Currently status of DSP PD is not notified to the user process. User
process is unaware if the User PD on DSP has been killed. Notify
status of DSP PD to user process.
Change-Id: Ied5c4eea9f1fba4842c4366a7cb9f088cadf05c0
Signed-off-by: Himateja Reddy <hmreddy@codeaurora.org>
Increase kgsl-events worker thread priority to run at high
priority among default worker threads to get some benefit
in fence signaling.
Change-Id: Ia3eeb1beabc4bca1ef69e37bc40ef9868e29572b
Signed-off-by: Hareesh Gundu <hareeshg@codeaurora.org>
Signed-off-by: Kamal Agrawal <kamaagra@codeaurora.org>
Adding check in msm_minidump_add_region() to handle scenarios for
out of bound access while adding region in minidump table in SMEM.
Change-Id: Ic20663dbd2fa8ae96899930a7f7ba79dc204ff5e
Signed-off-by: Komal Bajaj <kbajaj@codeaurora.org>
Add llcc-qcom module to first stage to enable CPU LLCC slice
early in the boot chain.
Change-Id: I285b1bd53b7d9c3411decbdc180074a254b39245
Signed-off-by: Prasad Sodagudi <psodagud@codeaurora.org>
USB audio server shall get notify when any client is crashing or restarting
using registered del_client API (i.e. uaudio_qmi_svc_disconnect_cb). There
is race condition seen when USB audio server is just registered with QMI
framework from probe() context and possibly preempted without updating
uaudio_dev global variable, and QMI framework is calling
uaudio_qmi_svc_disconnect_cb() API on other worker thread context causing
NULL pointer dereference while accessing uaudio_dev. Fix this issue by
adding explicitly check against uaudio_dev as well initialized uaudio_dev
and qmi_disconnect_work before registering USB audio server.
Change-Id: If535ba808811e78c80ea769dac3d645074739d3b
Signed-off-by: Mayank Rana <mrana@codeaurora.org>
To help ensure that the GKI kernel and device kernel are kept in sync,
test whether abi_gki_aarch64.xml matches between the GKI kernel and
msm-5.10.
This check is not fail-safe however. If an ABI impacting change is
committed to either msm-5.10 or GKI tree without the corresponding ABI
update, then this check will incorrectly proceed. The correct way is to
do a full ABI comparison between msm-5.10 and the GKI tree with every
build, however that is time-intensive. Thus, prefer to do the simple
check for now.
Change-Id: Id801eeda347ce821fc6a29ae0e987c185a856bce
Signed-off-by: Elliot Berman <eberman@codeaurora.org>
Currently dlkm in kernel core drivers is not
supported for TUI. Compile touchscreen FTS driver
statically for TUI session.
Change-Id: I27453706e69b61bc70c2a2a8fa7c52f7fef8eb01
Signed-off-by: Chandan Uddaraju <chandanu@codeaurora.org>
This change Enhances IPC logs to add SE instance to know which
GPII is attached to which SE device, this helps in debug. Also,
removed unnecessary read of iommu-dma property.
Change-Id: I6a5a9312d80abab2addc0c63642429a787f9065c
Signed-off-by: Mukesh Kumar Savaliya <msavaliy@codeaurora.org>
When reading many registers with a count value that leads to more
than PAGE_SIZE bytes being returned, regmap_data_read_file()
outputs an extra PAGE_SIZE bytes of register address/value lines.
Correct this by mapping file position offsets based on the
register address and count values provided from userspace.
This also ensures that the proper set of registers is output in
the case of a regmap with reg_stride > 1 or a regmap which has
non-readable register regions.
Change-Id: I4ac5f112e176f994413a63ee8d91149978e076dd
Signed-off-by: David Collins <collinsd@codeaurora.org>
Implement our callbacks for .get_pool_size for the system and secure
system heaps.
Change-Id: Ic52ae76cb1bf10005a6bc6445b22e1d241daad92
Signed-off-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
The delayed unmap feature should only be used along with our DMA-BUF
heaps - enforce this at compile time.
Change-Id: Ib7fcfbc4a3e4cd7f1e389c71d13ed888fa791181
Signed-off-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
Since dma_heap_get_drvdata() is now exported, there's no reason to
maintain a reference to a struct dma_heap in our struct cma_heap
(since we can just access the cma_heap through
dma_heap_get_drvdata(dma_heap)). Remove all cma_heap->heap references
accordingly.
Change-Id: Id7b733e84f6357fac3977ec5d7cc91e03dead232
Signed-off-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
Do not import memalloc buffers (such as the secure preemption
buffers) into GMU VA if GMU access bit is not set.
Change-Id: Ic90bc21e96aebe31a9337e74f30fd53bbc89c943
Signed-off-by: Harshdeep Dhatt <hdhatt@codeaurora.org>
This brings over the support of vulkan semaphores to
hwscheduler.
Change-Id: I6267e070d58e46efee7fcb8f5c285f9191dc78c8
Signed-off-by: Harshdeep Dhatt <hdhatt@codeaurora.org>
Currently, we can only send 83 IBs due to the size limits
enforced by the HFI interface.Hence, we allocate a GMU buffer
to send upto a maximum of 2000 IBs across to the GMU. At any
time there may only be a single big IB inflight. This avoids
any overly large allocations which might appear in conformance
testing scenarios.
Change-Id: Ic5f2881f9e3bf77673a9add22dbfa8bce9e0ecb7
Signed-off-by: Harshdeep Dhatt <hdhatt@codeaurora.org>
Checking if there is enough room in between the offset of the
two subsequent field so that data of field 2 will not overlap
the data of field 1.
Change-Id: I96f656bb25878a302e7de109dd8f981045ed52e7
Signed-off-by: Nitin LNU <nlakra@codeaurora.org>
