In pstate_cmp the plane_state structures are accessed
without checking it is allocated or not, which leads to
null pointer dereference. To fix it NULL check is added.
Change-Id: I5982138b396b70979205d87bc9aa260d9501fee0
Signed-off-by: Mahadevan <mahap@codeaurora.org>
Linux message queues(msgq) are used to communicate between VM's
This change creates a display specific msgq and a set of
framework api's for the VM layer to invoke communication
between the VMs'.
Display msgq is bidirectional. Tx end is created by default for
both the VM's. Rx is kept optional to avoid unnecessary creation
of listener threads.
Change-Id: If4e6410045293d6b21087a76ec018d4784dd5238
Signed-off-by: Jeykumar Sankaran <jsanka@codeaurora.org>
If event register is called from LA VM when the HW is owned by trusted
VM, it will lead to a crash. This change adds checks to prevent illegal
register access when the HW is not owned by LA VM.
Change-Id: Ib4d956fdce4b5aae41e765d7355082c5e233efe2
Signed-off-by: Ping Li <pingli@codeaurora.org>
RSC is updated during encoder enable/disable/modeset or power-
collapse/wakeup. The RSC vote for primary display will change
depending on QSYNC status which might change without modeset or
power-collapse. Add a call to update the primary display's vote
when the QSYNC property is updated.
Change-Id: I664754e4aae689ef2db5b1ef7b8b55a4c5fad3be
Signed-off-by: Steve Cohen <cohens@codeaurora.org>
Add checks to block transition to trusted-vm when multiple encoders
are attached to a crtc. This ensures, concurrent writeback is
disabled before trusted-ui usecase. Additionally, avoid enabling /
disabling the IRQs associated with the cwb encoder during all
the transitions to avoid any unbalanced calls.
Change-Id: I022077018ac9b7dfb62506cfaddcb60cb8b35ed8
Signed-off-by: Veera Sundaram Sankaran <veeras@codeaurora.org>
Iterate over the crtc old/new state and decide which ones
to process by checking the active state of both during
the vm release in error cases.
Change-Id: Iab20e89792c53fb72e0c00f1fa5091616c7afbf3
Signed-off-by: Veera Sundaram Sankaran <veeras@codeaurora.org>
Fix the checks done as part of the trsuted-ui transition.
Allow full validation to go through whenever the new vm request
or old vm request is not NONE. Split the checks and the vm acquire,
so vm acquire can be done after all necessary checks are over.
Change-Id: I91165010ad110193c2ca3947af8c6504cd259919
Signed-off-by: Veera Sundaram Sankaran <veeras@codeaurora.org>
Add changes to increment refcount of dma_fence before accessing
the input_fence structure in plane_state through debug_fs node and
decrement once access is finished. This prevents the use after free
issues when accessed through debug_fs node.
Change-Id: I262336ce2b86a94d7662d834e8fd3509dadda1bd
Signed-off-by: Jayaprakash <jmadiset@codeaurora.org>
Currently, in SST mode, sink edid is read once after hpd
and reused on subsequent mode enumeration calls. But in
MST mode, there is no caching and the driver re-reads the
edid from the sink on every get_modes call for each monitor.
Each read takes more than 500ms causing unnecessary delays
during MST enumeration.
This change reads the edid once per hpd and uses the
cached data on subsequent calls.
Change-Id: I27545a44b9f9bd40000dde60735815f9c47fa54c
Signed-off-by: Sudarsan Ramesh <sudarame@codeaurora.org>
This change adds support to program both qsync and variable
refresh rate in the same atomic commit. During vrr
usecase, if qsync is enabled, avr ctrl gets programmed
during prepare phase as well as after configuring timing
engine. This change also handles such scenarios to prevent
double programming of avr ctrl.
Change-Id: I19461423b0ae08c8204b5edeb98e3d73ce16a21b
Signed-off-by: Yashwanth <yvulapu@codeaurora.org>
In current implementation qsync min fps is single value.
It is same for all the list of supported dfps list.
Added support for new dt entry dsi-supported-qsync-min-fps-list
corresponding to the fps supported in the dfps list
dsi-supported-dfps-list.
Change-Id: Ifd5309c2f51865a3c0d9fadb65cbcd291b6ef42b
Signed-off-by: Raviteja Tamatam <travitej@codeaurora.org>
Currently vblank enable/disable events are processed based
on the vm ownership check. Use the vblank helper functions
instead, during the transitions to avoid processing the
vblank from DRM framework level.
Change-Id: Icb2bd569e73d9206a234489a288320fecf631845
Signed-off-by: Veera Sundaram Sankaran <veeras@codeaurora.org>
Add changes to extend the current evtlogs for ease
of debug purposes.
Change-Id: Ibe136afbf25b8b8decfc06ef3868e3f628df4dcb
Signed-off-by: Jayaprakash <jmadiset@codeaurora.org>
Avoid crtc_state active_changed flag and explicitly check new/old
crtc_state active to decide if trusted vm transition checks are
required. As the active_changed might not be set at this point
in atomic_check sequence.
