| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
| |
Change-Id: Ie04e3bcae4aa23a9c11f37347d110e61a141f9fb
Signed-off-by: Erik Gilling <konkers@android.com>
|
|
|
|
|
|
|
| |
DSSCOMP required PVR_SGX, so add a dependency in Kconfig.
Change-Id: Ieba8c475688bf6845e160b0eb4f5821eca1d8fd6
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Cropping logic was using the wrong variables for picture size.
Thus, cropping sometimes failed if crop region was outside the
incorrectly set picture size.
This is now fixed.
Change-Id: I0d4427d0ba1899011c46e44c09997ffdcd8766fe
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Added reference count for each time an NV12 buffer is queued to
the display. It is decremented when the buffer is swapped with
another one (including the same one queued again).
This is a crude solution since SGX is not keeping proper readops
count for displayed buffers.
Using atomic operations to change the reference count as the
incrementing and decrementing are done in different blocks
(graphics and display) under different locks. This should
still be OK because the incrementing and the checking is in
the graphics context.
Requires corresponding change in DDK.
Change-Id: I361577a3104e002cf9e5c5452fb2ce967545766b
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
|
|
|
|
|
|
|
|
|
| |
Using color-phase rotation we can swap red and blue color
components on LCD outputs to support BGRA and ABGR color formats.
YUV color conversion matrices are also adjusted so YUV colors
are not affected after color field swapping.
Change-Id: Iffd24ad0355b52bd76ff7dd783b25f694fef3d1c
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
|
|
|
|
|
|
| |
Added support for setting color-phase rotation coefficients for
(LCD) managers.
Change-Id: Id4f32c70cf6b6502d3a98a12b9d871670f4997b0
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
|
|
|
|
|
| |
Now can set color conversion coefficients for YUV pipelines.
Change-Id: If0b7f8c0f2828367e84513a8f8b44a227dbb7c22
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|\
| |
| |
| | |
Change-Id: I9123d2c339b3118ff0b8158ae5d2bb706147e3ea
|
| |
| |
| |
| |
| |
| |
| | |
Added color conversion coefficients to overlay info structures.
Change-Id: If8fe3127d94d5e41fef3bc64505fbd19fbf830cf
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Add Color Phase Rotation (CPR) support and sysfs files to enable CPR and
to set the CPR coefficient matrix.
CPR is enabled via manager?/cpr_enable file, and the coefficient matrix
is set via manager?/cpr_coef file. The values in cpr_coef are in the
following order:
RR RG RB GR GG GB BR BG BB
Signed-off-by: Tomi Valkeinen <tomi.valkeinen@ti.com>
|
| |
| |
| |
| |
| |
| |
| | |
early suspend has no way of erroring out.
Change-Id: I3a6dc7059ed989792d9f28ea88dd1f7b1273f012
Signed-off-by: Erik Gilling <konkers@android.com>
|
| |
| |
| |
| |
| |
| |
| | |
If there are DSS errors, DSSCOMP's early suspend could block forever.
Change-Id: I0adc8008696033abf1f9b9fef628d7f7ffcfa9ca
Signed-off-by: Erik Gilling <konkers@android.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We should not queue or get callbacks from invalid compositions,
or compositions in invalid states, so these can be BUGs.
Also made composition validating safer by not dereferencing
the composition in unknown state (while being validated).
Change-Id: I6327aee8ece622eab18b94050d3c427401879951
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Multiple compositions now can be waiting to be applied, and
apply was completing potentially an earlier composition. This
happened mainly during blanking.
Added an APPLYING state to mark compositions to be in the
queue.
Also fixed a concurrency issue with tiler slot freeing.
Change-Id: I99cf99f1b57e12a5c0bacebf06593086ae5da09a
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Keep mutex on during callback to avoid using a stale composition
pointer.
Also fixed an issue during dsscomp_apply() where the callback
might have already been programmed even though apply() or
wait_for_vsync() fails. In this case, we cannot automatically
call the RELEASED callback because DSS will also call the now
registered callback. This results in accessing a freed composition
handle.
Change-Id: Iee13591e1079074270eecf3e7a6f90f16c365775
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
waiting for a tiler slot to become available is already
(effectively) waiting for a vsync. Calling wait_for_vsync() in
dsscomp_apply() creates an effective swap interval of 2. We can
wait for vsync after applying DSS settings to get the same
smoothing effect (resulting in effective swap interval of 1.)
