Guidance from genpd team is to use a different power domain for each
device to avoid race condition in genpd framework when multiple devices
on same power domain runtime suspend in parallel.
Change-Id: I683ff80e5cef9583b65c14aa24db0f24d4ed981c
Signed-off-by: Patrick Daly <quic_pdaly@quicinc.com>
When all clients remove DDR bandwidth vote, DDR may power collapse.
As part of its shutdown sequence, it waits for an 'active' signal to
no longer be asserted by the gpu cx gdsc. Thus, if SW votes for the
gdsc to be active, but not for DDR bandwidth, this sequence may
get stuck.
Change-Id: I48d704f08cfe6d17159eb04d02f5ed123809f967
Signed-off-by: Patrick Daly <quic_pdaly@quicinc.com>
Newly added optional quirk "qcom,sleep-clk-bcr" adds delay of
200-250us after deasserting the USB3 BCR. This is needed on
some targets where sleep clk is used for BCR demet.
Change-Id: I88370838c29f679f2d2d90f565d3884d48bcdff2
Signed-off-by: Prashanth K <quic_prashk@quicinc.com>
Spilt memdump entries to static and dynamic dumps. Static dumps
are enabled by default. Dynamic mem dumps are enabled/disabled
using sysfs nodes and it's disabled in perf build.
Change-Id: I1f1f105d796391ec6e6001c249add9e8ba4b91ce
Signed-off-by: Saranya R <quic_sarar@quicinc.com>
Set RX settings mode to zero for Ultrashort channel
settings for sun PCIe controller.
Change-Id: I50b7896e6dabb2cda069c9242340dee02a225b8c
Signed-off-by: Vivek Pernamitta <quic_vpernami@quicinc.com>
Add memory and clock support for qup for Parrot-VM target.
Change-Id: I2bdd7fea970fafc9c245277c31d5dc4003294734
Signed-off-by: Prakash Yadachi <quic_pyadachi@quicinc.com>
Add idle states for CPUs and CPU clusters, added PSCI device,
to enable CPUs to enter deeper LPMs.
Disabled the idle states till Rumi validations are done.
Additionally. updated APPS RSC device to be in cluster power domain
to handle RSC activities when cluster is powering off.
Change-Id: I0dc50ff04bb480eb9ebdfa0bbaebfdf954c7c41b
Signed-off-by: Sneh Mankad <quic_smankad@quicinc.com>
The tui_test heap will be used by the large_dmabuf test
on sun-vm.
Change-Id: I84e5aee85c03e2cc809acc307509ce00aa74d967
Signed-off-by: Georgi Djakov <quic_c_gdjako@quicinc.com>
When all clients remove DDR bandwidth vote, DDR may power collapse.
As part of its shutdown sequence, it waits for an 'active' signal to
no longer be asserted by the gpu cx gdsc. Thus, if SW votes for the
gdsc to be active, but not for DDR bandwidth, this sequence may
get stuck.
Change-Id: I48d704f08cfe6d17159eb04d02f5ed123809f967
Signed-off-by: Patrick Daly <quic_pdaly@quicinc.com>
APPS needs to place proxy votes to ddr and cnoc when the SOCCP is in D0.
Change-Id: Idfa93910b51c6df033ea010480c1a8adeacd4af5
Signed-off-by: Gokul krishna Krishnakumar <quic_gokukris@quicinc.com>
SOCCP_SOCCP_SPARE_REG0 is used to check D0 status of SOCCP.
TCSR_SOCCP_SLEEP_STATUS is used to check D3 status of SOCCP.
Change-Id: Icee37cddb0b7ef303962cab0d9a8f37a211a05da
Signed-off-by: Gokul krishna Krishnakumar <quic_gokukris@quicinc.com>
ZRAM is being disabled on tuivm due to observed stability issues due to
ZRAM's use of GFP_MEMALLOC. Increase maximum movable zone size to
compensate.
When kswapd swaps out pages, PF_MEMALLOC is set, which results in zram
ignoring the min_watermark, and lowmem_reserve. This is problematic
because we intend for movable memory to stay in ZONE_MOVABLE, rather
than falling back to ZONE_NORMAL. If the later occurs to such a high
degree that all ZONE_NORMAL memory is consumed, as is observed to be
the case during a "C2PA" testcase, attempts to plug in additional
ZONE_MOVABLE memory may fail due to requiring a small amount of
GFP_KERNEL memory.
Change-Id: Ic847e90514f80ccb91daa98849440d7a9b97f156
Signed-off-by: Patrick Daly <quic_pdaly@quicinc.com>
"qcom,initial-movable-zone-size" creates a movable zone of the given size
before userspace starts. This is useful in a system where a userspace
component (psi-daemon) is used to hotplug in memory on a on-demand basis.
With swap disabled, the current tvm carveout size (68 Mb) hits
out-of-memory condition before psi-deamon is started.
Once psi-daemon is running, it detects the number of blocks added to the
movable zone via "qti_virtio_mem/kernel_plugged" sysfs file, and
can reduce or increment the number as necessary.
Change-Id: I5b62b6227e8a357ba3012979ed4bf937f1810437
Signed-off-by: Patrick Daly <quic_pdaly@quicinc.com>
This property describes how large of a movable zone should be created
when the virtio_mem device probes.
Also, fix all errors reported by make dt_binding_check.
Change-Id: I487ad7592d54021ddbb3caddb20774d3e076c766
Signed-off-by: Patrick Daly <quic_pdaly@quicinc.com>
The virtio-mem device has always had a dependency on mem-buf for plugging
in memory. Previously, the ordering was not a concern, since hotplug was
not done until userspace was up. However, in subsequent patches we
will instead hotplug in memory as soon as the virtio-mem device probes.
Change-Id: I9c4e4728a15ab32d65ffaea0ba2681a816a4abf0
Signed-off-by: Patrick Daly <quic_pdaly@quicinc.com>
Child node support for SCID heuristics compatible device.
Change-Id: Id1fb1e190181d39053dce629c6807262032744ad
Signed-off-by: Avinash Philip <quic_avinashp@quicinc.com>
Bindings file for WALT cycle counter driver is in incorrect location.
Move it to the correct location where all other bindings files are
present.
Change-Id: I9e8ef0a87ac6b311931535a82ccf3c784bcdc896
Signed-off-by: Sai Harshini Nimmala <quic_snimmala@quicinc.com>
Parrot need to follow non-relocatable absolute addresses for VM due to
firmware constraints. VM kernel load address is the start address of the
vm_mem_region and build system requires the load address of the kernel
to be 2MB aligned.
This patch adjusts the start address of the trust_ui_vm_mem region for
Parrot to align to 2MB. This also means reducing its size by 1MB.
Change-Id: I1816ec5f6ff18f55ecc4dec39958d442151f8cb0
Signed-off-by: Hrishabh Rajput <quic_hrishabh@quicinc.com>
Signed-off-by: Swetha Chikkaboraiah <quic_schikk@quicinc.com>
Binding document to support MPAM(Memory System Resource Partitioning
and Monitoring) MSC(Memory System Component) interface framework added
to support multiple MSC components. Currently SLC(System Level Cache)
MSC support integrated to MSC interface with API support to configure
SLC Capacity and monitors.
Change-Id: I866b0e8cd6106f86535baf004d25e81e406e3e12
Signed-off-by: Avinash Philip <quic_avinashp@quicinc.com>
