Amit Nagal [Wed, 15 Feb 2023 13:13:55 +0000 (18:43 +0530)]
feat(zynqmp): add support for custom sip service
Add support for custom sip service.
Bare minimum implementation for custom_smc_handler is provided
by platform. Actual definition for custom_smc_handler will be provided
by custom pkg.
This feature is going to be used by external libraries. For example
for checking it's status.
The similar approach is also used by qti/{sc7180,sc7280} platforms
by providing a way to select QTISECLIB_PATH.
This code is providing a generic way how to wire any code
via custom $(CUSTOM_PKG_PATH)/custom_pkg.mk makefile with also an
option to wire custom SMC. SMC functionality depends on "package".
Change-Id: Icedffd582f76f89fc399b0bb2e05cdaee9b743a0 Signed-off-by: Amit Nagal <amit.nagal@amd.com>
Michal Simek [Wed, 15 Feb 2023 11:39:22 +0000 (12:39 +0100)]
fix(versal): check smc_fid 23:16 bits
23:16 bits when they gets to SMC handler should be all zeros but be
inside SIP Service Calls range which is defined as 0x82000000-0x8200ffff
or 0xc2000000-0xc200ffff. That's why make sure that code won't handle
any SMCs in SIP range out of predefined range.
Also fix MASK values to check the same range for PM/IPI calls to make
sure that masking covers all required bits including 23:16. Bits 15:12
are used for different class of requests.
Change-Id: I9d3e91aa521d6bb90f6b15b71ff1e89fa77ee379 Signed-off-by: Michal Simek <michal.simek@amd.com>
Michal Simek [Tue, 14 Feb 2023 12:10:29 +0000 (13:10 +0100)]
fix(zynqmp): check smc_fid 23:16 bits
23:16 bits when they gets to SMC handler should be all zeros but be
inside SIP Service Calls range which is defined as 0x82000000-0x8200ffff
or 0xc2000000-0xc200ffff. That's why make sure that code won't handle
any SMCs in SIP range out of predefined range. Because EM SMC is out of
this range already on this SOC check it after it (EMC SMC will be
handled separately).
Also fix MASK values to check the same range for PM/IPI/EM calls to make
sure that masking covers all required bits including 23:16. Bits 15:12
are used for different class of requests.
Change-Id: If23ac769c91d206e47758aeaa1f14e8b9c3dc7bb Signed-off-by: Michal Simek <michal.simek@amd.com>
In the ZynqMP, 0x36 EEMI API ID is used for PM_FPGA_GET_VERSION and 0x37
is used for PM_FPGA_GET_FEATURE_LIST. The same ID numbers in the Versal
are used for PM_ADD_SUBSYSTEM and PM_DESTROY_SUBSYSTEM and it leads to
the EEMI API ID conflict between the platforms. To fix this issue this
patch updates the PM_FPGA_GET_VERSION and PM_FPGA_GET_FEATURE_LIST EEMI
API ID's to 0x48 and 0x49.
In linux zynqmp_pm_fpga_get_version() and
zynqmp_pm_fpga_get_feature_list() API's are uses PM_FPGA_GET_VERSION
and PM_FPGA_GET_FEATURE_LIST to get the xilfpga version and
xilfpga-supported feature list info. These API's are called only in
zynqmp-fpga.c as part of the probe. In case of this caller API's are
failed it will fall to the default feature list and this default
feature list is same as latest xilfpga-supported feature list (No new
feature was added in the xilfpga after adding these APIs). So, these
updated IDs will not cause any functional issues between Linux, TF-A,
and firmware components.
Signed-off-by: Nava kishore Manne <nava.kishore.manne@amd.com>
Change-Id: I14d974dd44651681ecbf726ad8b6940e1850cbec
Michal Simek [Wed, 8 Feb 2023 12:34:47 +0000 (13:34 +0100)]
fix(versal): sync location based on IPI_ID macros
IPI_ID_* macros available at include/plat_ipi.h are using PMC/APU/RPU0..
order which is not how versal_ipi_table array is composed. That's why
swap APU and PMC to follow the same order as is described by macros.
Change-Id: Ieaa3a967650e298e7cff45fafde0df96294c09fe Signed-off-by: Michal Simek <michal.simek@amd.com>
Akshay Belsare [Wed, 15 Feb 2023 05:19:52 +0000 (10:49 +0530)]
fix(zynqmp): with DEBUG=1 move bl31 to DDR range
Due to size constraints in OCM memory range keeping the bl31 with
DEBUG=1 overlaps with the memory range from other Firmware thus
affecting the bootflow on target.
bl31 binary can not be placed in OCM memory range when built with
DEBUG=1.