Change-Id: I4a0526ee1f0983f44cf3f53e340679ddf92c8227
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Prior compositions are completed in the early suspend handler,
and new compositions are ignored (immediately completed) while
display is blanked.
Change-Id: I436e8a804ed1f5ed14c3ab90e43a5c26a9501ae3
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Therefore, each composition must apply. Disable overlays that are
not configured correctly so composition applies. Also ignore
failing manager settings.
Change-Id: Ia93b6a321c9554fd043aed70458fe6300beb43e9
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
- Made dsscomp_mgr_callback delayed (in workqueue)
- Added mutex in dsscomp_mgr_callback
- Fixed double deletion issue when composition is blank
- Removed delay in dsscomp_gralloc_cb since mgr_callback will
call it in a workqueue now.
Change-Id: I5bd3dea46a83c2caf5376a5ed722342605312b4b
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Disable overlays that were part of the previous frame, but are
not part of the current frame.
Change-Id: I9fb2be9c451c18aa8be23d4b4a348b21283e0d99
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We may want to delay apply() as it is blocking. However, we will
not get the error code if the apply failed, only an ECLIPSED_SET
callback. This should be enough to handle the lifecycle of this
composition.
dsscomp_gralloc_queue() now calls delayed apply.
Change-Id: I494073a338b75fc85d570a7f030a9bb73dcec6a3
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Added dsscomp_gralloc_queue() to queue a DSS composition from gralloc.
This allows passing page list information for non-contiguous buffers,
and a callback funtion.
This gralloc layer automatically uses TILER1D areas to pin
non-contiguous buffers, so that they become contiguous and can
can be displayed by DSS. We only use dsscomp_gralloc_queue() for
displaying a composition, so set the mode for DISPLAY.
Also added new ioctl to test gralloc adaptaion layer.
dsscomp_gralloc_init() does two things: it receives pointer to
the device, and allocates TILER container areas.
DSSCOMP is initialized before TILER, which causes problems.
Thus, dsscomp_gralloc_init() stores the device pointer separately
from allocating the memory. This way each part can fail
independently.
Added dsscomp_gralloc_init() within dsscomp_gralloc_queue() as
it may be called before using the dsscomp device vs the file
pointer.
Using semaphore to keep track of available tiler slots. This
is needed because we are releasing tiler slots in a work
queue, that may get scheduled too late:
- we release tiler slots when a composition is eclipsed in a
work queue, as tiler_unpin_memory may be blocking.
- we also release tiler slots if a composition has failed to
be set.
PVRSVC's callback is called when the composition is first
displayed or eclipsed. It also gets called if a composition
fails to queue to avoid a similar check and call from the
called of queue().
Change-Id: I69d0a0c9fe452f05d846c6a38e314c5ac52d70e6
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Removed sync_id because compositions will be specified at once.
Moved to simple queue as all compositions will get displayed.
Delete composition in dsscomp_drop()
Print composition address instead of sync-id, which is now 0.
Change-Id: Ic5078f3da4d10d0d5a27c3df0d33e9ad5de362b6
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This patch links in the DSS composition (dsscomp) module into the
build and configuration system. DSS composition is enabled by default.
Change-Id: Ic9fe20416116997541bbc043ab4681d52db61a5b
Signed-off-by: Lajos Molnar <molnar@ti.com>
Signed-off-by: Mark Tyler <mark.tyler@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This patch implements a new DSS composition module.
DSSCOMP allows specifying a whole composition for a DSS display:
1. set overlay information for all overlays on a manager
2. reroute the overlays to the manager (overlay must be disabled
just as when using sysfs)
3. set manager information
4. optionally call manager->apply() that programs the DISPC
4. optionally do an update (after a sync call)
DSS already implements coordinated updates by separating overlay
information setting from the applicaion of those settings (which
happens only in manager->apply()). However current users of DSS
call manager->apply() for each change, which make coordinated
change impossible.
This API also implements auto cropping of all layers to the display
region. This makes switching displays and handling display resolution
changes easier (without getting "failed to setup overlay" messages.)
DSSCOMP operates on 3 levels.
base.c contains the basic DSS operations, such as setting DSS overlay
and managers using DSSCOMP's setting structures. Theoretically,
DSSCOMP could be used via only these operations.
queue.c contains the queuing mechanism. This module maintains
compositions queued to each overlay manager (the basic DSS composition
entity). Each composition is referred to by a unique sync-id.
