This patch introduces a build flag which allows the xlat tables
to be mapped in a read-only region within BL31 memory. It makes it
much harder for someone who has acquired the ability to write to
arbitrary secure memory addresses to gain control of the
translation tables.
The memory attributes of the descriptors describing the tables
themselves are changed to read-only secure data. This change
happens at the end of BL31 runtime setup. Until this point, the
tables have read-write permissions. This gives a window of
opportunity for changes to be made to the tables with the MMU on
(e.g. reclaiming init code). No changes can be made to the tables
with the MMU turned on from this point onwards. This change is also
enabled for sp_min and tspd.
To make all this possible, the base table was moved to .rodata. The
penalty we pay is that now .rodata must be aligned to the size of
the base table (512B alignment). Still, this is better than putting
the base table with the higher level tables in the xlat_table
section, as that would cost us a full 4KB page.
Changing the tables from read-write to read-only cannot be done with
the MMU on, as the break-before-make sequence would invalidate the
descriptor which resolves the level 3 page table where that very
descriptor is located. This would make the translation required for
writing the changes impossible, generating an MMU fault.
The caches are also flushed.
Signed-off-by: Petre-Ionut Tudor <petre-ionut.tudor@arm.com>
Change-Id: Ibe5de307e6dc94c67d6186139ac3973516430466
Till now TF-A allows limited number of external images to be made part
of FIP. With SPM coming along, there may exist multiple SP packages
which need to be inserted into FIP. To achieve this we need a more
scalable approach to feed SP packages to FIP.
This patch introduces changes in build system to generate and add SP
packages into FIP based on information provided by platform.
Platform provides information in form of JSON which contains layout
description of available Secure Partitions.
JSON parser script is invoked by build system early on and generates
a makefile which updates FIP, SPTOOL and FDT arguments which will be
used by build system later on for final packaging.
"SP_LAYOUT_FILE" passed as a build argument and can be outside of TF-A
tree. This option will be used only when SPD=spmd.
For each SP, generated makefile will have following entries
- FDT_SOURCES += sp1.dts
- SPTOOL_ARGS += -i sp1.img:sp1.dtb -o sp1.pkg
- FIP_ARGS += --blob uuid=XXXX-XXX...,file=SP1.pkg
Signed-off-by: Manish Pandey <manish.pandey2@arm.com>
Change-Id: Ib6a9c064400caa3cd825d9886008a3af67741af7
There are chances a denial-of-service attack, if an attacker
removes the SPE firmware from the system. The console driver
would end up waiting for the firmware to respond indefinitely.
The console driver must detect such scenarios and uninit the
interface as a result.
This patch adds a timeout to the interaction with the SPE
firmware and uninits the interface if it times out.
Change-Id: I06f27a858baed25711d41105b4110865f1a01727
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
Tegra210 SoCs need the sc7entry-fw to enter System Suspend mode,
but there might be certain boards that do not have this firmware
blob. To stop the NS world from issuing System suspend entry
commands on such devices, we ned to disable System Suspend from
the PSCI "features".
This patch removes the System suspend handler from the Tegra PSCI
ops, so that the framework will disable support for "System Suspend"
from the PSCI "features".
Original change by: kalyani chidambaram <kalyanic@nvidia.com>
Change-Id: Ie029f82f55990a8b3a6debb73e95e0e218bfd1f5
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
Code complexity is a good indication of maintainability versus
testability of a piece of software.
ISO26262 introduces the following thresholds:
complexity < 10 is accepted
10 <= complexity < 20 has to be justified
complexity >= 20 cannot be accepted
Rationale is that number of test cases to fully test a piece of
software can (depending on the coverage metrics) grow exponentially
with the number of branches in the software.
This patch removes redundant conditionals from 'ipc_send_req_atomic'
handler to reduce the McCabe Cyclomatic Complexity for this function
Change-Id: I20fef79a771301e1c824aea72a45ff83f97591d5
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
This patch provides platforms an opportunity to relocate the
BL32 image, during cold boot. Tegra186 platforms, for example,
relocate BL32 images to TZDRAM memory as the previous bootloader
relies on BL31 to do so.
