msleep API runs in atomic context and can add extra delay
after sleep timeout. This can add extra delays for panel
regulator turn on/off sequence which affects device
suspend/resume timing. Moving to usleep_range enables the
delay backed by hrtimers.
Change-Id: Id9e50ab02c0166de41316eeb948e93cdce432e15
Signed-off-by: Dhaval Patel <pdhaval@codeaurora.org>
[cip@codeaurora.org: Resolved merge conflicts]
Signed-off-by: Clarence Ip <cip@codeaurora.org>
If the regulator is already enabled, driver doesn't need to
wait till post on delay is complete. This change fixes to avoid
delays during regulator enable.
Change-Id: If7c7d1f2895abaa12b94e9f5c78ba61f28b87758
Signed-off-by: Jie Cheng <rockiec@codeaurora.org>
Signed-off-by: Jeevan Shriram <jshriram@codeaurora.org>
Currently, all the event timers are associated to CPU0. These event
timers are not bind to any of the IRQs and always handled(hrtimer)
by CPU0. This event timers framework make sure event timers are
bound to clients IRQs and when the IRQ affinity has changed event
timer framework will make sure to move associated event timer to
a new CPU based on IRQ affinity.
CRs-fixed: 719258
Change-Id: Idbce3d5195655f558f74c208216f75515d235b8e
Signed-off-by: Murali Nalajala <mnalajal@codeaurora.org>
[cip@codeaurora.org: Remove drivers/soc/qcom/event_timer.c changes]
Signed-off-by: Clarence Ip <cip@codeaurora.org>
Janks in display are occurring due to delayed scheduling of rotator
work. Making it high priority would prevent such occurrences.
Change-Id: I85a7e4ca46596cfad56b29256ede30f1a13f0123
Signed-off-by: Terence Hampson <thampson@codeaurora.org>
Signed-off-by: Ujwal Patel <ujwalp@codeaurora.org>
VBIF debug bus test points provides the ARB/XIN
client status. These test points information is
required when MDSS core has requested some transaction
but they are not returning back. This change adds
the support for common code dump/log the test points
and adds the offset for msm8996 configuration.
Change-Id: I376a9e50aaaedb170bf4904551450206a85e4a1c
Signed-off-by: Dhaval Patel <pdhaval@codeaurora.org>
While switching resolution the fb memory requirement may increase and
the current memory may not be able to support the new resolution.
However since we can still go through overlay path, this requirement
shouldn't be mandatory. Allow it to go through, there are enough
checks during mmap to fail if trying to use this memory.
Change-Id: I76aafecd568763c456ea75234582b8a7aa9f83f4
Signed-off-by: Adrian Salido-Moreno <adrianm@codeaurora.org>
If command mode interface is being stopped, adding vsync handler can
cause the clocks to go on and cause improper state after.
There are checks for panel off inside add_vsync_handler however, there
needs to be synchronization while the stop is in progress.
Change-Id: I99bfca53a18bc818eecc94007357abdf4a251e50
Signed-off-by: Adrian Salido-Moreno <adrianm@codeaurora.org>
The mask for setting the six zone LUT is incorrect. This
change fixes the mask to make sure all data is written to
the hardware.
Change-Id: Ic8cb82cbf9d34d857c71c03decc8b64f3a3e9699
Signed-off-by: Benet Clark <benetc@codeaurora.org>
Use the output format instead of the input format when validating X
and Y offset
Change-Id: I0a41c5f2d3f17b720da35cecb8d5c1def76b2d0c
Signed-off-by: Terence Hampson <thampson@codeaurora.org>
The layer mixer GC hardware is meant to be programmed and enabled only
once. In cases where we want to reprogram the values in the hardware,
there is the possibility of corruption. This change forces the mixer
setup to skip GC programming if it is currently enabled. The user
should disable the feature before enabling it with different values.
Change-Id: I3c049d87a230be6f65ffa354099a2f1f1096f78b
Signed-off-by: Benet Clark <benetc@codeaurora.org>
On mdp3 targets iommu refcount is not properly protected
by mutex lock. This is causing iommu to be detached while
hardware is fetching the data from the iommu mapped
address leading to SNOC errors. Add mutex lock while
using the iommu refcounts to fix the issue.
Change-Id: Ife59791b08b123f6b60d3ffa6d8061839244934c
Signed-off-by: Shivaraj Shetty <shivaraj@codeaurora.org>
Scaling is not supported on DMA and CURSOR pipes. This change fails layer
validation when there is a scaling on any of these pipes.
