core: add CFG_BOOT_MEM and boot_mem_*() functions

Adds CFG_BOOT_MEM to support stack-like memory allocations during boot
before a heap has been configured.

Signed-off-by: Jens Wiklander <jens.wikl

core: add CFG_BOOT_MEM and boot_mem_*() functions

Adds CFG_BOOT_MEM to support stack-like memory allocations during boot
before a heap has been configured.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: add VCORE_FREE_{PA,SZ,END_PA}

Add VCORE_FREE_{PA,SZ,END_PA} defines to identify the unused and free
memory range at the end of TEE_RAM_START..(TEE_RAM_START +
TEE_RAM_VA_SIZE).

VCORE_FREE_SZ

core: add VCORE_FREE_{PA,SZ,END_PA}

Add VCORE_FREE_{PA,SZ,END_PA} defines to identify the unused and free
memory range at the end of TEE_RAM_START..(TEE_RAM_START +
TEE_RAM_VA_SIZE).

VCORE_FREE_SZ is 0 in a pager configuration since all the memory is
used by the pager.

The VCORE_FREE range is excluded from the TEE_RAM_RW area for
CFG_NS_VIRTUALIZATION=y and instead put in a separate NEX_RAM_RW area.
This makes each partition use a bit less memory and leaves the
VCORE_FREE range available for the Nexus.

The VCORE_FREE range is added to the TEE_RAM_RW area for the normal
configuration with CFG_NS_VIRTUALIZATION=n and CFG_WITH_PAGER=n. It's in
practice unchanged behaviour in this configuration.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: thread: clarify thread_rpc_alloc_*_payload() inline description

The thread_rpc_alloc_*_payload APIs are consistent with the thread_shm_type
enumerations, as seen in the alloc_shm function. How

core: thread: clarify thread_rpc_alloc_*_payload() inline description

The thread_rpc_alloc_*_payload APIs are consistent with the thread_shm_type
enumerations, as seen in the alloc_shm function. However, the name and
comments of thread_rpc_alloc_payload might be misleading, suggesting it is
a universal API. To maintain backward compatibility, only the code comments
have been updated for clarity.

Signed-off-by: Yitong Cheng <santongding@foxmail.com>
Acked-by: Etienne Carriere <etienne.carriere@foss.st.com>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

show more ...

core: kernel: dt: add and use fdt_reg_info()

Implement fdt_reg_info() instead of fdt_reg_base_address() and
fdt_reg_size() to optimize look up in the DT due to finding parent node.

Signed-off-by: E

core: kernel: dt: add and use fdt_reg_info()

Implement fdt_reg_info() instead of fdt_reg_base_address() and
fdt_reg_size() to optimize look up in the DT due to finding parent node.

Signed-off-by: Etienne Carriere <etienne.carriere@foss.st.com>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

show more ...

core: dt: cache embedded DTB node information

Optimize OP-TEE boot time regarding parsing of the embedded DTB
content when using libfdt. The library functions fdt_parent_offset()
and fdt_node_offset

core: dt: cache embedded DTB node information

Optimize OP-TEE boot time regarding parsing of the embedded DTB
content when using libfdt. The library functions fdt_parent_offset()
and fdt_node_offset_by_phandle() are not very efficient since they
parse the DTB from root node to target node to look up for,
respectively, the node offset of a node parent and the node offset
related to a node phandle. Helper functions fdt_reg_base_address()
and fdt_reg_size() are also affected since they are based on
fdt_parent_offset() to find the #address-cells and #size-cells
properties of a node parent.

Optimize this by parsing the embedded DT once and caching node
information (parent node, phandle value, parent node #address-cells
and #size-cells values) in a array. Parse the array instead of the
DT for find these information.

We made few tests to use bisection or hash tables for look up the
information in the cache array. The gain was very small, likely due
to the number of DT node involved in the platform is relatively small
(only several hundreds or nodes).

This feature is enabled upon configuration switch CFG_DT_CACHED_NODE_INFO.

