Statically allocate 'gpt_bitlock' array of fine-grained
'bitlock_t' data structures in arm_bl31_setup.c.
The amount of memory needed for this array is controlled
by 'RME_GPT_BITLOCK_BLOCK' build option and 'PLAT_ARM_PPS'
macro defined in platform_def.h which specifies the size
of protected physical address space in bytes.
'PLAT_ARM_PPS' takes values from 4GB to 4PB supported by
Arm architecture.
Change-Id: Icf620b5039e45df6828d58fca089cad83b0bc669
Signed-off-by: AlexeiFedorov <Alexei.Fedorov@arm.com>
GPT library function fill_l1_tbl() gets 'first' and
'last' parameters which are the start addresses of
the 1st and the last granules in the range to fill
L1 GPT table. When RME_GPT_MAX_BLOCK build option
is not 0, condition for 'while' loop should be
changed from 'first < last' to 'first <= last' in
the case of 'first' = 'last' when a single granule
is passed.
Change-Id: I9b49a78b5a2f7a01f51dbce43bd3f3cfbb458fa2
Signed-off-by: AlexeiFedorov <Alexei.Fedorov@arm.com>
This patch fixes fill_l1_tbl() function bug
for RME_GPT_MAX_BLOCK build option set to 0
disabling filling L1 tables with Contiguous
descriptors.
Change-Id: I3eedd6c1bb55b7c207bb3630d1ab2fda8f72eb17
Signed-off-by: AlexeiFedorov <Alexei.Fedorov@arm.com>
This patch adds support in GPT library for configuration
of the memory block size protected by one bit of 'bitlock'
structure. Build option 'RME_GPT_BITLOCK_BLOCK' defines the
number of 512MB blocks covered by each bit. This numeric
parameter must be a power of 2 and can take the values from
0 to 512. Setting this value to 0 chooses a single spinlock
for all GPT L1 table entries. The default value is set to 1
which corresponds to 512MB per bit.
Change-Id: I710d178072894a3ef40daebea701f74d19e8a3d7
Signed-off-by: AlexeiFedorov <Alexei.Fedorov@arm.com>
This patch adds support for large GPT mappings using
Contiguous descriptors. The maximum size of supported
contiguous block in MB is defined in RME_GPT_MAX_BLOCK
build parameter and takes values 0, 2, 32 and 512 and
by default set to 2 in make_helpers/defaults.mk.
Setting RME_GPT_MAX_BLOCK value to 0 disables use of
Contiguous descriptors.
Function gpt_tlbi_by_pa_ll() and its declaration
are removed from lib/aarch64/misc_helpers.S and
include/arch/aarch64/arch_helpers.h, because the
GPT library now uses tlbirpalos_xxx() functions.
Change-Id: Ia9a59bde1741c5666b4ca1de9324e6dfd6f734eb
Signed-off-by: AlexeiFedorov <Alexei.Fedorov@arm.com>
This patch modifies GPT library comments and makes
logging messages consistent with PRIx64 usage and
TF-A format used in other modules.
Minor changes are made to make the code compliant
with MISRA C requirements.
Change-Id: Ic40e1b7ac43cd9602819698d00e1ce3a8c7183ce
Signed-off-by: AlexeiFedorov <Alexei.Fedorov@arm.com>
This patch removes 'gpt_' prefix from the
names of static functions for better code
readability.
Change-Id: I0398b55047a73209da598b708240fcba47c779f7
Signed-off-by: AlexeiFedorov <Alexei.Fedorov@arm.com>
Use cache clean and invalidate of data and allocation tags by PA to PoPA
maintenance instruction (dc cigdpapa) in the GPT library upon changing
the PAS for a memory region. This is required to flush allocation tags
when MTE2 (and above) is implemented.