With DEBUG=1, by default bl31 is moved to DDR memory range
0x1000-0x7FFFF.
The user can provide a custom DDR memory range during build time using
the build parameters ZYNQMP_ATF_MEM_BASE and ZYNQMP_ATF_MEM_SIZE.
Akshay Belsare [Wed, 15 Feb 2023 11:59:42 +0000 (17:29 +0530)]
fix(zynqmp): update MAX_XLAT_TABLES for DDR memory range
An assert is observed when the bl31 is placed in DDR memory range and
DEBUG is also enabled. To resolve this, increase the size of
MAX_XLAT_TABLES to 8 when bl31 is placed in DDR memory range.
Ronak Jain [Mon, 13 Feb 2023 12:48:06 +0000 (04:48 -0800)]
feat(zynqmp): bump up version of query_data API
As per the current code base, the version of the PM_QUERY_DATA EEMI
API is 2 in the Versal but in ZynqMP it returns the base version.
Since this EEMI API ID support similar functionality for Versal and
ZynqMP, hence there should not be any difference in the versioning
as well.
In version 2, the feature check API supports the bitmask functionality
of the QUERY_DATA API, so the user can query the supported QUERY_DATA
ID first and if the ID is supported then the user can perform the
actual functionality of the same.
Hence, bump up the version of PM_QUERY_DATA API Id to 2.
Harrison Mutai [Mon, 13 Feb 2023 18:30:04 +0000 (18:30 +0000)]
docs: fix broken Juno links
Certain links to Juno documentation point to a location that were
removed at some point, or are unused. Fix links to point to the latest
available version on Arm's public documentation site, and remove those
that are no longer being used.
Signed-off-by: Harrison Mutai <harrison.mutai@arm.com>
Change-Id: I59202767db8834e9c302b2826f3faee47d3a5edd
Add code in SPMD to forward calls to FFA_PARTITION_INFO_GET_REGS. This
is a new ABI that allows getting partition information without the need
for rx/tx buffer, that helps in situations where having an rx/tx buffer
mapped and available is difficult (ex. uefi runtime services).
Currently, the spmc at el3 does not support this new ABI.
The new ABI uses registers x8-x17 to return partition information so
changes are made to ensure those registers are passed through to the
SPMC and restored on the return path.
Boyan Karatotev [Mon, 9 Jan 2023 11:50:24 +0000 (11:50 +0000)]
build(docs): update Python dependencies
Update the python dependencies for building the project's Sphinx
documentation. Sphinx plugins are updated to the latest version, while
Sphinx itself is only updated to 5.3.0 (latest 5.x.x revision) due to
sphinx-rtd-theme not supporting any higher (both require incompatible
versions of docutils). Myst-parser is also updated to the latest version
to prevent a docutils clash as well.
The effect of this is to bump certifi to version 2022.12.7 and wheel to
0.38.4 as suggested by dependabot.
Boyan Karatotev [Thu, 9 Feb 2023 15:59:39 +0000 (15:59 +0000)]
fix(docs): make required compiler version == rather than >=
TF-A carries its own compiler-rt so higher versions of the compilers may
not necessarily work. Because TF-A is only tested on the specified
versions in the CI, any breakage remains unknown. Update the
prerequisites guide to make it more apparent that this is the case.
Boyan Karatotev [Thu, 9 Feb 2023 15:55:34 +0000 (15:55 +0000)]
fix(deps): add missing aeabi_memset.S
This file provides __aeabi_memclr8 builtin which the Ubuntu 22.04
version of clang 14 needs to compile. Add it to prevent this oddity from
failing the build.
Michal Simek [Mon, 13 Feb 2023 12:11:28 +0000 (13:11 +0100)]
fix(zynqmp): fix DT reserved allocated size
BL31_LIMIT is not size but reserved node reg property contains base
address and size that's why BL31_LIMIT - BL31_BASE + 1 is correct size
of reseved space for BL31.
Also update warning message to cover that it is for BL31.
Change-Id: I53f53d2170eb873f758f9ba250d54f57f0b562b4 Signed-off-by: Michal Simek <michal.simek@amd.com>
Chris Kay [Mon, 16 Jan 2023 18:57:26 +0000 (18:57 +0000)]
build: permit multiple linker scripts
This change allows platforms to provide more than one linker script to
any image utilizing the `MAKE_BL` build system macro.
This is already done by some MediaTek platforms via the
`EXTRA_LINKERFILE` build system variable, which has now been removed.