Queueing operations consist of creating a composition, setting/getting
manager/overlay information for the composition, applying the
composition to the display (which also displays it on manually
updated panels), waiting on various states of a composition.
For now the basic queuing mechanism of DSSCOMP is "queue and forget".
Therefore, it is not necessary to dequeue each frame queued. A
consequence of this methodology is that if one applies a composition
to a display, any prior unapplied compositions will be dropped.
The queuing interface tracks which overlay is assigned to which
manager. This is done at the DSS programming level, as that is the
most reliable place to monitor overlay ownership. Nonetheless,
the device interface uses overlay information to verify overlay
ownership - which may be slightly out of sync. The user of
DSSCOMP should maintain overlay ownership to ensure flawless
sharing of overlays between managers. (E.g. should not use an
overlay on a new manager, until the overlay has been disabled on
the previous manager, and that composition has been programmed.)
device.c contains the device hooks to operating system, and the file
interface (via /dev/dsscomp's ioctls). /dev/dsscomp works on top of
the queueing mechanism.
There are 3 levels of header files.
linux/dsscomp.h: basic dsscomp structures and ioctls
plat/dsscomp.h: kernel dsscomp interface (on top of linux/dsscomp.h)
local dsscomp.h: common implementation structures and shared methods
Note: plat/dsscomp.h defines a handle typedef that causes a
checkpatch warning. I feel that the creation of a handle typedef
is warranted.
Limitations:
- no WB support
- unsure whether to call sync on non-manual update panels
- cannot get overlay/manager information on a composition without
first having set it
- not fully operational - still debugging some unit test issues
Change-Id: Ie62e8eeccfd6f3cae9e39e30d5105fa2bad68c62
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|\ \
| |/
| |
| | |
Change-Id: Ida1a3053c2e75c4bfa554319bade0e78011b5e4c
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Includes the following fixes:
- fixed color format modes for non-GFX pipelines
- fixed NV12 support in 1D mode buffer
- added all color formats to configure_overlay()
- fix maxdownscale limits for OMAP4 in DISPC setup plane.
Predecimation
- added predecimation support for NV12/YUV/rotated/SDMA buffers
- added scaling decision (predecimation)
- added conservative fclk calculation
- added sysfs controls
- tied into DSSCOMP
Change-Id: Ie730ac2b2da48b5de8e27f03035a8213696612a8
Signed-off-by: Lajos Molnar <molnar@ti.com>
Signed-off-by: Mark Tyler <mark.tyler@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This patch allows tracking when a particular overlay or manager
change has taken place, and when it is eclipsed (no longer used).
In DSS2 overlay/manager information travels through 4 stages:
1: (software) overlay/manager info
2: (software) cache (on manager->apply())
3: (hardware) shadow registers (on configure())
4: (hardware) DISPC (on vsync or go/enable)
Callback information for each settings can be passed as part of
the info structure (info->cb), which is the following struct:
struct omapdss_ovl_cb {
void (*fn)(void *data, int id, int status);
void *data;
};
id contains to the overlay/manager index.
If fn is NULL, no callback will take place. Otherwise, callbacks
will be generated on programming (when settings get into the
DISPC), when the settings are displayed, and on eclipse (when settings
are overwritten by newer settings). Depending on the stage where the
eclipse happens, a different callback is generated.
Status is one of the DSS_COMPLETION_... enum values specifying the event.
DSS_COMPLETION_ECLIPSED_SET - info was overwritten at stage 1
DSS_COMPLETION_ECLIPSED_CACHE - info was overwritten at stage 2
DSS_COMPLETION_ECLIPSED_SHADOW - info was overwritten at stage 3
DSS_COMPLETION_RELEASED - info was overwritten at stage 4 (after successfully
being displayed)DSS_COMPLETION_TORN - info was overwritten at stage 4 (before it was
successfully displayed)
DSS_COMPLETION_PROGRAMMED - info moved from stage 3 to stage 4 (this follows
the prior info's callback of DSS_COMPLETION_RELEASED/TORN)
DSS_COMPLETION_DISPLAYED - info in stage 4 has been successfully displayed.
This callback is received on every frame refresh. If only the first display
is required, it needs to be filtered out in the callback.
You can use the DSS_COMPLETION_RELEASED flag to see if an info has been
eclipsed (so you don't have to check for all 5 values).
