core: rename to core_mmu_init_phys_mem()

Rename core_mmu_init_ta_ram() to core_mmu_init_phys_mem() for a more
accurate name of the function.

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

core: rename to core_mmu_init_phys_mem()

Rename core_mmu_init_ta_ram() to core_mmu_init_phys_mem() for a more
accurate name of the function.

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

show more ...

core: replace tee_mm_sec_ddr with phys_mem functions

Replace the tee_mm_sec_ddr mm pool with the phys_mem functions. This
doesn't change the behaviour.

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

core: replace tee_mm_sec_ddr with phys_mem functions

Replace the tee_mm_sec_ddr mm pool with the phys_mem functions. This
doesn't change the behaviour.

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

show more ...

core: add phys_mem allocation functions

Add nex_phys_mem and phys_mem allocation functions. These functions are
intended to replace the previous calls to tee_mm functions on with the
virt_mapper_poo

core: add phys_mem allocation functions

Add nex_phys_mem and phys_mem allocation functions. These functions are
intended to replace the previous calls to tee_mm functions on with the
virt_mapper_pool or tee_mm_sec_ddr as arguments.

The pool of physical memory is divided into two parts, core and ta. All
physical TA memory allocations are done from the core pool if a ta pool
isn't added. This might be the case if core and ta physical memory
resides in the same physical memory range.

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

show more ...

core: mm: add core_mmu_for_each_map()

Add core_mmu_for_each_map() to iterate over all memory regions, struct
tee_mmap_region.

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

core: mm: add core_mmu_for_each_map()

Add core_mmu_for_each_map() to iterate over all memory regions, struct
tee_mmap_region.

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

show more ...

core: add_phys_mem(): fix mergeable physical memory

The test in add_phys_mem() to see if two physical memory ranges can be
merged only checks for overlapping memory ranges, but consecutive ranges
ar

core: add_phys_mem(): fix mergeable physical memory

The test in add_phys_mem() to see if two physical memory ranges can be
merged only checks for overlapping memory ranges, but consecutive ranges
are not detected even if they can be merged. Fix this by also checking
if the byte after the lowest range matches the beginning of the next
range.

The resulting merged entry might be mergeable with the previous or next
entry, so add checks for that and merge if possible.

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

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 ...

core: mobj_phys_alloc(): remove unused memory types

In mobj_phys_alloc() remove support for the two unused memory types
CORE_MEM_TEE_RAM and CORE_MEM_TA_RAM.

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

core: mobj_phys_alloc(): remove unused memory types

In mobj_phys_alloc() remove support for the two unused memory types
CORE_MEM_TEE_RAM and CORE_MEM_TA_RAM.

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

show more ...

core: remove unused mobj_sec_ddr

Remove the unused global variable mobj_sec_ddr.

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

core: remove unused mobj_sec_ddr

Remove the unused global variable mobj_sec_ddr.

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

show more ...

core: mm: ensure all pager VA space is mapped with small pages

Fix can_map_at_level() to ensure all memory areas related the pager
pageable virtual memory are mapped with small pages. This change
fi

core: mm: ensure all pager VA space is mapped with small pages

Fix can_map_at_level() to ensure all memory areas related the pager
pageable virtual memory are mapped with small pages. This change
fixes an issue found when the pager physical RAM ends on a section
boundary (e.g. 512MB or 2MB on LPAE case) making the virtual memory
mapping above that boundary to be prepared with pgdir or wider MMU tables
while pager implementation expects 4kB page MMU tables.

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

show more ...

core: mm: strip the tag of VA when check_pa_matches_va

When CFG_TEE_CORE_DEBUG=y and CFG_MEMTAG=y, core will crash, if there are
some modules call virt_to_phys() or core_vbuf_is() and so on.
Because

core: mm: strip the tag of VA when check_pa_matches_va

When CFG_TEE_CORE_DEBUG=y and CFG_MEMTAG=y, core will crash, if there are
some modules call virt_to_phys() or core_vbuf_is() and so on.
Because the MM can not find map by the tagged VA, so strip the tag of
VA when check_pa_matches_va().