Signed-off-by: Etienne Carriere <etienne.carriere@foss.st.com>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

show more ...

core: pm: add unregister_pm_cb()

Add unregister_pm_cb() API function and its helper variants to
allow unregistering a PM callback entry. This can be needed for
example in the GPIO framework where gp

core: pm: add unregister_pm_cb()

Add unregister_pm_cb() API function and its helper variants to
allow unregistering a PM callback entry. This can be needed for
example in the GPIO framework where gpio_put() can release a GPIO
that a driver no more consumed. In case a PM callback was previously
registered for such a GPIO, consumer driver needs mean to unregister
it.

This change implies that the PM callbacks list is protected from
concurrent accesses hence add a lock for that purpose.

Signed-off-by: Gatien Chevallier <gatien.chevallier@foss.st.com>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>
Acked-by: Jerome Forissier <jerome.forissier@linaro.org>

show more ...

core: timeout_elapsed(): fix expiration threshold

Fix timeout threshold that should return once the time expires,
not 1 tick after it has expired. This change makes udelay(0) to
immediately return a

core: timeout_elapsed(): fix expiration threshold

Fix timeout threshold that should return once the time expires,
not 1 tick after it has expired. This change makes udelay(0) to
immediately return and not wait a generic timer counter increment.

This change fixes an issue where a driver stops the counter feeding
Arm generic timer counter and indirectly calls IO_READ32_POLL_TIMEOUT()
with a delay of 0us. It that case, since counter never increments, the
udelay(0) call in IO_READ32_POLL_TIMEOUT() never returns while we expect
the macro to endlessly poll (as timeout would never be detected) until
poll condition is met or system watchdog is triggered.

Signed-off-by: Etienne Carriere <etienne.carriere@foss.st.com>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>

show more ...

core: add THREAD_SHM_CACHE_USER_RPMB

Add an enum to cache RPMB shared memory allocations.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Jerome Forissier <jerome.forissier@l

core: add THREAD_SHM_CACHE_USER_RPMB

Add an enum to cache RPMB shared memory allocations.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: open-code thread_init_stack()

The implementations of thread_init_stack() are identical and trivial for
both arm and riscv. So simplify code further and open-code it where it's
called from in c

core: open-code thread_init_stack()

The implementations of thread_init_stack() are identical and trivial for
both arm and riscv. So simplify code further and open-code it where it's
called from in core/kernel/thread.c.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Alvin Chang <alvinga@andestech.com>
Tested-by: Alvin Chang <alvinga@andestech.com>

show more ...

core: arm: introduce boot_init_primary_final()

Introduce boot_init_primary_final() and move the call to
call_finalcalls() into that function.

This is needed in later patches to enabled PAUTH before

core: arm: introduce boot_init_primary_final()

Introduce boot_init_primary_final() and move the call to
call_finalcalls() into that function.

This is needed in later patches to enabled PAUTH before
boot_init_primary_final() is called.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: mm: introduce struct memory_map

Introduce struct memory_map to keep track of the array of struct
tee_mmap_region, covering number of used entries and number of allocated
entries.

core_mmap_is

core: mm: introduce struct memory_map

Introduce struct memory_map to keep track of the array of struct
tee_mmap_region, covering number of used entries and number of allocated
entries.

core_mmap_is_end_of_table() and MEM_AREA_END are now unused so remove
them.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Jerome Forissier <jerome.forissier@linaro.org>

show more ...

kernel: pm: add suspend type hint

When entering suspend type, the suspend level is platform dependent and
can be set to various values depending on these platform. In order to
allow platforms settin

kernel: pm: add suspend type hint

When entering suspend type, the suspend level is platform dependent and
can be set to various values depending on these platform. In order to
allow platforms setting it in a generic way when entering suspend,
reserve some bits in the suspend/resume hint to pass this information.
Driver can then used it in a platform independent way to execute
specific code depending on it.

