbovIC+IyY49CgxiewnwOeZO52dmU4gQ7EiXzq4y2wsGYX4SMd52B2q
dqTs4BzNw2B/kcrPw3Qb
=Fi2j
-----END PGP SIGNATURE-----
Merge tag 'bypass' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator into for-3.7
regulator: Bypass mode support
Allow regulators to be put into a non-regulating mode bypassing the
input straight to the output, mostly used by low power retention modes.
An output entity represented by the struct omap_dss_output connects to a
omap_dss_device entity. Add functions to set or unset an output's device. This
is similar to how managers and devices were connected previously. An output can
connect to a device without being connected to a manager. However, the output
needs to eventually connect to a manager so that the connected panel can be
enabled.
Keep the omap_overlay_manager pointer in omap_dss_device for now to prevent
breaking things. This will be removed later when outputs are supported
completely.
Signed-off-by: Archit Taneja <archit@ti.com>
Add output structs to output driver's private data. Register output instances by
having an init function in the probes of the platform device drivers for
different outputs. The *_init_output for each output registers the output and
fill up the output's plaform device, type and id fields. The *_uninit_output
functions unregister the output.
In the probe of each interface driver, the output entities are initialized
before the *_probe_pdata() functions intentionally. This is done to ensure that
the output entity is prepared before the panels connected to the output are
registered. We need the output entities to be ready because OMAPDSS will try
to make connections between overlays, managers, outputs and devices during the
panel's probe.
Signed-off-by: Archit Taneja <archit@ti.com>
The current OMAPDSS design contains 3 software entities: Overlays, Managers and
Devices. These map to pipelines, overlay managers and the panels respectively in
hardware. One or more overlays connect to a manager to represent a composition,
the manager connects to a device(generally a display) to display the content.
The part of DSS hardware which isn't represented by any of the above entities
are interfaces/outputs that connect to an overlay manager, i.e blocks like DSI,
HDMI, VENC and so on. Currently, an overlay manager directly connects to the
display, and the output to which it is actually connected is ignored. The panel
driver of the display is responsible of calling output specific functions to
configure the output.
Adding outputs as a new software entity gives us the following benefits:
- Have exact information on the possible connections between managers and
outputs: A manager can't connect to each and every output, there only limited
hardware links between a manager's video port and some of the outputs.
- Remove hacks related to connecting managers and devices: Currently, default
links between managers and devices are set in a not so clean way. Matching is
done via comparing the device type, and the display types supported by the
manager. This isn't sufficient to establish all the possible links between
managers, outputs and devices in hardware.
- Make panel drivers more generic: The DSS panel drivers currently call
interface/output specific functions to configure the hardware IP. When making
these calls, the driver isn't actually aware of the underlying output. The
output driver extracts information from the panel's omap_dss_device pointer
to figure out which interface it is connected to, and then configures the
corresponding output block. An example of this is when a DSI panel calls
dsi functions, the dsi driver figures out whether the panel is connected
to DSI1 or DSI2. This isn't correct, and having output as entities will
give the panel driver the exact information on which output to configure.
Having outputs also gives the opportunity to make panel drivers generic
across different platforms/SoCs, this is achieved as omap specific output
calls can be replaced by ops of a particular output type.
- Have more complex connections between managers, outputs and devices: OMAPDSS
currently doesn't support use cases like 2 outputs connect to a single
device. This can be achieved by extending properties of outputs to connect to
more managers or devices.
- Represent writeback as an output: The writeback pipeline fits well in OMAPDSS
as compared to overlays, managers or devices.
Add a new struct to represent outputs. An output struct holds pointers to the
manager and device structs to which it is connected. Add functions which can
register/unregister an output, or look for one. Create an enum which represent
each output instance.
Signed-off-by: Archit Taneja <archit@ti.com>
The functions dss_mgr_wait_for_go() and dss_mgr_wait_for_go_ovl() check if there
is an enabled display connected to the manager before trying to see the state of
the GO bit.
The checks related to the display can be replaced by checking the state of the
manager, i.e, whether the manager is enabled or not. This makes more sense than
checking with the connected display as the GO bit behaviour is more connected
with the manager state rather than the display state. A GO bit can only be set
if the manager is enabled. If a manager isn't enabled, we can safely assume that
the GO bit is not set.
Signed-off-by: Archit Taneja <archit@ti.com>
Many of the DSI functions receive the connected panel's omap_dss_device pointer
as an argument. The platform device pointer is then derived via omap_dss_device
pointers.