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

show more ...

core: mm: core_mmu: skip unmapped regions when searching by PA

Currently, OP-TEE OS on RISC-V platforms uses the external
device-tree located at 'Domain1 Next Arg1' passed by previous
boot stages (i

core: mm: core_mmu: skip unmapped regions when searching by PA

Currently, OP-TEE OS on RISC-V platforms uses the external
device-tree located at 'Domain1 Next Arg1' passed by previous
boot stages (i.e. U-Boot SPL and OpenSBI).
For example:

Domain1 Name : trusted-domain
...
Domain1 Next Address : 0x0000000010000000 // OP-TEE OS
> Domain1 Next Arg1 : 0x000000000187f760 // device-tree
Domain1 Next Mode : S-mode

In this case, 0x0187f760 lies in SHM_VASPACE, which is not yet
mapped and not intended to load an external DT:

type TEE_RAM_RX va 0x10000000..0x10092fff pa 0x10000000..0x10092fff
type TEE_RAM_RW va 0x10093000..0x101fffff pa 0x10093000..0x101fffff
type RES_VASPACE va 0x10200000..0x10bfffff pa 0x00000000..0x009fffff
> type SHM_VASPACE va 0x10c00000..0x12bfffff pa 0x00000000..0x01ffffff
type TA_RAM va 0x12c00000..0x139fffff pa 0x10200000..0x10ffffff
type IO_SEC va 0x13a00000..0x13bfffff pa 0xf0200000..0xf03fffff

To address this issue, add a memory region attribute check to
skip such regions. In this way, when init_external_dt() calls
core_mmu_get_type_by_pa(), it can properly return MEM_AREA_MAXTYPE
(i.e. valid region not found) and map a MEM_AREA_EXT_DT region.

Note that this bug cannot be reproduced on QEMU virt machine,
as its memory regions have no overlapping with the external DT.

Signed-off-by: Yu Chien Peter Lin <peterlin@andestech.com>
Reviewed-by: Alvin Chang <alvinga@andestech.com>
Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>

show more ...

core: mm: add CFG_PGT_CACHE_ENTRIES

Add CFG_PGT_CACHE_ENTRIES to allow platforms to customize the page
table cache size. This is needed for example when a platform is to
support very large TAs of se

core: mm: add CFG_PGT_CACHE_ENTRIES

Add CFG_PGT_CACHE_ENTRIES to allow platforms to customize the page
table cache size. This is needed for example when a platform is to
support very large TAs of several dozen of Mbytes of private memory
(code/data).

Move PGT_CACHE_SIZE macro definition from pgt_cache.h to pgt_cache.c
since it is used only in that source file.

By the way, fix pgt_cache.h layout to have header files includes first
followed by macro definitions.

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

show more ...

core: rename tee_mm_shm to core_virt_shm_pool

Rename tee_mm_shm to core_virt_shm_pool to make it clear that it handles
virtual memory allocations for shared memory.

Signed-off-by: Jens Wiklander <j

core: rename tee_mm_shm to core_virt_shm_pool

Rename tee_mm_shm to core_virt_shm_pool to make it clear that it handles
virtual memory allocations for shared memory.

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: rename tee_mm_vcore to core_virt_mem_pool

Rename tee_mm_vcore to core_virt_mem_pool to make it clear that it
handles virtual memory allocations.

Signed-off-by: Jens Wiklander <jens.wiklander@

core: rename tee_mm_vcore to core_virt_mem_pool

Rename tee_mm_vcore to core_virt_mem_pool to make it clear that it
handles virtual 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: move tee_mm_{sec_ddr,vcore,shm}

Move tee_mm_sec_ddr, tee_mm_vcore, and tee_mm_shm into core/mm/core_mmu.c.

Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Jerome Foriss

core: move tee_mm_{sec_ddr,vcore,shm}

Move tee_mm_sec_ddr, tee_mm_vcore, and tee_mm_shm into core/mm/core_mmu.c.

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: move mobj_dyn_shm.c to mm

To enable RISC-V platforms to utilize dynamic shared memory,
relocate the non-architecture-specific mobj_dyn_shm.c to the
mm directory.

Signed-off-by: Alvin Chan

core: mm: move mobj_dyn_shm.c to mm

To enable RISC-V platforms to utilize dynamic shared memory,
relocate the non-architecture-specific mobj_dyn_shm.c to the
mm directory.

Signed-off-by: Alvin Chang <alvinga@andestech.com>
Reviewed-by: Yu Chien Peter Lin <peterlin@andestech.com>
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>

show more ...

core: mmu: fix memory regions found from ff-a manifest

Fix the 6th parameter of add_phys_mem() in collect_device_mem_ranges()
that has to be the size of memory region and not the end address of the

core: mmu: fix memory regions found from ff-a manifest

Fix the 6th parameter of add_phys_mem() in collect_device_mem_ranges()
that has to be the size of memory region and not the end address of the
region.