Change-Id: Ibb864901e43aca5bf55d8c79e918b598c12e8a28
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
Code complexity is a good indication of maintainability versus
testability of a piece of software.
ISO26262 introduces the following thresholds:
complexity < 10 is accepted
10 <= complexity < 20 has to be justified
complexity >= 20 cannot be accepted
Rationale is that number of test cases to fully test a piece of
software can (depending on the coverage metrics) grow exponentially
with the number of branches in the software.
This patch removes redundant conditionals from 'bl31_early_platform_setup'
handler to reduce the McCabe Cyclomatic Complexity for this function.
Change-Id: Ifb628e33269b388f9323639cd97db761a7e049c4
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
This patch sets the "secure" bit to mark the PMC hardware block
as accessible only from the secure world. This setting must be
programmed during cold boot and System Resume.
The sc7entry-fw, running on the COP, needs access to the PMC block
to enter System Suspend state, so "unlock" the PMC block before
passing control to the COP.
Change-Id: I00e39a49ae6b9f8c8eafe0cf7ff63fe6a67fdccf
Signed-off-by: kalyani chidambaram <kalyanic@nvidia.com>
This patch modifies the delay timer driver to switch to the ARM
secure physical timer instead of using Tegra's on-chip uS timer.
The secure timer is not accessible to the NS world and so eliminates
an important attack vector, where the Tegra timer source gets switched
off from the NS world leading to a DoS attack for the trusted world.
This timer is shared with the S-EL1 layer for now, but later patches
will mark it as exclusive to the EL3 exception mode.
Change-Id: I2c00f8cb4c48b25578971c626c314603906ad7cc
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
This patch moves the MHZ_TICKS_PER_SEC macro to utils_def.h
for other platforms to use.
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
Change-Id: I6c4dc733f548d73cfdb3515ec9ad89a9efaf4407
This patch locks most of the stream id security config registers as
per HW guidance.
This patch keeps the stream id configs unlocked for the following
clients, to allow some platforms to still function, until they make
the transition to the latest guidance.
- ISPRA
- ISPFALR
- ISPFALW
- ISPWA
- ISPWA1
- ISPWB
- XUSB_DEVR
- XUSB_DEVW
- XUSB_HOSTR
- XUSB_HOSTW
- VIW
- VIFALR
- VIFALW
Change-Id: I66192b228a0a237035938f498babc0325764d5df
Signed-off-by: Pritesh Raithatha <praithatha@nvidia.com>
The PMC hardware block resume handler was called for Tegra210
platforms, only if the sc7entry-fw was present on the device.
This would cause problems for devices that do not support this
firmware.
This patch fixes this logic and resumes the PMC block even if
the sc7entry-fw is not present on the device.
Change-Id: I6f0eb7878126f624ea98392f583ed45a231d27db
Signed-off-by: Kalyani Chidambaram <kalyanic@nvidia.com>
This patch adds the macro to enable legacy FIQ handling to the common
Tegra makefile. The default value of this macro is '0'. Platforms that
need this support should enable it from their makefiles.
This patch also helps fix violation of Rule 20.9.
Rule 20.9 "All identifiers used in the controlling expression of #if
of #elif preprocessing directives shall be #define'd before
evaluation"
Change-Id: I4f0c9917c044b5b1967fb5e79542cd3bf6e91f18
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
The CPUACTLR_EL1 register on Cortex-A57 CPUs supports a bit to enable
non-cacheable streaming enhancement. Platforms can set this bit only
if their memory system meets the requirement that cache line fill
requests from the Cortex-A57 processor are atomic.
This patch adds support to enable higher performance non-cacheable load
forwarding for such platforms. Platforms must enable this support by
setting the 'A57_ENABLE_NONCACHEABLE_LOAD_FWD' flag from their
makefiles. This flag is disabled by default.