In its place, additional linker scripts may be added to the
`<IMAGE>_LINKER_SCRIPT_SOURCES` variable.
BREAKING-CHANGE: The `EXTRA_LINKERFILE` build system variable has been
replaced with the `<IMAGE>_LINKER_SCRIPT_SOURCES` variable. See the
commit message for more information.
Change-Id: I3f0b69200d6a4841fd158cd09344ce9e67047271 Signed-off-by: Chris Kay <chris.kay@arm.com>
These new names better reflect how each variable is used:
1. the default linker script is passed via `-dT` instead of `-T`
2. linker script source files are first preprocessed
Additionally, linker scripts are now placed in the build directory
relative to where they exist in the source directory. For example,
the `bl32/sp_min/sp_min.ld.S` would now preprocess to
`sp_min/sp_min.ld` instead of just `bl32.ld`
BREAKING-CHANGE: The `LINKERFILE`, `BL_LINKERFILE` and
`<IMAGE_LINKERFILE>` build system variables have been renamed. See the
commit message for more information.
Change-Id: If8cef65dcb8820e8993736702c8741e97a66e6cc Signed-off-by: Chris Kay <chris.kay@arm.com>
Chris Kay [Thu, 29 Sep 2022 13:36:53 +0000 (14:36 +0100)]
style: normalize linker script code style
There are a variety of code styles used by the various linker scripts
around the code-base. This change brings them in line with one another
and attempts to make the scripts more friendly for skim-readers.
Change-Id: Ibee2afad0d543129c9ba5a8a22e3ec17d77e36ea Signed-off-by: Chris Kay <chris.kay@arm.com>
Chris Kay [Thu, 2 Feb 2023 14:39:03 +0000 (14:39 +0000)]
fix(pie): pass `-fpie` to the preprocessor as well
When PIE is enabled, the `-fpie` flag is passed to the compiler but not
to the preprocessor. This change ensures that both tools are aware of
when the image is position-independent when preprocessing, which impacts
some pre-defined preprocessor definitions.
Change-Id: I5208a591d60ee01312f6bf3dd7343abe6535ee61 Signed-off-by: Chris Kay <chris.kay@arm.com>
Thomas Viehweger [Mon, 23 Jan 2023 10:26:37 +0000 (11:26 +0100)]
fix(tsp): loop / crash if mmap of region fails
In test_memory_send the variable i is of unsigned type, so
it is never negative. If i is 0, the result of i-- is 4294967295. Don't know what happens if trying to
access composite->address_range_array[4294967295].
Made i a signed integer.
Signed-off-by: Thomas Viehweger <Thomas.Viehweger@rohde-schwarz.com>
Change-Id: I8b4e532749b5e86e4b5acd238e72c3f88e309ff2
Merge changes from topic "mb/tos-fw-config-load-refactor" into integration
* changes:
feat(spmd): map SPMC manifest region as EL3_PAS
feat(fvp): update device tree with load addresses of TOS_FW config
refactor(fvp): rename the DTB info structure member
feat(fconf): rename 'ns-load-address' to 'secondary-load-address'
There were some late comments to the prior change (18635) which are
address in this commit. There was also an invalid return value check
which was changed and the wrong result was being returned via the SMC
call for loading OP-TEE which is now fixed.
laurenw-arm [Wed, 8 Feb 2023 19:14:54 +0000 (13:14 -0600)]
fix(cert-create): change WARN to VERBOSE
SAVE_KEYS is set to '0' by default, causing cert_create to
show the 'Key filename not specified' message on each run
even though this is perfectly normal. Show the message only
in the VERBOSE log level.
Maksims Svecovs [Thu, 9 Feb 2023 16:48:34 +0000 (16:48 +0000)]
feat(git-hooks): add pre-commit hook
Adds a pre-commit git hook to keep track of copyright year.
Checks staged files for Arm copyright header and suggests a change if
the year is outdated. Works with both single-year format and
from_year-to_year format.
The tos_fw_config is currently loaded into memory by BL2 and
consumed by SPMD (part of BL31) and BL32 firmwares. This does
not work in RME-enabled systems as BL31 uses the root PAS memory
and does not trust secure PAS memory.
A first attempt was made to map the TOS_FW_CONFIG region as root PAS,
and then to remap to secure PAS after SPMD consumption, but this was
not suitable for RME systems where memory encryption is enabled.
This can be solved by copying the TOS FW config (SPMC manifest) from
the Root PAS region to the Secure PAS region so that BL32 can consume
it.
