Age | Commit message (Collapse) | Author |
|
'partner-android-review.googlesource.com/2755569'] into android13-msm-pixelwatch-5.15-24Q1-release.
Change-Id: I4ff7aa83b6d58ccb9834b109e071e3c9ad2d62f2
Signed-off-by: Coastguard Worker <android-build-coastguard-worker@google.com>
|
|
kgsl: Gen7: Enable access protection for LPAC pipeline
Bug: 323918714
Bug: 324446026
Signed-off-by: Fred Lundquist <flundquist@google.com>
(cherry picked from commit f40a4b93e210f4bb8ee2374ab5bf73a8bcec9cf5)
(cherry picked from https://partner-android-review.googlesource.com/q/commit:f71334cb85eedd3ad3692645b589609d7ad291b6)
Merged-In: I146715b1b710a2c3d5ebef37c678b83ea49374ee
Change-Id: I146715b1b710a2c3d5ebef37c678b83ea49374ee
|
|
msm: kgsl: Do not free sharedmem if it cannot be unmapped
Bug: 324446026
Bug: 318393843
Signed-off-by: Fred Lundquist <flundquist@google.com>
(cherry picked from commit bad264f896099d73edb52172594bdf8cba1172a1)
(cherry picked from https://partner-android-review.googlesource.com/q/commit:6b7ddcfdb4a0ef731aabc8e0cf73f7993a02e203)
Merged-In: Ie6922a3d63be6c2494567167378038ad45fe06aa
Change-Id: Ie6922a3d63be6c2494567167378038ad45fe06aa
|
|
into android13-msm-pixelwatch-5.15-eos-release.
Change-Id: Ib3623b5713b5b95bb8d20109c9b46aaa3aa0d201
|
|
msm: kgsl: Simplify timelineobj cleanup
msm: kgsl: Defer drawobj_sync_timeline_fence_work() to a workqueue
Bug: 307604078
Signed-off-by: Andrew Evans <andrewevans@google.com>
(cherry picked from https://partner-android-review.googlesource.com/q/commit:5f0124adcdfe68e348874b5d5b12db3bcae3c104)
Merged-In: I82919a18ee097083ca50727ababa5cd69827880c
Change-Id: I82919a18ee097083ca50727ababa5cd69827880c
|
|
into android13-msm-pixelwatch-5.15-eos-release.
Change-Id: I93956866216dd73969761a66f1c58215d3f24af5
|
|
msm: kgsl: Limit the syncpoint count for AUX commands
msm: kgsl: Prevent wrap around during user address mapping
Bug: 299649795
Bug: 300941008
Signed-off-by: Andrew Evans <andrewevans@google.com>
(cherry picked from https://partner-android-review.googlesource.com/q/commit:abecb8dd35388c365dfa69617949ffb0997b25e6)
Merged-In: I01a27482be96fa5c68b593fbd84f9189834afb96
Change-Id: I01a27482be96fa5c68b593fbd84f9189834afb96
|
|
Prevent obscure build errors caused by having sourced
`build/envsetup.sh` from an Android tree.
Bug: 223899851
Change-Id: I780f12bfe288da3eded2e424e12a496a5d44bc0b
Signed-off-by: Andrew Evans <andrewevans@google.com>
|
|
This updates our sources to CodeLinaro release
AU_LINUX_ANDROID_LW.UM.2.1.R1.11.00.00.200.040 (LW.UM.2.1.r1-04000-SW5100.0),
which corresponds to ChipCode LW 2.0 r00078.3a.
Bug: 280095980
Change-Id: Icee1ff7849842f77853fbc9f1d9a59ac97413c51
Signed-off-by: Daniel Price <danielprice@google.com>
|
|
Postamble packets are executed in privileged mode by gpu. So we should keep
them in a privileged scratch buffer to block userspace access. For
targets with APRIV feature support, we can mark the preemption scratch
buffer as privileged too to avoid similar issues in future
Change-Id: Ifda360dda251083f38dfde80ce1b5dc83daae902
Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Signed-off-by: Kaushal Sanadhya <quic_ksanadhy@quicinc.com>
Signed-off-by: Amit Kushwaha <quic_amitkush@quicinc.com>
|
|
Point to kernel/private/msm-google:OWNERS_pixelwatchbsp
Bug: 247160135
Change-Id: I4b424c6b505f75611e7d14e51e05db93456c1393
Signed-off-by: Andrew Evans <andrewevans@google.com>
|
|
Change-Id: I1a7803126bbf0104c6f4d1f5b6c8a3acb124642f
|
|
This updates our sources to CodeLinaro release
AU_LINUX_ANDROID_LW.UM.2.1.R1.11.00.00.200.031 (LW.UM.2.1.r1-03100-SW5100.0),
which corresponds to ChipCode LW 2.0 r00062.1a.