There is a fundamental issue with tracking DSS settings in the current
DSS2 framework. Pipeline/manager settings are programmed first into
an info structure that stays around. These settins can be modified
piece-by-piece - as they are done using the sysfs framework. Theoretically,
these cause the old settings to be eclipsed by the new settings. However,
sysfs interface is used to augment the other DSS2 users that would register
for these callbacks. Therefore, we need to treat these partial updates
specially. For now - if the callback function and data are the same -
a different status is used: DSS_COMPLETION_CHANGED_SET.
Now these auxiliary interfaces also apply the changes automatically, so a
similar method needs to be used at level 2. (DSS_COMPLETION_CHANGED_CACHE)
This, however, causes callback info to stay around at the level 1
interface making it not useful for tracking the status. We cannot
guarantee that the information is tracked from setting to release/eclipse
if future settings may end up reusing (ignorantly) the same callback info.
For now the callback info is cleared at level 1 when transferred to
level 2. If other - callback unaware - DSS2 users modify overlay/manager
settings, (e.g. using sysfs controls in any way), they will not reset
the callback. We err on the side of sysfs, and we will not treat a
settings application (transferring level 1 info to level 2) as eclipsed
if the level-1 callback method is empty. This works for sysfs changes,
but we will be a missed callback if V4L2 or FB is changing the base
address, and we are using the callbacks to track buffer usage.
The other issue is using sysfs to enable/disable an overlay. We may
get a RELEASED event on disable, but no usage callback is done on
subsequent enable because sysfs will not request a callback.
Change-Id: Iebc2551ad1d5bce94bd668cf2c885c895809d416
Signed-off-by: Lajos Molnar <molnar@ti.com>
Signed-off-by: Mark Tyler <mark.tyler@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Added support for displaying buffers located in the TILER 2D containers.
Buffers in TILER 1D containers can be displayed using OMAP_DSS_ROT_DMA.
Change-Id: Ibd8eb2c8c705cb4c2a1aec4b78ef63d238af50ae
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Earlier the default color format,frmamebuffer has,was xRGB24U but SGX
is producing data in ARGB32. Also,SGX,the FB driver user,doesnt negotiates
with FB driver for the format and thus FB needs to provide ARGB32 to align
with SGX format.
Based on original commit from
http://git.omapzoom.org/?p=kernel/omap.git;a=commit;h=c75a7e1b83d446650b0f393746f8a810242a8739
Change-Id: I086ebeec0ccd00ffff9ddf3e5fd2e69d2074e99b
Signed-off-by: Mark Tyler <mark.tyler@ti.com>
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Add zorder Support on OMAP4, this feature allows deciding the visibility order
of the overlays based on the zorder value provided as a sysfs attribute of the
overlay object.
Since this feature is not supported on OMAP3 and OMAP2, the sysfs attibutes
always return zorder as 0 and do not write to dispc registers for non OMAP4
architectures.
Based on original commit from Archit Taneja on k38.
Change-Id: I3a11d6294352983e4feeeecf269f7811d56c29ff
Signed-off-by: Sumit Semwal <sumit.semwal@ti.com>
Signed-off-by: Archit Taneja <archit@ti.com>
Signed-off-by: Mark Tyler <mark.tyler@ti.com>
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
VID3 and Writeback pipeline registers have a linear relation, introduce these
registers and access these registers in DISPC functions. Add a dss_feature for
VIDEO3 pipeline.
Configure color conversion and global alpha for this new pipeline.
Also, create overlay object for the new video pipe.
Based on original k38 patch from Archit Taneja.
Change-Id: Ia7115eaa325289269121ab916c92f19cd0d4a7fb
Signed-off-by: Archit Taneja <archit@ti.com>
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Alpha blending is always enabled on OMAP4. The meaning of the
alpha enabled bits in DISPC_CONFIG registers now enabled OMAP3
compatibility mode which disables ZORDER settings. This is not
what we want when we set alpha blending.
This patch sets alpha blending always to true on OMAP4.
It also turns off OMAP3 compatibility mode for now, and reports
the compatibility mode results when querying dispc_alpha_blending.
Change-Id: Id8239cafdc6b6b55ab2c29454edd0108b9d69c6e
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
Add global alpha settings for video 1 overlay in
_dispc_setup_global_alpha. On OMAP4 VID1 has global alpha.