Change-Id: Ib27e55dd68d11a50962c0bbc5b89072208b4bac5
Signed-off-by: Varun Wadekar <vwadekar@nvidia.com>
Removed duplicate error code present for SMCCC and used
proper error code for "SMCCC_ARCH_WORKAROUND_2" call.
Signed-off-by: Manish V Badarkhe <Manish.Badarkhe@arm.com>
Change-Id: I76fc7c88095f78a7e2c3d205838f8eaf3132ed5c
Fix mailbox driver incompatible cast bug and control flow issue that
was flagged by Coverity Scan.
Signed-off-by: Abdul Halim, Muhammad Hadi Asyrafi <muhammad.hadi.asyrafi.abdul.halim@intel.com>
Change-Id: I3f34e98d24e40139d31cf7d5b9b973cd2d981065
Use CREATE_SEQ helper macro to create sequence of valid chip counts
instead of manually creating the sequence. This allows a scalable
approach to increase the valid chip count sequence in the future.
Change-Id: I5ca7a00460325c156b9e9e52b2bf656a2e43f82d
Signed-off-by: Vijayenthiran Subramaniam <vijayenthiran.subramaniam@arm.com>
Add `CREATE_SEQ` function to generate sequence of numbers starting from
1 to allow easy comparison of a user defined macro with non-zero
positive numbers.
Change-Id: Ibcb336a223d958154b1007d08c428fbaf1e48664
Signed-off-by: Vijayenthiran Subramaniam <vijayenthiran.subramaniam@arm.com>
This patch allows to use DDR address in memory node because on FPGA we
typically use DDR instead of shared RAM.
This patch also modifies the kernel arguments to allow the rootfs to be
mounted from a direct mapping of the QSPI NOR flash using the physmap
driver in the kernel. This allows to support CRAMFS XIP.
Change-Id: I4e2bc6a1f48449c7f60e00f5f1a698df8cb2ba89
Signed-off-by: Vishnu Banavath <vishnu.banavath@arm.com>
Signed-off-by: Abdellatif El Khlifi <abdellatif.elkhlifi@arm.com>
Fixes for the following MISRA violations:
- Missing explicit parentheses on sub-expression
- An identifier or macro name beginning with an
underscore, shall not be declared
- Type mismatch in BL1 SMC handlers and tspd_main.c
Change-Id: I7a92abf260da95acb0846b27c2997b59b059efc4
Signed-off-by: Zelalem <zelalem.aweke@arm.com>
* changes:
Fix boot failures on some builds linked with ld.lld.
trusty: generic-arm64-smcall: Support gicr address
trusty: Allow gic base to be specified with GICD_BASE
trusty: Allow getting trusty memsize from BL32_MEM_SIZE instead of TSP_SEC_MEM_SIZE
Fix clang build if CC is not in the path.
When TF-A is built with RESET_TO_BL31=1 option, BL31 is the
first image to be run and should have all the memory allocated
to it except for the memory reserved for Shared RAM at the start
of Trusted SRAM.
This patch fixes FVP BL31 load address and its image size for
RESET_TO_BL31=1 option. BL31 startup address should be set to
0x400_1000 and its maximum image size to the size of Trusted SRAM
minus the first 4KB of shared memory.
Loading BL31 at 0x0402_0000 as it is currently stated in
'\docs\plat\arm\fvp\index.rst' causes EL3 exception when the
image size gets increased (i.e. building with LOG_LEVEL=50)
but doesn't exceed 0x3B000 not causing build error.
Change-Id: Ie450baaf247f1577112f8d143b24e76c39d33e91
Signed-off-by: Alexei Fedorov <Alexei.Fedorov@arm.com>
The TBBR implementation extracts hashes from certificates and stores
them in static buffers. TF-A supports 3 variants of SHA right now:
SHA-256, SHA-384 and SHA-512. When support for SHA-512 was added in
commit 9a3088a5f5 ("tbbr: Add build flag
HASH_ALG to let the user to select the SHA"), the hash buffers got
unconditionally increased from 51 to 83 bytes each. We can reduce that
space if we're using SHA-256 or SHA-384.