Signed-off-by: Olivier Deprez <olivier.deprez@arm.com>
Change-Id: I4b70afb297f693b1d446839607922c47111ce063
This patch fixes compilation error for gpt_init_l0_tables()
function in lib/gpt_rme/gpt_rme.c reported by GCC 13.0.0:
"gpt_rme/gpt_rme.c:765:5: error: conflicting types for
'gpt_init_l0_tables' due to enum/integer mismatch;
have 'int(unsigned int, uintptr_t, size_t)'
{aka 'int(unsigned int, long unsigned int, long unsigned int)'}"
Signed-off-by: AlexeiFedorov <Alexei.Fedorov@arm.com>
Change-Id: I38f28be290337e7d37d59b52cad7bde5b96b8d51
Since GPC control register fields are permitted to be cached in a TLB,
invalidate TLB after setting fields to ensure future checks are using
the updated values.
Signed-off-by: Kathleen Capella <kathleen.capella@arm.com>
Change-Id: I95630b40b673363bbf74da2705deca03089fff3a
This patch adds documentation for the missing RMM-EL3
runtime services:
* RMM_RMI_REQ_COMPLETE
* RMM_GTSI_DELEGATE
* RMM_GTSI_UNDELEGATE
This patch also fixes a couple of minor bugs on return codes
for delegate/undelegate internal APIs.
Signed-off-by: Javier Almansa Sobrino <javier.almansasobrino@arm.com>
Change-Id: Ic721005e7851e838eebaee7865ba78fadc3309e4
The previous delegating/undelegating sequence was incorrect as per the
specification DDI0615, "Architecture Reference Manual Supplement, The
Realm Management Extension (RME), for Armv9-A" Sections A1.1.1 and
A1.1.2
Off topic:
- cleaning the gpt_is_gpi_valid and gpt_check_pass_overlap
Change-Id: Idb64d0a2e6204f1708951137062847938ab5e0ac
Signed-off-by: Robert Wakim <robert.wakim@arm.com>
Under certain configurations of PPS and L0GPTSZ a macro could result in
a right shift by 64 bits. This patch removes that possibility by
limiting the total size of each shift to the maximum width of the L0 or
L1 index field in a physical address. In addition, it adds more detail
about how these values are calculated.
Signed-off-by: John Powell <john.powell@arm.com>
Change-Id: Ie71c8e6f922a5bb522a6169701bfc36fc99f765a
sha 4ce3e99a3 introduced printf format specifiers for fixed width
types, which uses PRI*64 instead of "ll" for 64 bit values.
Signed-off-by: Manish Pandey <manish.pandey2@arm.com>
Change-Id: I30472411467061d58cc6ee22407ed3bad2552751
This patch adds necessary barriers after GPT entries are modified
so that the writes are observed correctly by the GPC hardware.
The shareability of GPC fetches are changed from OSH to ISH so
that they align with the shareability of MMU attributes for the
region. Thus by adding a dsbishst() between the GPT L1 entry
write as part of granule migration at runtime, we can now remove
the clean cache maintenance operation (CMO) for that region.
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
Signed-off-by: Robert Wakim <robert.wakim@arm.com>
Change-Id: Ib9e405b106f0db95c7fbdb26773c0ed41663a5b4
This patch updates and refactors the GPT library and fixes bugs.
- Support all combinations of PGS, PPS, and L0GPTSZ parameters.
- PPS and PGS are set at runtime, L0GPTSZ is read from GPCCR_EL3.
- Use compiler definitions to simplify code.
- Renaming functions to better suit intended uses.
- MMU enabled before GPT APIs called.
- Add comments to make function usage more clear in GPT library.
- Added _rme suffix to file names to differentiate better from the
GPT file system code.
- Renamed gpt_defs.h to gpt_rme_private.h to better separate private
and public code.
- Renamed gpt_core.c to gpt_rme.c to better conform to TF-A precedent.
Signed-off-by: John Powell <john.powell@arm.com>
Change-Id: I4cbb23b0f81e697baa9fb23ba458aa3f7d1ed919