Change-Id: I54428a6bf5516fcb66b41825a1d156d8052c5d84
Signed-off-by: Jeevan Shriram <jshriram@codeaurora.org>
When DSI underflow is reported, ctl reset is preformed which requires
certain critical regions to be protected by mutex. However this reset
itself doesn't need to be protected by koff_lock spinlock and thus move
it out of spinlock to prevent mutex locking while holding spinlocks.
Change-Id: I9a5da66c3db5c3839fa45a25898a5d610134594e
Signed-off-by: Ujwal Patel <ujwalp@codeaurora.org>
Autorefresh feature does not need to issue a manual
flush in order to start processing the frames.
Remove the manual flush and make sure that we wait
for one vsync before suspend.
Change-Id: I557a1a67017d9f0171915c209479e22bd908db36
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
Whenever resolution switch happens, ad needs to be reconfigured with
new resolution. Detect these changes and program frame size
accordingly.
Change-Id: Iac4e36c19d55d56de99318e13fd7435bdfac3a84
Signed-off-by: Adrian Salido-Moreno <adrianm@codeaurora.org>
Prior to programming the pipe for rotation, rotator needs to
calculate valid pixel extension so that it is properly configured
Change-Id: Idd92905d46d79f8feddbaf2d03653668360b7c10
Signed-off-by: Terence Hampson <thampson@codeaurora.org>
When dma tx timeout happens, the internal check to verify state of
hardware is actually checking for DMA interrupt enabled bit instead of
the actual state of DMA done.
Change-Id: I2c03e3e405289c223358b82e37e49203d0d0978b
Signed-off-by: Adrian Salido-Moreno <adrianm@codeaurora.org>
Format on pipe is changed to solid fill when it is
un-staged from mixer and put in cleanup list. This is
not required because HW is not going to pick up this pipe
when it is not present on any stage. This change removes
the unnecessary programming during cleanup and updates the
sw pixel extension configuration for solid fill pipe.
Change-Id: Idabd264dc9a3e93ef94e347e7f48fd6f88bd987b
Signed-off-by: Dhaval Patel <pdhaval@codeaurora.org>
[cip@codeaurora.org: Resolved merge conflicts]
Signed-off-by: Clarence Ip <cip@codeaurora.org>
MDSS driver stages the pipe on certain mixer but does not
update the stage index to unused after un-stagging the pipe
from kickoff. It is better that pipe free api checks the
mixer stage before setting pipe mixer null. This helps
to catch the bugs where pipe is getting staged without
commit call.
This change tries to update the pipe free API with proper
error checks. It also avoids mixer free twice from ctl
destroy to avoid error messages at each suspend/resume.
Change-Id: I6facef86b0816378d79fd17737b0f335c59ebfa6
Signed-off-by: Dhaval Patel <pdhaval@codeaurora.org>
During the pipe configuration process it was possible for multiple
thread contexts to read and write common register data, resulting in
a race condition. In order to prevent a race condition, we are now
protecting access to common registers are with reg_lock mutex.
Change-Id: I724e7870e856b3303448edbdf00f334ad38483f6
Signed-off-by: Terence Hampson <thampson@codeaurora.org>
There is a race condition between mdss_mdp_idle_pc_restore
and mdss_mdp_display_commit. Use ov_lock to serialize
execution of these two threads to fix race condition.
Change-Id: I6454e3d0edf2f3655e442d591fb3cac48f88a712
Signed-off-by: Kuogee Hsieh <khsieh@codeaurora.org>
In the current implementation, the DSI h/w configuration and DSI
PLL source are configured based on the MDSS DSI DT properties.
These can be changed dynamically based on the panel that is being
configured in LK. Add support to override these properties based on
the display command line string that is passed from LK to kernel.
Change-Id: I7ba5a2d16b3527537c91c79e7977a55d8b2e23c3
Signed-off-by: Padmanabhan Komanduru <pkomandu@codeaurora.org>
There is a race condition between kickoff and early
wakeup events; add mutex to protect the state of the
resources and prevent race condition.
Change-Id: Ibf3284f1d1fce26b88ea0a1f56f42a4682b16c77
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
Postprocessing userspace is a client of backlight notifications from
driver. There are two notifications exposed, one for AD BL attenuation
and one for regular backlight updates. Previously, the notification
would be for one type or the other. This change makes the notifications
occur together if both notifications are needed.
Change-Id: Ia5bc520a183106e1f9a876583cb3a17435be08ed
Signed-off-by: Benet Clark <benetc@codeaurora.org>
DSI clock manager is registering the post clk on twice
but does not register the post clk off. CORE/CTRL/PHY
power modules were never turned off because post clk
off callback is not registered.