Change-Id: I8eef8345366199cb0e367db883c34a5b5136465d Signed-off-by: Manish V Badarkhe <Manish.Badarkhe@arm.com>
BL31 already reports that there is missing CPU workaround for this
erratum as
"WARNING: BL31: cortex_a53: CPU workaround for 1530924 was missing!"
That's why enable it by default as was done by other platforms for
example by commit 74665119f04d ("allwinner: Enable workaround for
Cortex-A53 erratum 1530924").
Change-Id: I251ffe3c307781b07477afb64f4e7af5dd9af9fe Signed-off-by: Michal Simek <michal.simek@amd.com>
Michal Simek [Thu, 9 Feb 2023 09:28:58 +0000 (10:28 +0100)]
feat(zynqmp): add SMCCC_ARCH_SOC_ID support
Add support for calling SMCCC_ARCH_SOC_ID which is used by Linux soc_id
driver for printing information about manufacturer and also chip version
and silicon ID code. SOC revision is directly mapped to chip ID code.
And SOC version is composed from manufacturer ID based on JEP-106 with
chip_id which contains bits mapped to CPU register 0xffca0044 platform
bits which differentiate between silicon, qemu and other emulated
platforms.
Function description is available at
docs/getting_started/porting-guide.rst.
Change-Id: I1f19e1973593897e71b39244dbdbceb6bd0e8a07 Signed-off-by: Michal Simek <michal.simek@amd.com>
Maksims Svecovs [Thu, 2 Feb 2023 16:10:22 +0000 (16:10 +0000)]
fix(context-mgmt): enable SCXTNUM access
Enable SCXTNUM_ELx access for lower ELs in non-secure state.
Make realm context setup take this build flag into account but enable it
by default when RME is used.
Merge changes I7bd311d7,Iea7dcfe3,I9d890934 into integration
* changes:
refactor(allwinner): use fdt_node_is_enabled() in AXP driver
fix(allwinner): check RSB availability in DT on H6
refactor(fdt): introduce common fdt_node_is_enabled()
laurenw-arm [Thu, 11 Aug 2022 20:29:56 +0000 (15:29 -0500)]
feat(psa): interface with RSS for NV counters
Adding AP/RSS interface for retrieving and incrementing non-volatile
counters.
The read interface implements the psa_call:
psa_call(RSS_PLATFORM_SERVICE_HANDLE,
RSS_PLATFORM_API_ID_NV_READ,
in_vec, 1, out_vec, 1);
where the in_vec indicates which of the 3 counters we want, and the
out_vec stores the counter value we get back from RSS.
The increment interface implements the psa_call:
psa_call(RSS_PLATFORM_SERVICE_HANDLE,
RSS_PLATFORM_API_ID_NV_INCREMENT,
in_vec, 1, (psa_outvec *)NULL, 0);
where, again, in_vec indicates the counter to increment, and we don't
get any output parameter from RSS.
Through this service, we will be able to get/increment any of the 3 NV
counters used on a CCA platform:
- NV counter for CCA firmware (BL2, BL31, RMM).
- NV counter for secure firmware.
- NV counter for non-secure firmware.
Signed-off-by: Lauren Wehrmeister <lauren.wehrmeister@arm.com> Signed-off-by: Raef Coles <raef.coles@arm.com>
Change-Id: I4c1c7f4837ebff30de16bb0ce7ecd416b70b1f62
feat(fvp): update device tree with load addresses of TOS_FW config
Provided both the root and secure addresses for TOS_FW config
in case of RME enabled systems where root address is in Root
SRAM and secure address is in Trusted DRAM.
Non-RME systems are unaffected by this change.
Change-Id: Ifb927c90fa5a68fe5362980858b4ddc5403ac95b Signed-off-by: Manish V Badarkhe <Manish.Badarkhe@arm.com>
feat(fconf): rename 'ns-load-address' to 'secondary-load-address'
The 'ns-load-address' property has been renamed to 'secondary-load-
address' in order to make it more generic. It can be used to copy
the configuration to any location, be it root, secure, or non-secure.
Change-Id: I122508e155ccd99082296be3f6b8db2f908be221 Signed-off-by: Manish V Badarkhe <Manish.Badarkhe@arm.com>
Andre Przywara [Wed, 1 Feb 2023 22:28:37 +0000 (22:28 +0000)]
fix(allwinner): check RSB availability in DT on H6
At the moment we access the RSB bus on all Allwinner H6 boards
unconditionally, even though some boards do not have any PMIC at all,
while others have some I2C devices connected to the same pins.
The latter case is just fragile, but the first case leads to a hang on
at least one board, as reported by Jernej.
Scan the devicetree, to check for the availability of the RSB bus node.
