Added a binding document for COT descriptors which is going
to be used in order to create COT desciptors at run-time.
Signed-off-by: Manish V Badarkhe <Manish.Badarkhe@arm.com>
Change-Id: Ic54519b0e16d145cd1609274a00b137a9194e8dd
Change handler of FFA version interface:
- Return SPMD's version if the origin of the call is secure;
- Return SPMC's version if origin is non-secure.
Signed-off-by: J-Alves <joao.alves@arm.com>
Change-Id: I0d1554da79b72b1e02da6cc363a2288119c32f44
Embed Arch Architecture SMCCC services in stm32mp1 SP_MIN. This
service is needed by Linux kernel to setup the SMCCC conduit
used by its SCMI SMC transport driver.
Change-Id: I454a7ef3048a77ab73fff945e8115b60445d5841
Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
SCMI shared memory is used to exchange message payloads between
secure SCMI services and non-secure SCMI agents. It is mapped
uncached (device) mainly to conform to existing support in
the Linux kernel. Note that executive messages are mostly short
(few 32bit words) hence not using cache will not penalize much
performances.
Platform stm32mp1 shall configure ETZPC to harden properly the
secure and non-secure areas of the SYSRAM address space, that before
CPU accesses the shared memory when mapped non-secure.
This change defines STM32MP_SEC_SYSRAM_BASE/STM32MP_SEC_SYSRAM_SIZE and
STM32MP_NS_SYSRAM_BASE/STM32MP_NS_SYSRAM_SIZE.
Change-Id: I71ff02a359b9668ae1c5a71b5f102cf3d310f289
Signed-off-by: Etienne Carriere <etienne.carriere@st.com>
When using SP_min as monitor, only sp_min_warm_entrypoint() is a valid
secure entry point.
Change-Id: I440cec798e901b11a34dd482c33b2e378a8328ab
Signed-off-by: Etienne Carriere <etienne.carriere@st.com>
Signed-off-by: Nicolas Toromanoff <nicolas.toromanoff@st.com>
Disable use of Neon VFP support for platform stm32mp1 when
building with SP_MIN runtime services as these can conflict with
non-secure world use of NEON support. This is preferred over a
systematic backup/restore of NEON context when switching
between non-secure and secure worlds.
When NEON support is disabled, this is done for both BL2 and BL32 as
build process uses common libraries built once for both binaries.
Change-Id: I4e8808dcb6ef58fc839e6f85fd6e45cfbaa34be0
Signed-off-by: Etienne Carriere <etienne.carriere@st.com>
BL32/SP_MIN configures platform security hardening from the shared
resources driver. At the end of SP_MIN initialization, all shared
resources shall be assigned to secure or non-secure world by
drivers. A lock prevent from further change on the resource
assignation. By definition, resources not registered are assign
to non-secure world since not claimed by any component on the BL.
No functional change as all resources are currently in state
SHRES_UNREGISTERED hence assigned to non-secure world as prior
this change in stm32mp1_etzpc_early_setup() and
sp_min_platform_setup().
Change-Id: Ic41fab47216c3b8b7a6a75b8358cfcec411ed941
Signed-off-by: Etienne Carriere <etienne.carriere@st.com>
Get number of pins in the GPIOZ bank with helper function
fdt_get_gpio_bank_pin_count(). Save the value in RAM to prevent
parsing the FDT several time for the same information.
Change-Id: Ie68e300804461ffce09914100a7d2962116023b5
Signed-off-by: Etienne Carriere <etienne.carriere@st.com>
Define enum stm32mp_shres for platform stm32mp1. The enumerated
type defines all resources that can be assigned to secure or
non-secure worlds at run time for the platform.
Change-Id: I5de20d72735856645f1efd0993643278e8d35bcb
Signed-off-by: Etienne Carriere <etienne.carriere@st.com>
STM32MP1 SoC includes peripheral interfaces that can be assigned to
the secure world, or that can be opened to the non-secure world.
This change introduces the basics of a driver that manages such
resources which assignation is done at run time. It currently offers
API functions that state whether a service exposed to non-secure
world has permission to access a targeted clock or reset controller.
Change-Id: Iff20028f41586bc501085488c03546ffe31046d8
Signed-off-by: Etienne Carriere <etienne.carriere@st.com>
Cortex A76 erratum 1800710 is a Cat B erratum, present in older
revisions of the Cortex A76 processor core. The workaround is to
set a bit in the ECTLR_EL1 system register, which disables allocation
of splintered pages in the L2 TLB.