Change-Id: I324e416241d9ec72e0aef8c4742849c14cffc37e
Signed-off-by: Dhaval Patel <pdhaval@codeaurora.org>
[cip@codeaurora.org: Resolved merge conflicts]
Signed-off-by: Clarence Ip <cip@codeaurora.org>
Idle power collapse API calls ctl restore for
each control which is powered on. This check will
pass for rotator ctls also. They have valid mixer
but mfd/panel_data null. Only mixer check is invalid
to call pp_resume for such ctls. It should check
the mfd/panel_data before restoring post processing
and compression block.
Change-Id: I09d599aae392e30d9ffae3cfd0e92bbda9aaa5e5
Signed-off-by: Dhaval Patel <pdhaval@codeaurora.org>
Modify the rotator bw vote to go to Active-Only in early stages when bw
is not needed, update the vote to Active+Sleep when bw must be
guaranteed.
Change-Id: I6a808831718426d93700fdf0c61b1eec1c0b2c1f
Signed-off-by: Terence Hampson <thampson@codeaurora.org>
In charging mode if charging code is not updating
the backlight then backlight will be turned off in
blank call and will not be turned on again on unblank.
Same is seen during ESD recovery. Set the unset
backlight level to the current backlight level during
blank so that after unblank, on first commit the
backlight value is restored. Also initialize the
brightness level to half of the max brightness.
Change-Id: Iee29b00a5e0fbbd4fee9d2c0dae1a40222b43939
Signed-off-by: Shivaraj Shetty <shivaraj@codeaurora.org>
Adjust the mdss event timer to wakeup before the readpointer
to help with latency issues.
Crs-Fixed: 840497
Change-Id: I384d7e13af1e4893aa200240c6a057ebc2bf92c6
Signed-off-by: Siddhartha Agrawal <agrawals@codeaurora.org>
When rotator fails on a request it would be helpful if information
about the request item was logged.
Change-Id: Ida8b3c9cd312a5142586323b1f9af613a44e40f3
Signed-off-by: Terence Hampson <thampson@codeaurora.org>
Add input event handler logic for video mode panels. In case of an
event, the idle notification should be prevented in anticipation of a
frame update.
Change-Id: I28c7b142b3b14375937d45f00fb552679db8f109
Signed-off-by: Vinu Deokaran <vinud@codeaurora.org>
Currently, ESD status check is default to ESD_MAX which
utilizes DSI BTA to perform ESD check as long as ESD check
enabled. Add ESD_NONE as default.
Change-Id: I85713f98a994f71a0e6b7bebe672f1147bd9ac52
Signed-off-by: Kuogee Hsieh <khsieh@codeaurora.org>
Add support to setup the layer mixer configuration
and select the number of DSC encoders in order to
support different panel configuration modes.
Change-Id: I36dbf8d9e325675bb3affa1794b2fc93ee6151b4
Signed-off-by: Ujwal Patel <ujwalp@codeaurora.org>
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
[cip@codeaurora.org: Removed dsi .dtsi changes]
Signed-off-by: Clarence Ip <cip@codeaurora.org>
Assertive display block can be configured in auto trigger mode,
which removes the requirement of issuing a start calculation
command. Change updates the ad calc worker to remove the kick off
of AD for each vsync and auto trigger mode is made the default mode.
Change-Id: Ie898a595cc9d3c8e521fffc8b3cde240dcbdce0e
Signed-off-by: Gopikrishnaiah Anandan <agopik@codeaurora.org>
Panels like AMOLED and dsi2hdmi bridge chip does not need
backlight control specified by DSI host. So no need to
fail dsi probe if backlight control is not defined.
Change-Id: I6e192b2e354d6e9146fb0890d2918feab9c7a2ed
Signed-off-by: Sandeep Panda <spanda@codeaurora.org>
Votes for register bus bandwidth are made from dsi clock manager. Create
separate handles for each dsi controller for bandwidth votes. MDP driver
does not maintain reference counting for register bus bandwidth votes
from each client. Voting from dsi clock manager removes the dependency
on reference counting.
Change-Id: I370053b143b6bc27358844a3958041da59281e92
Signed-off-by: Vinu Deokaran <vinud@codeaurora.org>
After 'commit 8635960b03b23f9c13c2214ed66cb0d621e7390c
("msm: mdss: refactor MDSS DSI driver")' there is no longer
the need to identify display destination types while reading
the dcs-cmd-by-left property.