Fixes: 72a6827a6353 ("core: arm: SPMC obtain device memory info from DTB")
Signed-off-by: Sungbae Yoo <sungbaey@nvidia.com>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

show more ...

core: mmu: add MEM_AREA_ROM_SEC in check_mem_map()

Handle MEM_AREA_ROM_SEC in check_mem_map() switch case to prevent OP-TEE
core from panicking when such area is mapped.

Fixes: fc7e0cc38b99 ("core:

core: mmu: add MEM_AREA_ROM_SEC in check_mem_map()

Handle MEM_AREA_ROM_SEC in check_mem_map() switch case to prevent OP-TEE
core from panicking when such area is mapped.

Fixes: fc7e0cc38b99 ("core: MEM_AREA_ROM_SEC maps secure read only cached memory")
Signed-off-by: Thomas Bourgoin <thomas.bourgoin@foss.st.com>
Reviewed-by: Etienne Carriere <etienne.carriere@foss.st.com>

show more ...

core: add is_nexus() and refactor is_unpaged()

Add the function is_nexus() to tell if an address is an address
available to the nexus when non-secure virtualization is enabled
(CFG_NS_VIRTUALIZATION

core: add is_nexus() and refactor is_unpaged()

Add the function is_nexus() to tell if an address is an address
available to the nexus when non-secure virtualization is enabled
(CFG_NS_VIRTUALIZATION=y). The function is stubbed to return true for
non-null arguments if non-secure virtualization isn't enabled, else
false.

Make the argument for is_unpaged() const void * for consistency with the
is_nexus() function. The stubbed version of is_unpaged() when paging
isn't enabled is updated to perform a NULL check on the argument.

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

show more ...

core: mmu: assert pointer to manifest device tree

Assert pointer to manifest device tree before using it in
fdt_node_offset_by_compatible().

Signed-off-by: Clement Faure <clement.faure@nxp.com>
Rev

core: mmu: assert pointer to manifest device tree

Assert pointer to manifest device tree before using it in
fdt_node_offset_by_compatible().

Signed-off-by: Clement Faure <clement.faure@nxp.com>
Reviewed-by: Jerome Forissier <jerome.forissier@linaro.org>

show more ...

core: remove unnecessary includes

Remove unnecessary includes.

Signed-off-by: Clement Faure <clement.faure@nxp.com>
Acked-by: Jerome Forissier <jerome.forissier@linaro.org>
Acked-by: Jens Wiklander

core: remove unnecessary includes

Remove unnecessary includes.

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

show more ...

tree wide: clarify PTA ABI

Explicitly describe the statistics PTA ABI using dedicated inline
comments and moving struct pta_stats_ta and struct pta_stats_alloc
from respectively tee_ta_manager.c and

tree wide: clarify PTA ABI

Explicitly describe the statistics PTA ABI using dedicated inline
comments and moving struct pta_stats_ta and struct pta_stats_alloc
from respectively tee_ta_manager.c and malloc.h to pta_stats.h
header file.

For that purpose, define ALLOC_ID_* macros to identify each allocator
one can query information from through the statistics PTA API.

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

show more ...

tree wide: rename struct malloc_stats

Rename struct malloc_stats to struct pta_stats_alloc to prepare
the creation of pta_stats.h header file that will define the
whole statistics PTA API and ABI, l

tree wide: rename struct malloc_stats

Rename struct malloc_stats to struct pta_stats_alloc to prepare
the creation of pta_stats.h header file that will define the
whole statistics PTA API and ABI, located in libutee/include/
as already done for all other PTAs.

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

show more ...

core: mmu: fix nsec ddr discovery regarding MEM_AREA_RAM_NSEC

Memory registered as MEM_AREA_RAM_NSEC can be part of the non-secure
shared memory. This change fixes core_mmu_set_discovered_nsec_ddr()

core: mmu: fix nsec ddr discovery regarding MEM_AREA_RAM_NSEC

Memory registered as MEM_AREA_RAM_NSEC can be part of the non-secure
shared memory. This change fixes core_mmu_set_discovered_nsec_ddr()
to not check such memory area do not overlap with non-secure RAM.
This is no issue physical pages from MEM_AREA_RAM_NSEC memory are also
handled and mapped as dynamic non-secure memory by OP-TEE.

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

show more ...

core: mm: use MEM_AREA_END to detect mapping area end

Change verify_special_mem_areas() prototype to remove the 'len'
argument that is useless since end of the static mapping areas array
is already

core: mm: use MEM_AREA_END to detect mapping area end

Change verify_special_mem_areas() prototype to remove the 'len'
argument that is useless since end of the static mapping areas array
is already defined by a MEM_AREA_END cell.

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

show more ...