Proceed only if the RSB DT node is actually enabled.
Change-Id: Iea7dcfe3e085e173334d098ec4ddcb6c4b085771 Signed-off-by: Andre Przywara <andre.przywara@arm.com> Reported-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Andre Przywara [Fri, 3 Feb 2023 11:11:18 +0000 (11:11 +0000)]
refactor(fdt): introduce common fdt_node_is_enabled()
There are several users in the tree which want to check whether a given
FDT node is enabled or not: the "status" property holds that
information. So far all those users provide private implementations,
some of them having issues.
Export a generic implementation of that function in fdt_wrappers.h, as
a "static inline" function to not increase code size.
Also replace the existing implementation in Arm's fconf code, which had
a tiny bug in needlessly using the property length:
"status = [6f 6b 61 79 20];" would pass the check, where it should not.
The proper solution is also simpler: status must be a string, and
strings must be NUL-terminated in a DT. strcmp() would terminate on the
first NUL in *either* of the two strings it compares, so it would never
walk beyond the property boundary in the DTB.
Change-Id: I9d89093432f127c09add6cf5c93a725bc534e5de Signed-off-by: Andre Przywara <andre.przywara@arm.com>
Soby Mathew [Wed, 1 Feb 2023 16:03:22 +0000 (17:03 +0100)]
Merge changes I1b092bc1,Ifc2461b4,I5176caa5 into integration
* changes:
docs(rme): update RMM-EL3 Boot Manifest structure description
feat(rme): read DRAM information from FVP DTB
feat(rme): set DRAM information in Boot Manifest platform data
Jay Buddhabhatti [Tue, 24 Jan 2023 07:32:35 +0000 (23:32 -0800)]
fix(versal-net): populate gic v3 rdist data statically
Currently gicv3_rdistif_probe() is called per CPU. In case of maxcpus=1,
only 1 core is initialized and gicr_base_addrs initialized for CPU 0
only. Because of this assertion is raised during Linux system suspend.
During Linux suspend, platform callback saves GIC v3 state which
internally invokes arm_gicv3_distif_pre_save(). This function tries to use
gicr_base for all CPUs. Since GICR base address for secondary CPUs are not
initialized, it raises assertion.
To fix the issue, populate GIC v3 rdist data statically (similar to Versal)
instead of dynamically initializing GIC v3 rdist per CPU.
Signed-off-by: Jay Buddhabhatti <jay.buddhabhatti@amd.com>
Change-Id: I98c97c03e451d05f4ebac358e197617ab9d9b71f
Akshay Belsare [Wed, 18 Jan 2023 10:24:12 +0000 (15:54 +0530)]
feat(versal-net): add jtag dcc support
Add support for JTAG Debug Communication Channel(DCC), using the dcc
console driver, for Versal NET platform.
UART0/UART1 is not configured when the JTAG DCC is used as console for
the platform.
Though DCC is not using any UART, VERSAL_NET_UART_BASE needs
to be defined in the platform code. If its not defined, build errors
are observed.
Now VERSAL_NET_UART_BASE by default points to UART0 base.
Check for valid console(pl011, pl011_0, pl011_1, dcc) is
being done in the platform makefile, the error condition in
setting the value of VERSAL_NET_UART_BASE is redundant, thus the error
message is removed from the code.
This adds the ability to load the OP-TEE image via an SMC called from
non-secure userspace rather than loading it during boot. This should
only be utilized on platforms that can ensure security is maintained up
until the point the SMC is invoked as it breaks the normal barrier
between the secure and non-secure world.
AlexeiFedorov [Thu, 29 Dec 2022 15:57:40 +0000 (15:57 +0000)]
feat(rme): read DRAM information from FVP DTB
This patch builds on the previous patch by implementing
support for reading NS DRAM layout of FVP model from
HW_CONFIG Device tree.
Macro _RMMD_MANIFEST_VERSION is renamed to
SET_RMMD_MANIFEST_VERSION to suppress MISRA-C
"rule MC3R1.D4.5: (advisory) Identifiers in
the same name space with overlapping visibility
should be typographically unambiguous" warning
Loic Poulain [Wed, 11 Jan 2023 15:08:48 +0000 (16:08 +0100)]
perf(imx): speed-up console/uart TX using FIFO
The current putc version test for TXEMPTY bit set (#6) instead
of waiting for TXFULL bit clear (#4), that slows the global
boot time as we are not taking benefit of the 32-byte FIFO.
We then need to implement the flush function to be sure the
transmit is complete (FIFO and shift register empty).
projects / arm-tf.git / log