This saves some BSS space in both BL1 and BL2:
- BL1 with SHA-256: saving 168 bytes.
- BL1 with SHA-384: saving 80 bytes.
- BL2 with SHA-256: saving 384 bytes.
- BL2 with SHA-384: saving 192 bytes.
Change-Id: I0d02e5dc5f0162e82339c768609c9766cfe7e2bd
Signed-off-by: Sandrine Bailleux <sandrine.bailleux@arm.com>
The TBBR implementation extracts public keys from certificates and
stores them in static buffers. DER-encoded ECDSA keys are only 91 bytes
each but were each allocated 294 bytes instead. Reducing the size of
these buffers saves 609 bytes of BSS in BL2 (294 - 91 = 203 bytes for
each of the 3 key buffers in use).
Also add a comment claryfing that key buffers are tailored on RSA key
sizes when both ECDSA and RSA keys are used.
Change-Id: Iad332856e7af1f9814418d012fba3e1e9399f72a
Signed-off-by: Sandrine Bailleux <sandrine.bailleux@arm.com>
Adding support for 32MHz UART clock and selecting it as the
default UART clock
Change-Id: I9541eaff70424e85a3b5ee4820ca0e7efb040d2c
Signed-off-by: Vishnu Banavath <vishnu.banavath@arm.com>
Signed-off-by: Abdellatif El Khlifi <abdellatif.elkhlifi@arm.com>
Running checkpatch.pl on the codebase and making required changes
Change-Id: I7d3f8764cef632ab2a6d3c355c68f590440b85b8
Signed-off-by: Avinash Mehta <avinash.mehta@arm.com>
Signed-off-by: Abdellatif El Khlifi <abdellatif.elkhlifi@arm.com>
Note: This patch implements in-band messaging protocol only.
ARM has launched a next version of MHU i.e. MHUv2 with its latest
subsystems. The main change is that the MHUv2 is now a distributed IP
with different peripheral views (registers) for the sender and receiver.
Another main difference is that MHUv1 duplex channels are now split into
simplex/half duplex in MHUv2. MHUv2 has a configurable number of
communication channels. There is a capability register (MSG_NO_CAP) to
find out how many channels are available in a system.
The register offsets have also changed for STAT, SET & CLEAR registers
from 0x0, 0x8 & 0x10 in MHUv1 to 0x0, 0xC & 0x8 in MHUv2 respectively.
0x0 0x4 0x8 0xC 0x1F
------------------------....-----
| STAT | | | SET | | |
------------------------....-----
Transmit Channel
0x0 0x4 0x8 0xC 0x1F
------------------------....-----
| STAT | | CLR | | | |
------------------------....-----
Receive Channel
The MHU controller can request the receiver to wake-up and once the
request is removed, the receiver may go back to sleep, but the MHU
itself does not actively put a receiver to sleep.
So, in order to wake-up the receiver when the sender wants to send data,
the sender has to set ACCESS_REQUEST register first in order to wake-up
receiver, state of which can be detected using ACCESS_READY register.
ACCESS_REQUEST has an offset of 0xF88 & ACCESS_READY has an offset
of 0xF8C and are accessible only on any sender channel.
This patch adds necessary changes in a new file required to support the
latest MHUv2 controller. This patch also needs an update in DT binding
for ARM MHUv2 as we need a second register base (tx base) which would
be used as the send channel base.
Change-Id: I1455e08b3d88671a191c558790c503eabe07a8e6
Signed-off-by: Tushar Khandelwal <tushar.khandelwal@arm.com>
Mark Dykes
Soby Mathew
Sandrine Bailleux
Petre-Ionut Tudor
joanna.farley
Manish Pandey
Varun Wadekar
kalyani chidambaram
Pritesh Raithatha
Manish V Badarkhe
Abdul Halim, Muhammad Hadi Asyrafi
Vijayenthiran Subramaniam
Rui Silva
Zelalem
Jimmy Brisson
Alexei Fedorov
Vishnu Banavath
Avinash Mehta
Khandelwal