Crs-Fixed: 874616
Change-Id: I765c810da1eb4f4bcd5e2c4e585ee7021a6274ee
Signed-off-by: Siddhartha Agrawal <agrawals@codeaurora.org>
Add msm bus scale settings for DSI driver. DSI driver will no longer
piggyback on MDP's bus bandwidth votes for AXI bandwidth.
Change-Id: If4e85c7b399f6e0f1664803d5489831c7b75d751
Signed-off-by: Vinu Deokaran <vinud@codeaurora.org>
[cip@codeaurora.org: Removed msm8996-mdss.dtsi changes]
Signed-off-by: Clarence Ip <cip@codeaurora.org>
In no panel connected case, when ESD is enabled we see
target is crashing because DSI controller is busy in waiting
for BTA from panel. This causes unwanted DSI PHY errors and
causes target crash due to watchdog bite.
This code change fixes the issue by extending the DSI PHY
recovery mechanism to BTA_TIMEOUT case also.
Crs-Fixed: 863383
Change-Id: I463a5b464cacca9c6c91a989218832aa7198522d
Signed-off-by: Sandeep Panda <spanda@codeaurora.org>
Signed-off-by: Siddhartha Agrawal <agrawals@codeaurora.org>
During ESD attack, sometimes the panel DDIC might not respond
to the ESD read command. This causes the data lane0 to get stuck
in reverse transmission causing DSI HS FIFO overflow errors
continuously in forward transmission. Enable the error interrupt
generation for LP_RX_TIMEOUT to handle this. Reset the DSI controller
and force DSI lanes to LP-11 to recover from this situation.
Crs-Fixed: 863383
Change-Id: Iee7d5e0c85fb77b9c8f899e569e37b54a100f67a
Signed-off-by: Padmanabhan Komanduru <pkomandu@codeaurora.org>
Signed-off-by: Sandeep Panda <spanda@codeaurora.org>
Signed-off-by: Siddhartha Agrawal <agrawals@codeaurora.org>
Signed-off-by: Vinu Deokaran <vinud@codeaurora.org>
Address offset was using hardcoded micro tile height of 8; Micro
tile height is not fixed and varies from format to format.
Change-Id: Idd033c955997ce56c64861637c60dd8e6f08c61e
Signed-off-by: Terence Hampson <thampson@codeaurora.org>
Attach an input event handler from mdp driver. In static screen cases
with command mode panel, mdp driver will wake up from idle state up on
receiving event from input driver. This will reduce latency during frame
kick off.
Change-Id: I60676915fea6f95a4aad5f49e09123d1a43f70cc
Signed-off-by: Vinu Deokaran <vinud@codeaurora.org>
In command mode panels, ROIs are manipulated differently based on
multiple factors like source split enabled or not, actual ROI, single
layer mixer, dual layer mixer, single DSI or Dual DSI. Current
implementation doesn't handle dual mixer single DSI configuration
correctly which results in pipeline hang when scaling is enabled. Fix
this by adding a check about this configuration.
Change-Id: Ie4a96948a280450c8f1f0638a55aa06f26063c38
Signed-off-by: Ujwal Patel <ujwalp@codeaurora.org>
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
Fix typo during the off work item; this can cause that
in case of an failure in the driver, the error goes silent
and power resources for the second display context
(i.e. bandwidth and clocks) stay enabled when not needed
without warning for this error.
Change-Id: Iff2683356ef6391ae57b3c5b711b544003584c7c
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
This change makes sure that all the clocks held by
the display driver gets disabled when the mdp clock
is also released and will switch the ahb vote to
active-only context, so the bus driver can also
free all the unneeded clocks when cpu is not active.
Change-Id: Id7ee92fdc6e5e91832e4dec61fb151ed20411ef1
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
This change modifies MDP vote to go to Active-Only in
an early stage when bandwidth is not needed and updates
the vote to Active+Sleep when bandwidth must be guaranteed.
Change-Id: I6ccb7a225f97416128518c990d881fa139ec1963
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
When continuous splash handoff is still pending, is not
needed to enable/release the iommu, since clocks and iommu
are already refcounted and enabled.
Add a check so by the time that aplash screen is still
on-going, we do not release the iommu.
Change-Id: I5ea948b2997aeac594f933ddca1096b755473884
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
Add support to gate the mdp clocks as soon as the
frame transfer is done for command mode panels.
Change-Id: I8325f26806ff3163edd87b9e3c01cd045f2aec77
Signed-off-by: Ingrid Gallardo <ingridg@codeaurora.org>
[cip@codeaurora.org: Resolved merge conflicts,
use debugfs_create_u32 for enable_gate]
Signed-off-by: Clarence Ip <cip@codeaurora.org>