Most of these functions don't really require omap_dss_device pointer anymore
since we now keep copies of parameters in the driver data which were previously
available only via omap_dss_device. Replace the arguments with platform device
pointers for such functions.
Signed-off-by: Archit Taneja <archit@ti.com>
Use 'late_initcall()' in UBI to make sure it initializes after MTD drivers.
Signed-off-by: Jiang Lu <lu.jiang@windriver.com>
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com>
tlpx_half bit field in DSI_DSIPHY_CFG1 is [20,16], not [22,16] as
accessed in the code currently. Fix this.
The bug should not have caused any problems on OMAP3/4, as the bits
21,22 are unused. They are used on OMAP5, though.
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
Remove the printk.h include: this header is already via kernel.h,
so, there's no need to explicitly add it at ivtv-alsa-pcm.c.
Signed-off-by: Hans Verkuil <hans.verkuil@cisco.com>
Acked-by: Andy Walls <awalls@md.metrocast.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com>
Otherwise when X starts we commonly get a black screen scanning
out nothing, its wierd dpms on/off from userspace brings it back,
With this on F18, multi-seat works again with my 1920x1200 monitor
which is above the sku limit for the device I have.
Cc: stable@vger.kernel.org
Reviewed-by: Alex Deucher <alexander.deucher@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
We don't allocate enough data for this struct. As soon as we start
modifying event->event on the next lines, then we're going beyond the
end of the memory we allocated.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@gmail.com>
These just silence some printks that we are seeing that we shouldn't
* 'drm-nouveau-fixes' of git://anongit.freedesktop.org/git/nouveau/linux-2.6:
drm/nvc0/ltcg: mask off intr 0x10
drm/nouveau: silence a debug message triggered by newer userspace
blkdev_mmap() isn't used outside of fs/block_dev.c, mark it as
static.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Pull v3.7 RCU commits from Paul E. McKenney:
"
0. A fix for a latent bug that has been in RCU ever since the
addition of CPU stall warnings. This bug results in
false-positive stall warnings, but thus far only on embedded
systems with severely cut-down userspace configurations.
This fix is located on an rcu/urgent branch, with the rest
of the commits based on top of it. This commit CCs stable.
Given that the merge window is coming quite soon and given
the small number of affected users, I do -not- recommend
pushing it to 3.6, but the separate branch makes it easy to
find if someone needs it.
1. Further reductions in latency spikes for huge systems, along
with additional boot-time adaptation to the actual hardware.
This is a large change, as it moves RCU grace-period
initialization and cleanup, along with quiescent-state forcing,
from softirq to a kthread. However, it appears to be in
quite good shape (famous last words). Posted to LKML at
https://lkml.org/lkml/2012/9/20/427.
2. Updates to documentation and rcutorture, the latter category
including keeping statistics on CPU-hotplug latencies and
fixing some initialization-time races. Posted to LKML at
https://lkml.org/lkml/2012/8/30/193.
3. Miscellaneous fixes and improvements, posted to LKML at
https://lkml.org/lkml/2012/8/30/199.
4. CPU-hotplug fixes and improvements, posted to LKML at
https://lkml.org/lkml/2012/8/30/292 for first three and at
https://lkml.org/lkml/2012/8/3/416.
5. Idle-loop fixes that were omitted on an earlier submission,
posted to LKML at https://lkml.org/lkml/2012/8/30/251.
"
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Finishes commit 02d719562e
Author: Daniel Vetter <daniel.vetter@ffwll.ch>
Date: Thu Aug 9 16:44:54 2012 +0200
drm/i915: properly guard ilk ips state
The core functions were annotated with their locking requirements, but
we overlooked that they were exported, without any control over the
locking, to debugfs. So in order to enable debugfs to read the registers
without triggering sanity checks, we change the exported entry points to
properly take the required locks before calling the core routines.
Reported-by: yangguang <guang.a.yang@intel.com>
Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=55304
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
... even if the actual infoframe is smaller than the maximum possible
size.
If we don't write all the 32 DIP data bytes the InfoFrame ECC may not
be correctly calculated in some cases (e.g., when changing the port),
and this will lead to black screens on HDMI monitors. The ECC value is
generated by the hardware.
I don't see how this should break anything since we're writing 0 and
that should be the correct value, so this patch should be safe.