This errata is explained in this SDEN:
https://static.docs.arm.com/sden885749/g/Arm_Cortex_A76_MP052_Software_Developer_Errata_Notice_v20.pdf
Signed-off-by: John Powell <john.powell@arm.com>
Change-Id: Ifc34f2e9e053dcee6a108cfb7df7ff7f497c9493
Cortex A76 erratum 1791580 is a Cat B erratum present in earlier
revisions of the Cortex A76. The workaround is to set a bit in the
implementation defined CPUACTLR2 register, which forces atomic store
operations to write-back memory to be performed in the L1 data cache.
This errata is explained in this SDEN:
https://static.docs.arm.com/sden885749/g/Arm_Cortex_A76_MP052_Software_Developer_Errata_Notice_v20.pdf
Signed-off-by: John Powell <john.powell@arm.com>
Change-Id: Iefd58159b3f2e2286138993317b98e57dc361925
* changes:
drivers/scmi-msg: smt entry points for incoming messages
drivers/scmi-msg: support for reset domain protocol
drivers/scmi-msg: support for clock protocol
drivers/scmi-msg: driver for processing scmi messages
It is desired to have the peripheral writes completed to clear the
interrupt condition and de-assert the interrupt request to GIC before
EOI write. Failing which spurious interrupt will occurred.
A barrier is needed to ensure peripheral register write transfers are
complete before EOI is done.
GICv2 memory mapped DEVICE nGnR(n)E writes are ordered from core point
of view. However these writes may pass over different interconnects,
bridges, buffers leaving some rare chances for the actual write to
complete out of order.
GICv3 ICC EOI system register writes have no ordering against nGnR(n)E
memory writes as they are over different interfaces.
Hence a dsb can ensure from core no writes are issued before the previous
writes are *complete*.
Signed-off-by: Sandeep Tripathy <sandeep.tripathy@broadcom.com>
Change-Id: Ie6362009e2f91955be99dca8ece14ade7b4811d6
This patch updates the 'bl31_check_ns_address()' helper function to
check that the memory address and size passed by the NS world are not
zero.
The helper fucntion also returns the error code as soon as it detects
inconsistencies, to avoid multiple error paths from kicking in for the
same input parameters.
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
Change-Id: I46264f913954614bedcbde12e47ea0c70cd19be0
Adjust the latest OP-TEE memory definitions to the
newest TF-A baseline.
Change-Id: Ib9c82b85f868adaf3c7285eb340486bda9c59c36
Signed-off-by: Konstantin Porotchkin <kostap@marvell.com>
Add llc_sram_enable() and llc_sram_disable() APIs to Marvell
cache_lls driver.
Add LLC_SRAM definition to Marvell common makefile - disabled
by the default.
Add description of LLC_SRAM flag to the build documentation.
Change-Id: Ib348e09752ce1206d29268ef96c9018b781db182
Signed-off-by: Konstantin Porotchkin <kostap@marvell.com>
Extend the CCU tables with secure SRAM window in all board
setups that uses SoCs based on AP806/AP807 North Bridges
Change-Id: I4dc315e4ea847562ac8648d8a8739244b548c70e
Signed-off-by: Konstantin Porotchkin <kostap@marvell.com>
It is not needed to reserve 64MB for FIP. Limit this to 4MB
for both supported Armada SoC families.
Change-Id: I58a8ce4408a646fe1afd3c1ea1ed54007c8d205d
Signed-off-by: Konstantin Porotchkin <kostap@marvell.com>
[Extract from bigger commit]
Signed-off-by: Marcin Wojtas <mw@semihalf.com>
Make sure the current CCU window is not in use before adding
a new address map during MSS BL2 image load preparations.
At BL2 stage the CCU Win-2 points to DRAM. If additional mapping is
added to MSS BL2 stage initialization, the DDR entry will be destroyed
and lead to the system hang.
Change-Id: I215e83508acc37d54dab6954d791b9a74cc883ca
Signed-off-by: Konstantin Porotchkin <kostap@marvell.com>
Add ccu_is_win_enabled() API for checking the CCU window
state using AP and window indexes.