Signed-off-by: Thomas Perrot <thomas.perrot@bootlin.com>
Acked-by: Jerome Forissier <jerome.forissier@linaro.org>

show more ...

core: kernel: Fix checking register convention r1/x1 value of transfer_list

According to recently firmware handsoff spec [1]'s
"Register usage at handoff boundary", Transfer List's signature value
w

core: kernel: Fix checking register convention r1/x1 value of transfer_list

According to recently firmware handsoff spec [1]'s
"Register usage at handoff boundary", Transfer List's signature value
was changed from 0x40_b10b (3 bytes) to 4a0f_b10b (4 bytes).

As updating of TL's signature, register value of x1/r1 should be:

In aarch32's r1 value should be
R1[23:0]: set to the TL signature (4a0f_b10b->masked value: 0f_b10b)
R1[31:24]: version of the register convention == 1
and
In aarch64's x1 value should be
X1[31:0]: set to the TL signature (4a0f_b10b)
X1[39:32]: version of the register convention == 1
X1[63:40]: MBZ
(See the [2] and [3]).

Therefore, it requires to separate mask and shift value for register
convention version field when checking each r1/x1 value.

This patch fix two problems:
1. breaking X1 value with updated specification in aarch64
- change of length of signature field.

2. previous error value set in R1 in arm32.
- length of signature should be 24, but it uses 32bit signature.

This patch is a breaking change. It works only TF-A is updated.

Link: https://github.com/FirmwareHandoff/firmware_handoff [1]
Link: https://github.com/FirmwareHandoff/firmware_handoff/issues/32 [2]
Link: https://github.com/FirmwareHandoff/firmware_handoff/commit/5aa7aa1d3a1db75213e458d392b751f0707de027 [3]
Fixes: 508e2476b232 ("core: update transfer list header and signature")
Signed-off-by: Levi Yun <yeoreum.yun@arm.com>
Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>

show more ...

core: kernel: dt_driver: fix nodes reference passed in dt_pargs

Correct the DT node references passed in struct dt_pargs. Field
phandle_node is expected to be the node offset of the provider driver

core: kernel: dt_driver: fix nodes reference passed in dt_pargs

Correct the DT node references passed in struct dt_pargs. Field
phandle_node is expected to be the node offset of the provider driver
but dt_driver_device_from_node_idx_prop_phandle() implementation passes
node offset -1, dt_driver_device_from_parent() passes the consumer
node offset and dt_driver_device_from_node_idx_prop() may pass the
phandle node offset from a previous index in the parsed DT node property.

Fix that by adding field consumer_node to reference the node offset of
the consumer device and fixing phandle_node where applicable and
update atmel_i2c and stm32_i2c drivers to get the I2C device address
from the consumer_node field instead of the phandle_node that is now
fixed.

Signed-off-by: Etienne Carriere <etienne.carriere@foss.st.com>
Acked-by: Gatien Chevallier <gatien.chevallier@foss.st.com>

show more ...

core: make generic notifications virtualization-aware

Makes the generic notification handling aware of virtualization. Atomic
events are delivered with a guest_id parameter and asynchronous
notifica

core: make generic notifications virtualization-aware

Makes the generic notification handling aware of virtualization. Atomic
events are delivered with a guest_id parameter and asynchronous
notifications are started per guest_id.

struct notif_data is added as guest specific data to be able to track if
notifications are started for a guest.

While this patch compiles it doesn't work as intended without patches
handling the ABI specific side of things.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: virt: add guest specific data

Add virt_add_guest_spec_data() and virt_get_guest_spec_data() to handle
guest specific data.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by:

core: virt: add guest specific data

Add virt_add_guest_spec_data() and virt_get_guest_spec_data() to handle
guest specific data.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: virt: support iterating over partitions

Add support to iterate over partitions using virt_next_guest().

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <e

core: virt: support iterating over partitions

Add support to iterate over partitions using virt_next_guest().