Notice that on IVB and older we actually have 64 bytes available for
VIDEO_DIP_DATA, but only bytes 0-31 actually store infoframe data: the
others are either read-only ECC values or marked as "reserved". On HSW
we only have 32 bytes, and the ECC value is stored on its own separate
read-only register. See BSpec.
This patch fixes bug #46761, which is marked as a regression
introduced by commit 4e89ee174b:
drm/i915: set the DIP port on ibx_write_infoframe
Before commit 4e89 we were just failing to send AVI infoframes when we
needed to change the port, which can lead to black screens in some
cases. After commit 4e89 we started sending infoframes, but with a
possibly wrong ECC value. After this patch I hope we start sending
correct infoframes.
Version 2:
- Improve commit message
- Try to make the code more clear
Cc: stable@vger.kernel.org
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=46761
Signed-off-by: Paulo Zanoni <paulo.r.zanoni@intel.com>
Reviewed-by: Rodrigo Vivi <rodrigo.vivi@gmail.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
. The new perf_evsel__tp_sched_test 'perf test' broke the build by setting the
'ret' variable but not using it, caught by newer gcc
-Werror=unused-but-set-variable, fix from Namhyung Kim.
. pevent_parse_event should return a proper PEVENT_ERRNO__ and call
pevent_free_format on its failure path, fixes from Namhyung Kim.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.14 (GNU/Linux)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=EqiG
-----END PGP SIGNATURE-----
Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux into perf/core
Pull perf/core fixes from Arnaldo Carvalho de Melo:
* The new perf_evsel__tp_sched_test 'perf test' broke the build by setting the
'ret' variable but not using it, caught by newer gcc
-Werror=unused-but-set-variable, fix from Namhyung Kim.
* pevent_parse_event should return a proper PEVENT_ERRNO__ and call
pevent_free_format on its failure path, fixes from Namhyung Kim.
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
This avoids cache line bouncing when many processes lock the semaphore
for read.
New percpu lock implementation
The lock consists of an array of percpu unsigned integers, a boolean
variable and a mutex.
When we take the lock for read, we enter rcu read section, check for a
"locked" variable. If it is false, we increase a percpu counter on the
current cpu and exit the rcu section. If "locked" is true, we exit the
rcu section, take the mutex and drop it (this waits until a writer
finished) and retry.
Unlocking for read just decreases percpu variable. Note that we can
unlock on a difference cpu than where we locked, in this case the
counter underflows. The sum of all percpu counters represents the number
of processes that hold the lock for read.
When we need to lock for write, we take the mutex, set "locked" variable
to true and synchronize rcu. Since RCU has been synchronized, no
processes can create new read locks. We wait until the sum of percpu
counters is zero - when it is, there are no readers in the critical
section.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The kernel may crash when block size is changed and I/O is issued
simultaneously.
Because some subsystems (udev or lvm) may read any block device anytime,
the bug actually puts any code that changes a block device size in
jeopardy.
The crash can be reproduced if you place "msleep(1000)" to
blkdev_get_blocks just before "bh->b_size = max_blocks <<
inode->i_blkbits;".
Then, run "dd if=/dev/ram0 of=/dev/null bs=4k count=1 iflag=direct"
While it is waiting in msleep, run "blockdev --setbsz 2048 /dev/ram0"
You get a BUG.
The direct and non-direct I/O is written with the assumption that block
size does not change. It doesn't seem practical to fix these crashes
one-by-one there may be many crash possibilities when block size changes
at a certain place and it is impossible to find them all and verify the
code.
This patch introduces a new rw-lock bd_block_size_semaphore. The lock is
taken for read during I/O. It is taken for write when changing block
size. Consequently, block size can't be changed while I/O is being
submitted.
For asynchronous I/O, the patch only prevents block size change while
the I/O is being submitted. The block size can change when the I/O is in
progress or when the I/O is being finished. This is acceptable because
there are no accesses to block size when asynchronous I/O is being
finished.
The patch prevents block size changing while the device is mapped with
mmap.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
When performing an online resize, we add a bunch of groups at one time
in ext4_flex_group_add, so in most cases a lot of group descriptors
will be in the same group block. But in the end of this function,
update_backups will be called for every group descriptor and the same
block will be copied and journalled again and again. It is really a
waste.
Fix things so we only update a particular bg descriptor block once and
skip subsequent updates of the same block.
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
bh_submit_read() is responsible for unlock bh on endio. In addition,
we need to use bh_uptodate_or_lock() to avoid races.
Signed-off-by: Dmitry Monakhov <dmonakhov@openvz.org>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