Change-Id: Ib955a2cac28b2729b0a763f3bbbea28b476a2fe4
Signed-off-by: Konstantin Porotchkin <kostap@marvell.com>
Make all LLC-related macros to start with the same prefix
Add more LLC control registers definitions
This patch is a preparation step for LLC SRAM support
Change-Id: I0a4f0fc83e8ef35be93dd239a85f2a9f88d1ab19
Signed-off-by: Konstantin Porotchkin <kostap@marvell.com>
The configuration space of each standalone CP was updated in BL31.
Loading FW procedure take places earlier in SCP_BL2.
It needs to be done after access to each CP is provided.
Moving the proper configuration from BL31 to BL2 solves it.
Change-Id: I44cf88dfd4ebf09130544332bfdd3d16ef2674ea
Signed-off-by: Ben Peled <bpeled@marvell.com>
As a preparation to support proper loading the OPTEE OS image,
enable the BL32 specific defines in case the SPD is used.
On the occasion move two BL32-related macros to marvell_def.h
and fix BL32_LIMIT definition.
Change-Id: Id4e2d81833bc1895650cca8b0fc0bfc341cf77f3
Signed-off-by: Konstantin Porotchkin <kostap@marvell.com>
Signed-off-by: Marcin Wojtas <mw@semihalf.com>
The phy porting layer uses defaults defined in
"phy-default-porting-layer.h" when board specific file
"phy-porting-layer.h" is not found. Because of the regression the board
specific directory was not included, therefore all boards used default
parameters.
Change-Id: I66e5e6eb8a39cca5aeeb4de6dab2ceddc39c1e31
Signed-off-by: Grzegorz Jaszczyk <jaz@semihalf.com>
Configuring common phy selector which was missing for AP mode.
Change-Id: I15be1ba50b8aafe9094734abec139d72c18bb224
Signed-off-by: Grzegorz Jaszczyk <jaz@semihalf.com>
1) Relay only on rx training, remove parts responsible for tx training
(trx training).
2) Add extra steps e.g. preconfigure FFE before starting training.
3) Remove some unnecessary steps like RRBS31 loopback setting which
shouldn't be relevant for tx_training.
Change-Id: Ib1e8567714f9ce33578186a262c339aa4b1c51f2
Signed-off-by: Grzegorz Jaszczyk <jaz@semihalf.com>
Before this patch the configuration took place only for CP0 and CP1, but
since new platforms can contains up to 3 CPs update is required.
Change-Id: Iebd50bbe7b9772063e2c4efb3a7ecbfd593e950d
Signed-off-by: Grzegorz Jaszczyk <jaz@semihalf.com>
This patch provides the platform level support to enable GICv3
drivers on future Tegra platforms.
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
Change-Id: I966a4502b2a4a7bd1ce66da843997c9ed605c59f
The previous sequence used by the driver to program the new memory
aperture settings and clear the non-overlapping memory was faulty.
The sequence locked the non-overlapping regions twice, leading to
faults when trying to clear it.
This patch modifies the sequence to follow these steps:
* move the previous memory region to a new firewall register
* program the new memory aperture settings
* clean the non-overlapping memory
This patch also maps the non-overlapping memory as Device memory to
follow guidance from the arch. team.
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
Change-Id: I7cf6e05b2dd372103dc7229e37b1b3fc269a57ae
The Marvell Armada 37xx SoCs-based platforms contain a bit
awkward directory structure because the currently only one
supported PLAT and PLAT_FAMILY are the same. Modify the latter
to 'a3k' in order to improve it and keep plat/marvell/armada
tree more consistent:
plat/marvell/
├── armada
│ ├── a3k
│ │ ├── a3700
[...]
│ ├── a8k
│ │ ├── a70x0
[...]
Change-Id: I693a6ef88e6ce49a326a3328875c90bbc186066a
Signed-off-by: Marcin Wojtas <mw@semihalf.com>
* changes:
ddr: a80x0: add DDR 32-bit ECC mode support
ble: ap807: improve PLL configuration sequence
ble: ap807: clean-up PLL configuration sequence
ddr: a80x0: add DDR 32-bit mode support
plat: marvell: mci: perform mci link tuning for all mci interfaces
plat: marvell: mci: use more meaningful name for mci link tuning
plat: marvell: a8k: remove wrong or unnecessary comments
plat: marvell: ap807: enable snoop filter for ap807
plat: marvell: ap807: update configuration space of each CP
plat: marvell: ap807: use correct address for MCIx4 register
plat: marvell: add support for PLL 2.2GHz mode
plat: marvell: armada: make a8k_common.mk and mss_common.mk more generic
marvell: armada: add extra level in marvell platform hierarchy
This change implements SCMI channels for reading a SCMI message from a
shared memory and call the SCMI message drivers to route the message
to the target platform services.