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: virt: get guest ID of a guest partition

Add the function virt_get_guest_id() to return the guest ID of a guest
partition.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: E

core: virt: get guest ID of a guest partition

Add the function virt_get_guest_id() to return the guest ID of a guest
partition.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: virt: get current guest partition

Add the function virt_get_current_guest() to get a pointer to the
current guest partition. The returned guest partition has its reference
counter increased wh

core: virt: get current guest partition

Add the function virt_get_current_guest() to get a pointer to the
current guest partition. The returned guest partition has its reference
counter increased which must be restored with a matching call to
virt_put_guest() when the caller is done with the guest partition.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: virt: get guest partition by guest ID

Add the function virt_get_guest() to get the pointer to a guest
partition with a certain guest ID. The returned guest partition has its
reference counter

core: virt: get guest partition by guest ID

Add the function virt_get_guest() to get the pointer to a guest
partition with a certain guest ID. The returned guest partition has its
reference counter increased which must be restored with a matching call
to virt_put_guest() when the caller is done with the guest partition.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: fix race condition on TA/PTA/StMM context loading

Fix race condition on creation of a context for single instance TAs,
PTAs or StMM application. Such race condition could occur and lead to
dup

core: fix race condition on TA/PTA/StMM context loading

Fix race condition on creation of a context for single instance TAs,
PTAs or StMM application. Such race condition could occur and lead to
duplicated contexts if connected close enough that they are created
after tee_ta_init_session() calls tee_ta_init_session_with_context()
and before the context are added in the centralized context list.

This is realized by keeping tee_ta_mutex held while tee_ctxes list is
looked up for matching a context or a new context is added into that
list with its .is_initializing field activated. For that purpose
user TA and StMM application initialization function are split in
2 functions, the 2nd one used to finalizes the context creation
started in the 1st function.

By the way, add inline description comments and fix indentation issues
in uaer_ta.h and remove the inline comment in pseudo_ta.c that refers
to TA loading whereas the function relates to PTA contexts creation.

Closes: https://github.com/OP-TEE/optee_os/issues/6801
Suggested-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>
Signed-off-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: add mutex_is_locked() helper function

Add mutex_is_locked() helper function to return whether a mutex is
locked or not. This helper function must be use with care since it
does not guarantee t

core: add mutex_is_locked() helper function

Add mutex_is_locked() helper function to return whether a mutex is
locked or not. This helper function must be use with care since it
does not guarantee that the mutex is held by the executing thread.

Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>
Signed-off-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: make is_initializing field a generic service context

Replace fields is_initializing from struct user_mode_ctx and
struct stmm_ctx with a common new field is_initialing in generic
struct tee_ta

core: make is_initializing field a generic service context

Replace fields is_initializing from struct user_mode_ctx and
struct stmm_ctx with a common new field is_initialing in generic
struct tee_ta_ctx so that it can be used in generic context loading
functions for contexts which initialization is done with tee_ta_mutex
released.

Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>
Signed-off-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: kernel: change DT getters to use manifest DT

This change get_dt() and get_secure_dt() to use manifest DT
as an alternative if there is no embedded DT or external DT.

Signed-off-by: Sungbae Yo

core: kernel: change DT getters to use manifest DT

This change get_dt() and get_secure_dt() to use manifest DT
as an alternative if there is no embedded DT or external DT.

Signed-off-by: Sungbae Yoo <sungbaey@nvidia.com>
Reviewed-by: Joakim Bech <joakim.bech@linaro.org>
Acked-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: kernel: move manifest DT functions to dt.c

This moves functions to handle manifest DT, used as a SP (secure partition)
manifest via FF-A interface, to core/kernel/dt.c and their definitions
to

core: kernel: move manifest DT functions to dt.c

This moves functions to handle manifest DT, used as a SP (secure partition)
manifest via FF-A interface, to core/kernel/dt.c and their definitions
to dt.h.
manifest DT is dependent with CORE_FFA that depends on CFG_DT, as the same
as dt.c.

Signed-off-by: Sungbae Yoo <sungbaey@nvidia.com>
Acked-by: Jerome Forissier <jerome.forissier@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...