Change-Id: If89777de7f877027c0bd6652facf1fe0ff3c266d
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| |
| |
| |
| |
| | |
This will be reapplied using individual patches.
This reverts commit a7e96985ea15176a46ed3d767c58baf3c36d510e.
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
| |
| |
| |
| | |
This reverts commit ba54e8d28c5d1735897bbfefcd4e499dd7c26b49.
|
|\ \ |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When releasing framebuffer, free colourmap allocations.
Signed-off-by: Daniel J Blueman <daniel.blueman@gmail.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Some pieces of userspace like debian-installer expect to find the fb0
driver name by readlink-ing /sys/class/graphics/fb0/device/driver but
this was broken with amba-clcd as it sets up fb_info manually and missed
the .device parent pointer.
Signed-off-by: Loïc Minier <loic.minier@linaro.org>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The Freescale DIU framebuffer driver defines two constants, MIN_PIX_CLK and
MAX_PIX_CLK, that are supposed to represent the lower and upper limits of
the pixel clock. These values, however, are true only for one platform
clock rate (533MHz) and only for the MPC8610. So the actual range for
the pixel clock is chip-specific, which means the current values are almost
always wrong. The chance of an out-of-range pixel clock being used are also
remote.
Rather than try to detect an out-of-range clock in the DIU driver, we depend
on the board-specific pixel clock function (e.g. p1022ds_set_pixel_clock)
to clamp the pixel clock to a supported value.
Signed-off-by: Timur Tabi <timur@freescale.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The situation in which the problem occurred was with a Plugable UGA-2K-A
connected to a Samsung EX2220X display. The driver indicates that
1920x1080 is a valid mode (the first mode available, in fact), but
proceeds to set the framebuffer size to 1600x1200.
The patch corrects what seems to be a logic error, regarding unsetting
the FB_MISC_1ST_DETAIL flag, if the first (top/best) mode is invalid.
The existing code unset the flag if ANY mode was invalid.
Signed-off-by: William Katsak <william.katsak@alcatel-lucent.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
| |\ \
| | | |
| | | |
| | | | |
git://git.kernel.org/pub/scm/linux/kernel/git/lethal/sh-3.x into fbdev-fixes-for-linus
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
The check was intended to test if we have a valid pointer to write into,
but it mistakenly checks the pointer contents instead.
Since a valid pointer is mandatory for the chroma data if a YCbCr format
is used, the pointer check has been removed.
Signed-off-by: Damian Hobson-Garcia <dhobsong@igel.co.jp>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
In hecubafb_probe(), after a successful try_module_get, vzalloc may
fail and make the hecubafb_probe return, but the module is not put on
this error path.
This patch adds an exit point that calls module_put in such situation.
Found by Linux Driver Verification project (linuxtesting.org).
Signed-off-by: Pavel Shved <shved@ispras.ru>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Fix section mismatch warning in sm501fb:
WARNING: drivers/video/sm501fb.o(.text+0x21d6): Section mismatch in reference from the function sm501fb_init_fb() to the variable .devinit.data:sm501_default_mode
The function sm501fb_init_fb() references
the variable __devinitdata sm501_default_mode.
This is often because sm501fb_init_fb lacks a __devinitdata
annotation or the annotation of sm501_default_mode is wrong.
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Fix a chain of section mismatches in geode driver, beginning with:
WARNING: drivers/video/geode/gx1fb.o(.data+0x70): Section mismatch in reference from the variable gx1fb_driver to the function .init.text:gx1fb_probe()
The variable gx1fb_driver references
the function __init gx1fb_probe()
If the reference is valid then annotate the
variable with __init* or __refdata (see linux/init.h) or name the variable:
*_template, *_timer, *_sht, *_ops, *_probe, *_probe_one, *_console
Making the changes that Paul pointed out resulted in a few more
changes being needed, so they are all included here.
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Paul Mundt <lethal@linux-sh.org>
|
|\ \ \
| | |/
| |/| |
|
| | |
| | |
| | |
| | |
| | | |
Change-Id: I18168c887e1384c07dc033a1ffc57abdacb26073
Signed-off-by: Arve Hjønnevåg <arve@android.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
DSSCOMP apply method is corrected to make sure that the requested composition
structure is applied before returning from the method.
Signed-off-by: Sunita Nadampalli <sunitan@ti.com>
Signed-off-by: Iliyan Malchev <malchev@google.com>
|
|\ \ \
| |/ / |
|