SMT refers to the shared memory management protocol which is used
to get/put message/response in shared memory. SMT is a 28byte header
stating shared memory state and exchanged protocol data.
The processing entry for a SCMI message can be a secure interrupt
or fastcall SMCCC invocation.
SMT description in this implementation is based on the OP-TEE
project [1] itself based in the SCP-firmware implementation [2].
Link: [1] a58c4d706d
Link: [2] https://github.com/ARM-software/SCP-firmware.git
Change-Id: I416c7dab5c67954c6fe80bae8d8cdfdcda66873e
Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
Adds SCMI reset domain protocol support in the SCMI message drivers
as defined in SCMI specification v2.0 [1]. Not all the messages
defined in the specification are supported.
scmi_msg_get_rd_handler() sanitizes the message_id value
against any speculative use of reset domain ID as a index since by
SCMI specification, IDs are indices.
This implementation is based on the OP-TEE project implementation [2]
itself based on the SCP-firmware implementation [3] of the SCMI
protocol server side.
Link: [1] http://infocenter.arm.com/help/topic/com.arm.doc.den0056a/DEN0056A_System_Control_and_Management_Interface.pdf
Link: [2] 56a1f10ed9
Link: [3] https://github.com/ARM-software/SCP-firmware.git
Change-Id: If7cf13de40a815dedb40dcd5af8b6bb6725d9078
Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
Adds SCMI clock protocol support in the SCMI message drivers as
defined in SCMI specification v2.0 [1] for clock protocol messages.
Platform can provide one of the plat_scmi_clock_*() handler for the
supported operations set/get state/rate and others.
scmi_msg_get_clock_handler() sanitizes the message_id value
against any speculative use of clock ID as a index since by
SCMI specification, IDs are indices.
This implementation is based on the OP-TEE project implementation [2]
itself based on the SCP-firmware implementation [3] of the SCMI
protocol server side.
Link: [1] http://infocenter.arm.com/help/topic/com.arm.doc.den0056a/DEN0056A_System_Control_and_Management_Interface.pdf
Link: [2] a7a9e3ba71
Link: [3] https://github.com/ARM-software/SCP-firmware.git
Change-Id: Ib56e096512042d4f7b9563d1e4181554eb8ed02c
Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
This change introduces drivers to allow a platform to create a basic
SCMI service and register handlers for client request (SCMI agent) on
system resources. This is the first piece of the drivers: an entry
function, the SCMI base protocol support and helpers for create
the response message.
With this change, scmi_process_message() is the entry function to
process an incoming SCMI message. The function expect the message
is already copied from shared memory into secure memory. The message
structure stores message reference and output buffer reference where
response message shall be stored.
scmi_process_message() calls the SCMI protocol driver according to
the protocol ID in the message. The SCMI protocol driver will call
defined platform handlers according to the message content.
This change introduces only the SCMI base protocol as defined in
SCMI specification v2.0 [1]. Not all the messages defined
in the specification are supported.
The SCMI message implementation is derived from the OP-TEE project [2]
itself based on the SCP-firmware implementation [3] of the SCMI protocol
server side.
Link: [1] http://infocenter.arm.com/help/topic/com.arm.doc.den0056a/DEN0056A_System_Control_and_Management_Interface.pdf
Link: [2] ae8c806809
Link: [3] https://github.com/ARM-software/SCP-firmware/tree/v2.6.0
Change-Id: I639c4154a39fca60606264baf8d32452641f45e9
Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
Load address of tb_fw_config is incorrectly mentioned
in below device trees:
1. rdn1edge_fw_config.dts
2. tc0_fw_config.dts
Till now, tb_fw_config load-address is not being retrieved from
device tree and hence never exeprienced any issue for tc0 and
rdn1edge platform.
For tc0 and rdn1edge platform, Load-address of tb_fw_config should
be the SRAM base address + 0x300 (size of fw_config device tree)
Hence updated these platform's fw_config.dts accordingly to reflect
this load address change.
Change-Id: I2ef8b05d49be10767db31384329f516df11ca817
Signed-off-by: Manish V Badarkhe <Manish.Badarkhe@arm.com>