Change-Id: I142befad68359ae5ee862a1306bdc75531d63e70
Signed-off-by: Veera Sundaram Sankaran <veeras@codeaurora.org>
As per current design in crtc disable sequence event thread
can cause deadlock.This change avoids mutex lock for crtc
flush event thread to flush all event work.
Change-Id: Ia7156032bee9245fe84fd1621cf1994920b8e574
Signed-off-by: Narendra Muppalla <NarendraM@codeaurora.org>
This change enables system cache support for shima
target which allows sde to read image from system
cache instead of DDR memory during static display
for video mode panels.
Change-Id: I2d7e17c4a6f6b477acf84fd2914c8db2d83df286
Signed-off-by: Krishna Manikandan <mkrishn@codeaurora.org>
This reverts commit 9ce6a2fbb3.
Revert the change for sde_encoder.c, and keep the change for
sde_connector.c.
Change-Id: I19ff26e4543b9b338ccaf363c99c5eb2c115f99b
Signed-off-by: Zhao, Yuan <yzhao@codeaurora.org>
Signed-off-by: Abhijit Kulkarni <kabhijit@codeaurora.org>
Commit Ifa100424733 ("disp: msm: sde: delay
encoder disable for clone mode") delays the CWB
encoder disable but it is also skipping the CWB
disable flush. That can cause the underrun on dp
display if it uses the same 3d_merge HW block. This
change reverts the portion of original code and
only delays the last cwb frame done wait. It still
keep the last CWB frame done wait as it is if crtc
is also moving to inactive state.
Change-Id: I3461188a35197f2925899ceea7ef705adf00a398
Signed-off-by: Dhaval Patel <pdhaval@codeaurora.org>
The worker function for enabling sys-cache re-uses the flushes
from the previous commit. This ends up flushing several
blocks which have no programming update. Instead, only flush
planes attached to the CRTC which is entering sys-cache state.
Change-Id: Ieed68a713e9e56a5718518cf1a6b632f37864fb9
Signed-off-by: Steve Cohen <cohens@codeaurora.org>
In few cases, even though input fence is freed, it might
get accessed through debugfs node while dumping fence list.
This change updates input fence to NULL in plane state
destroy once it is freed to prevent accessing it further.
Change-Id: I1a3799249e6a177af87653e17d014d1dad7ca638
Signed-off-by: Yashwanth <yvulapu@codeaurora.org>
Check for invalid topology value before extracting the number
of LMs in dither setup path. This occurs when there is a race
condition between runtime supend/resume and dynamic FPS switch.
Change-Id: Ibf4bd2fe5f094ddec11ebc5b85ea9a5cfa0e91a7
Signed-off-by: Amine Najahi <anajahi@codeaurora.org>
Currently, trusted VM accepts and releases the resources
on RM notifications.
In cases, where TUI fails to submit the first frame on the trusted
vm, display cannot rely on the teardown commit IOCTL to release the
resources back to primary VM.
To handle scenarios where RM notification drops, the VM's should be
able to ACCEPT/RECLAIM resources without relying on the RM.
To address above scenarios, this change moves the resource handling
calls from notification handlers to user prompt. With this change,
trusted VM will ACCEPT the resource only on the first frame commit and
primary VM can RECLAIM the resources back if the TUI use case fails
or any of the RM notifications fail to deliver.
Change-Id: Iebb1724a7558e52567f8af1a49e38f8adbec88a0
Signed-off-by: Jeykumar Sankaran <jsanka@codeaurora.org>
When continuous splash is enabled, KMS populates the modes of
a connector during splash config. With TUI VM enabled, this path
will also be exercised by trusted vm for splash handoff, where for
a given display we try to use the same HW blocks configured by
the primary VM.
Since the trusted VM invokes the path under connection_mutex lock,
calling fill_modes will lead to a deadlock. So, move the connector
mode population during init for DSI displays, since the modes are
available for the connector during initialization.
Change-Id: I12f920f4af84b7bccc97a5f5edb5117ada49b960
Signed-off-by: Jeykumar Sankaran <jsanka@codeaurora.org>
When HW resource is not available, return ENODEV so that user space can
check the errno and do proper handling.
Change-Id: If5422f1020c262ede4ac7226a274da2c5713b446
Signed-off-by: Ping Li <pingli@codeaurora.org>
During stability stress testing, there are many instances where
the panel is unreachable which results in the ESD check mechanism
failing continuously with PANEL_DEAD events. There are two possible
reasons. Either the device is bad or DSI controller is hung. As of
today, logs/xlog is flooded and info on root cause is missed in the
logs. To catch the root cause for such issues, this patch induces
forced crash when there are five consecutive ESD check failures.
Change-Id: Id0a5762ac977f8a209af651f65cbe9da199cb8d0
Signed-off-by: Lipsa Rout <lrout@codeaurora.org>
The last commit in primary-vm before transitioning to
trusted-vm during trusted-ui usecase will be NULL
commit for video mode. In the current splash path,
rm reserve during cont splash res init will succeed
only if there are sspps staged on the mixer. Avoid
this check for vm handoff usecase.
Change-Id: I9829c66f8eb856cc009edc35e489bc9adad1c5d7
Signed-off-by: Krishna Manikandan <mkrishn@codeaurora.org>
