Age | Commit message (Collapse) | Author |
|
Changing file permissions to 644
Signed-off-by: Ricardo Perez Olivares <x0081762@ti.com>
|
|
Changing file permissions to 644
Signed-off-by: Ricardo Perez Olivares <x0081762@ti.com>
|
|
git://dev.omapzoom.org/pub/scm/tisyslink/kernel-syslink into L24x9
Conflicts:
arch/arm/configs/omap_4430sdp_defconfig
drivers/Kconfig
drivers/Makefile
Signed-off-by: Ricardo Perez Olivares <x0081762@ti.com>
|
|
Changing to 644 file permissions
|
|
Changing to 644 file permissions
|
|
git://dev.omapzoom.org/pub/scm/axelcx/kernel-display into L24x9
Conflicts:
arch/arm/configs/omap_4430sdp_defconfig
arch/arm/mach-omap2/board-4430sdp.c
arch/arm/mach-omap2/include/mach/tiler.h
drivers/media/video/tiler/tiler.c
Signed-off-by: Ricardo Perez Olivares <x0081762@ti.com>
|
|
This patch adds a configuration option to syslink Kconfig to
enable fault recovery.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
This patch has changes to syslink ipc to enable recovery in
case of Ducati fault. Changes for recovery include:
1. Getting a handle to iommu and listening for events.
2. When fault event is received
- unregister for events and put back iommu handle
- block new file open requests until recovery is complete
- reject ioctl calls until recovery is complete
3. Recovery is complete when all open file handles are released.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
This patch allows procmgr to be built as a module without forcing
iommu to be built as a module. Iommu selection is moved to its
own configuration that selects OMAP_IOMMU by default and procmgr
no longer depends on OMAP_IOMMU.
[Hari] - Amended the patch to remove unused IOMMU_DEBUG config option.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
Signed-off-by: Hari Kanigeri <h-kanigeri2@ti.com>
|
|
This setting is required for bringing Tesla out of reset. Without
this the Tesla clock is not getting enabled.
Signed-off-by: Suman Anna <s-anna@ti.com>
|
|
gptimer5 clock is used by the DSP processor to provide clock functions
to BIOS. The clock needs to be enabled to allow the Task_sleep.
Without this, any Task_sleep in the BIOS-code would lead to an indefinite
suspend of the task.
Signed-off-by: Suman Anna <s-anna@ti.com>
|
|
This patch would add clock in dss hwmod
and irq, base address etc
Signed-off-by: Mayuresh Janorkar <mayur@ti.com>
|
|
|
|
|
|
|
|
Signed-off-by: Mythri P K <mythripk@ti.com>
|
|
Signed-off-by: Mythri P K <mythripk@ti.com>
|
|
Signed-off-by: Mythri P K <mythripk@ti.com>
|
|
VID* Attributes
Signed-off-by: Mythri P K <mythripk@ti.com>
|
|
Truncation logic has to be enabled if the color mode is RGB16 or RGB24.
Signed-off-by: Mythri P K <mythripk@ti.com>
|
|
Signed-off-by: Mythri P K <mythripk@ti.com>
|
|
|
|
|
|
Now, by installing parse-edid tool (from read-edid package) you can parse
/sys/devices/omapdss/displayN/edid to make for easier debugging of monitor issues.
Signed-off-by : Rob Clark <rob@ti.com>
|
|
A callback can be registered by the dssdev client in order to be notified
of resolution changes, for example an external monitor that is hot-plugged.
Note: the seemingly unrelated change to omap_dss_get_overlay() was needed
to avoid an opps in case of a race condition between the list of overlays
being built at boot time and a callback to the dssdev client.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This patch adds a tiler notifier callback mechanism
for events. This covers file close events.
Signed-off-by: Angela Stegmaier <angelabaker@ti.com>
|
|
Currently, the free page stack keeps 40MB worth of pages on hand.
Since there are still some system workarounds that are being resolved
using tiler memory, we need to temporarily adjust this figure. Also,
we haven't quite figured out what amount to keep that would strike a
balance between performance and memory consumption.
Signed-off-by: David Sin <davidsin@ti.com>
|
|
Signed-off-by: Bryan Wu <bryan.wu@canonical.com>
Signed-off-by: David Sin <davidsin@ti.com>
|
|
This patch will resolve the data inconsistency problems
seen in omap 4 es2.0. For es1.0, we were creating a
memory barrier using dsb() around the PAT reg accesses.
Additionally, we have added wmb() calls before the call to the
PAT refill function to ensure that the data reaches the main
memory beforehand.
Signed-off-by: David Sin <davidsin@ti.com>
|
|
|
|
Add logic to cap the total amount of pages to keep on hand.
Signed-off-by: David Sin <davidsin@ti.com>
|
|
Instead of allocating and freeing PAT page array
memory each time, allocate 128k upfront and reuse the
memory. This will avoid the possibilty of not being
able to obtain the memory after driver initialization.
Signed-off-by: David Sin <davidsin@ti.com>
|
|
Move division point in the container manager
to the right most corner, and extend division
points to allow for 'zero' and 'container_width'.
Signed-off-by: David Sin <davidsin@ti.com>
|
|
Clear PAT IRQ STATUS RAW and set PAT_DESC_0 to NULL before
exiting the refill function. Check the PAT STATUS
reg before exiting to make sure PAT 0 is in the ready state.
Signed-off-by: David Sin <davidsin@ti.com>
|
|
This resolves incorrect colors on NV12 images, as well as incorrectly
rotated images, which may result in DSS OCP_ERR.
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
slab.h needs to be included in the sources when using kmalloc and
kfree for 2.6.34
Signed-off-by: David Sin <davidsin@ti.com>
|
|
set all dmm and tiler related files to 644
Signed-off-by: David Sin <davidsin@ti.com>
|
|
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
Fixed size calculation for the allocated buffer into which the
packing takes place.
Tested using V4L2 unit tests and debug prints as only V4L2 uses this
function.
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
Since Kconfig entries have been moved to drivers/media/Kconfig,
we need to enable media support in order for dmm/tiler to build.
Signed-off-by: David Sin <davidsin@ti.com>
|
|
Signed-off-by: David Sin <davidsin@ti.com>
|
|
The reference count of blocks increases on alloc, and by registered
buffers that contain that block.
When all tiler drivers of a process are closed:
- blocks allocated by it are marked "freed"
- buffers registered by it are unregistered
All blocks that are no longer marked "allocated" and are not
referenced by registered buffers are actually freed.
Blocks allocated by the kernel API are tracked separately, and
are not release when tiler drivers are closed. They are only
freed using tiler_free, or when the kernel driver closes.
Fixed a bug with querying buffers by address, which did not
account for the new TCM support for 1D buffers and block management.
Added support for clearing PAT for freed blocks.
Added extended tiler kernel API-s for specifying group ids, pids,
alignment and offset.
tiler_allocx
tiler_reserve
tiler_reservex
tiler_map
tiler_mapx
Signed-off-by: Lajos Molnar <molnar@ti.com>
|
|
Removed unnecessary true/false identical macro values
Changed tcm_area.type field to is2d
Shortened variable names
Renamed functions based on what they do
Simplified some of the logic
Removed most of the debug messages and unused/duplicate variables
Removed duplicate parameter checks and made fuctions that always
succeed void.
Moved tcm_dbg.h macros into tcm_utils.h
Tested using memmgr_1.0 tests.
Signed-off-by: Lajos Molnar <molnar@ti.com>
|