Bug: 276757489
Change-Id: I3971246723ffdc9a80b977a7934c8270eb98783a
Signed-off-by: Daniel Price <danielprice@google.com>
|
|
|
|
Targets from gen7_9_0 have CPU vs GPU policy differentiation for perfmode
and hence the provision to do independent GPU perfmode for improved
performance.
Add support to configure the ACV vote from GMU that sets the perfmode.
Since for gen7_9_0 the perfmode is set through GMU, disable the icc path as
well.
Change-Id: I05574bcfca6080cba147c92a2e3939dbbe7cfadf
Signed-off-by: Mohammed Mirza Mandayappurath Manzoor <quic_mmandaya@quicinc.com>
|
|
Currently, there is no dependency between kgsl driver probe and smmu
driver probe, this results in issues in case smmu driver probe is not
done before kgsl driver probe. To ensure kgsl driver probe/bind
happens only post smmu driver probe, add kgsl MMU and CBs as a component
device.
As CB nodes contain "iommus" property, fw_devlink ensures that CBs
are probed only after corresponding supplier (smmu driver) probe is
done. Adding CBs as a component device ensures master bind (adreno_bind)
is called only once all components are probed. So adding CBs as component
device ensures kgsl probe/bind are done only post smmu driver probe.
Change-Id: I369c406128dc7f5c8904792f5979b3d0692ced5a
Signed-off-by: Deepak Kumar <quic_dkumar@quicinc.com>
|
|
|
|
|
|
|
|
When AQE is enabled, CP actively uses 112KB(48(BR)+48(BV)+16(AQE0))
of preemption record. So, for Gen7 snapshot increase preemption
record dump size from 64KB to 112KB.
Change-Id: Ibca2703294ec7dd689985f1e9d6b8ccd49063129
Signed-off-by: Hareesh Gundu <quic_hareeshg@quicinc.com>
|
|
Print the cmdline string of the culprit process during snapshot to
identify the exact command to trigger the same gpu hang. This will help
to make it easier to reproduce the gpu issue in scenarios involving
cmdline applications.
Change-Id: I26baa671aa7a5465cfc88cebe08aeee550e22922
Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
|
|
Enable needed configs for KGSL process reclaim feature
on Khaje.
Change-Id: I2ce784384cd623fc3204235217b401d5563be388
Signed-off-by: Archana Sriram <quic_c_apsrir@quicinc.com>
|
|
Change-Id: I0b666ac62f86cc21ea1fc299eeca3c78d6bbbc65
|
|
|
|
This reverts msm-5.10 commit b84bd97e37f8ac3a0f3194ebcbb1fa962e42cd73.
Warnings for direct dma clients during cache operations are now being
handled by dma-buf driver. Instead of doing dma_buf_unmap_attachment
early if we do it in destroy path, it helps in saving cycles and
improving performance during app launch.
Change-Id: Ic66dd3b66136318abf59685f95fee17890377fd4
Signed-off-by: Pankaj Gupta <quic_gpankaj@quicinc.com>
Signed-off-by: Archana Sriram <quic_c_apsrir@quicinc.com>
|
|
Set KGSL_MMU_SUPPORT_VBO feature at probe time.
PRR registers are not supported in A650 through HLOS.
Thus, disable this feature for A650 target.
Change-Id: Ied897f4cff4de1cd361e63d5df6ed3a80444fe01
Signed-off-by: Abhishek Barman <quic_abarman@quicinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Enabling some perf counter groups needs sending commands
to GPU. Add support to send these commands via HFI for
HW-Sched to enable these perfcounters.
Change-Id: I3441c024cf3a4bc56998d27153201e6cf9236d26
Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
Signed-off-by: Hareesh Gundu <quic_hareeshg@quicinc.com>
|
|
iommu->user_context can get detached in kgsl_iommu_close and later
get accessed in the deferred_destroy path. Flush the workqueue to
ensure pagetables are destroyed before proceeding for context detach.
Also handle set_ttbr0_cfg properly in the close path.
Change-Id: Idd936957e9c7617d978056604241e2e24dac8af4
Signed-off-by: Harshitha Sai Neelati <quic_hsaineel@quicinc.com>
|
|
Recently, there was a fix for a gpu bw starvation issue in a gen7 gpu.
This fix increases the GPU IB vote when gpu is at the lowest opp corner
and when it detects a high 'ram wait' above a particular threshold.
Some GPUs are less capable in terms of ddr channels, system cache,
smmu tbu size etc. So it is natural to see a much higher 'ram waits'
on these gpus. Due to this, there is now a very high IB vote in some
a6xx GPUs causing severe power regression in some use cases.
So, lets keep this ddr stall fix enabled only on Gen7 GPUs for now. We
have not seen a similar ddr starvation issue in a6xx/a5xx/a3xx family.
Change-Id: Ic89a9a90665566749e7715446ab16fdcbadb37d6
Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
|
|
Currently, we generate bus modifier hint (to update bus level) in gpubw
governor to identify the bus modifier value. Also, we leak
"wait_active_percent" stat from gpubw governor to kgsl_bus.c to again
update the bus level, something which a bus hint could have done. To make
it cleaner, we can keep the entire logic of bus hint generation logic
within the gpubw governor and pass a new hint to increase the buslevel
twice.
And refactor the bus hint generation logic in gpubw governer to make it
a little bit more readable.
Change-Id: Id996c31f20efd6735b7f79da642bd3db1aa3184e
Signed-off-by: Akhil P Oommen <quic_akhilpo@quicinc.com>
|
|
Use CONFIG_DEEPSLEEP to unload ZAP shader and passed
context info during SMMU aperture set.
Change-Id: I455c01822e5bc657689784f5dfcf3077ae6363ac
Signed-off-by: Amit Kushwaha <quic_amitkush@quicinc.com>
Signed-off-by: Abhishek Barman <quic_abarman@quicinc.com>
|
|
|
|
|
|
This reverts commit 45683098e772a463ce9d206dad0923a4340aaae4.
Users are allowed to override the default frequency behaviors and
request lower GPU frequencies. Revert that patch that was restricting
the user from making such requests.
Change-Id: Ie9995babc59aa12cde8a7a2d9f7d04b609538553
Signed-off-by: Carter Cooper <quic_ccooper@quicinc.com>
|
|
If GMU detects that a certain external hardware fence (part of a sync
object) is not signaled via the hardware path, GMU will query the
software status of that fence. Say, kgsl indicates to GMU that the
software status of the fence is signaled, GMU will wait for another
epoch. GMU will then send a F2H_MSG_CONTEXT_BAD packet if it still detects
that the hardware fence is not signaled via the hardware path. This
packet contains the context id (that was waiting for this external
hardware fence) and the sync object timestamp (to which the external
hardware fence belongs). Kgsl will log all the fences (and some metadata)
that are part of this sync object and trigger snapshot and recovery.
Change-Id: If7ece0d4e45b3ff4e0af05f9233752b99b0bb123
Signed-off-by: Harshdeep Dhatt <quic_hdhatt@quicinc.com>
|
|
If GMU detects that certain external hardware fences (part of a
sync object) haven't signaled in a given amount of time, GMU will
send F2H_MSG_SYNCOBJ_QUERY packet to query the software status of
these hardware fences. In response to this packet, kgsl will look
up these hardware fences (in the queried sync object) and indicate
to GMU whether the corresponding dma fences are signaled (or not)
via the same packet. In case a dma fence is signaled, add a log to
indicate the dma fence is signaled and the hardware fence isn't.
Change-Id: Iada8a8a2641cf74c591604b8236c8555d1fb4d2f
Signed-off-by: Harshdeep Dhatt <quic_hdhatt@quicinc.com>
|
|
Snapshot parsers need this to be able to identify this buffer
uniquely (using type=SNAPSHOT_GMU_MEM_HW_FENCE) to parse
it accordingly. Also, remove KGSL_MEMDESC_IOMEM flag as it is
no longer needed.
Change-Id: If0bf9c0efe9a099a8b6790737642b0fe2d69ce29
Signed-off-by: Harshdeep Dhatt <quic_hdhatt@quicinc.com>
|
|
|
|
|
|
|
|
|
|
|
|
Allocation of memory to user_ctxt_record is done based on adreno_device
flag. If a context is created when preemption is disabled, the memory
to user_ctxt_record is not allocated. Now if we enable preemption,
through the sysfs knob, the device flag is again set. We may try to
access user_ctxt_record in preemption_pre_ibsubmit(). So allocate memory
to user_ctxt_record based on the preemption feature flag instead
of preemption device flag.
Change-Id: I947ef5a1bfbd8bff471427fae60d7fcd507b85a2
Signed-off-by: SIVA MULLATI <quic_smullati@quicinc.com>
|