Newer cores in upcoming platforms may refuse to power down. The PSCI
library is already prepared for this so convert platform code to also
allow this. This is simple - drop the `wfi` + panic and let common code
deal with the fallout. The end result will be the same (sans the
message) except the platform will have fewer responsibilities. The only
exception is for cores being signalled to power off gracefully ahead of
system reset. That path must also be terminal so replace the end with
the same psci_pwrdown_cpu_end() to behave the same as the generic
implementation. It will handle wakeups and panic, hoping that the system
gets reset from under it. The dmb is upgraded to a dsb so no functional
change.
Change-Id: I381f96bec8532bda6ccdac65de57971aac42e7e8
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
The simplistic view of a core's powerdown sequence is that power is
atomically cut upon calling `wfi`. However, it turns out that it has
lots to do - it has to talk to the interconnect to exit coherency, clean
caches, check for RAS errors, etc. These take significant amounts of
time and are certainly not atomic. As such there is a significant window
of opportunity for external events to happen. Many of these steps are
not destructive to context, so theoretically, the core can just "give
up" half way (or roll certain actions back) and carry on running. The
point in this sequence after which roll back is not possible is called
the point of no return.
One of these actions is the checking for RAS errors. It is possible for
one to happen during this lengthy sequence, or at least remain
undiscovered until that point. If the core were to continue powerdown
when that happens, there would be no (easy) way to inform anyone about
it. Rejecting the powerdown and letting software handle the error is the
best way to implement this.
Arm cores since at least the a510 have included this exact feature. So
far it hasn't been deemed necessary to account for it in firmware due to
the low likelihood of this happening. However, events like GIC wakeup
requests are much more probable. Older cores will powerdown and
immediately power back up when this happens. Travis and Gelas include a
feature similar to the RAS case above, called powerdown abandon. The
idea is that this will improve the latency to service the interrupt by
saving on work which the core and software need to do.
So far firmware has relied on the `wfi` being the point of no return and
if it doesn't explicitly detect a pending interrupt quite early on, it
will embark onto a sequence that it expects to end with shutdown. To
accommodate for it not being a point of no return, we must undo all of
the system management we did, just like in the warm boot entrypoint.
To achieve that, the pwr_domain_pwr_down_wfi hook must not be terminal.
Most recent platforms do some platform management and finish on the
standard `wfi`, followed by a panic or an endless loop as this is
expected to not return. To make this generic, any platform that wishes
to support wakeups must instead let common code call
`psci_power_down_wfi()` right after. Besides wakeups, this lets common
code handle powerdown errata better as well.
Then, the CPU_OFF case is simple - PSCI does not allow it to return. So
the best that can be done is to attempt the `wfi` a few times (the
choice of 32 is arbitrary) in the hope that the wakeup is transient. If
it isn't, the only choice is to panic, as the system is likely to be in
a bad state, eg. interrupts weren't routed away. The same applies for
SYSTEM_OFF, SYSTEM_RESET, and SYSTEM_RESET2. There the panic won't
matter as the system is going offline one way or another. The RAS case
will be considered in a separate patch.
Now, the CPU_SUSPEND case is more involved. First, to powerdown it must
wipe its context as it is not written on warm boot. But it cannot be
overwritten in case of a wakeup. To avoid the catch 22, save a copy that
will only be used if powerdown fails. That is about 500 bytes on the
stack so it hopefully doesn't tip anyone over any limits. In future that
can be avoided by having a core manage its own context.
Second, when the core wakes up, it must undo anything it did to prepare
for poweroff, which for the cores we care about, is writing
CPUPWRCTLR_EL1.CORE_PWRDN_EN. The least intrusive for the cpu library
way of doing this is to simply call the power off hook again and have
the hook toggle the bit. If in the future there need to be more complex
sequences, their direction can be advised on the value of this bit.
Third, do the actual "resume". Most of the logic is already there for
the retention suspend, so that only needs a small touch up to apply to
the powerdown case as well. The missing bit is the powerdown specific
state management. Luckily, the warmboot entrypoint does exactly that
already too, so steal that and we're done.
All of this is hidden behind a FEAT_PABANDON flag since it has a large
memory and runtime cost that we don't want to burden non pabandon cores
with.
Finally, do some function renaming to better reflect their purpose and
make names a little bit more consistent.
Change-Id: I2405b59300c2e24ce02e266f91b7c51474c1145f
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
When GICR_WAKER.ProcessorSleep == 1 (i.e. after gicv3_cpuif_disable())
the GIC will assert the WakeRequest signal to try and wake the core up
instead of delivering an interrupt. This is useful when a core is in
some kind of suspend state.
However, when the core is properly off (CPU_OFF), it shouldn't get woken
up in any way other than a CPU_ON call. In the general case interrupts
would be routed away so this doesn't matter. But in case they aren't, we
want the core to stay off.
So turn the redistributor off on CPU_OFF calls. This will prevent the
WakeRequest from being sent.
Change-Id: I7f20591d1c83a4a9639281ef86caa79d6669b536
Signed-off-by: Boyan Karatotev <boyan.karatotev@arm.com>
This patch adds support for preserving DSU PMU registers
over a power cycle in TC platform.
These PMU registers need to be manually saved/restored
because they are part of cluster power domain and OS
doesn't know when DSU is powered OFF.
Change-Id: Ife9573f205d99d092039cb95674e7434bb5f9239
Signed-off-by: Arvind Ram Prakash <arvind.ramprakash@arm.com>
In certain instances a platform may need to make modifications to an
image after it has been loaded by BL2. The existing common
implementation is a thin wrapper for a more generic arm post image
handler. To enable platforms to make changes to images when
they're loaded, move this into platform code.
Change-Id: I44025391056adb2d8a8eb4ea5984257b02027181
Signed-off-by: Harrison Mutai <harrison.mutai@arm.com>
This reverts commit 9cf7f355ce.
Above mentioned commit was writing to cluster power required bit of
CLUSTERPWRDN register, which provides an advisory status to the power
controller.
Bit definition indication:
0 : Cluster power is not required when all cores are powered down
1 : Cluster power is required even when all cores are powered down
RESET value of this bit is 0
The current implementation in TF-A just programs this bit to 0 when
cluster power down is done but it never sets it to 1. Which actully
does not change any behaviour as the value of this bit always remains 0.
Ideally this bit has to be set to 1 when a core powers up (as RESET
value is 0) and set it to 0 for any core power down except if its last
man standing, in that case we need to ensure the target power level
from OS is cluster then we can do set it to 0.
There also are some investigation needs to be done to find that whether
we need a explicit message to power controller for turning cluster OFF
or it will happen automatically.
Considering this needs a bit of analysis as well as a platform to test
it on, revert the changes which impact the programming during cluster
power down and just keep register defnition.
Change-Id: I4c4ebedae7ca9cd081fb1e0605b9d906d77614d9
Signed-off-by: Manish Pandey <manish.pandey2@arm.com>
Commit 4d8c181963
introduced an invalid redistributor power off
where we turn off the redistributor without
checking if the system power domain level is
turning off, otherwise we can turn off a
redistributor when other cores or clusters are
sharing it, also if it does indeed needs
powering off during suspend we do it twice.
This change fixes this by checking on the
system power state first then turning off
the redistributor.
Signed-off-by: Waleed Elmelegy <waleed.elmelegy@arm.com>
Change-Id: Id202bc2316ab7c516298fa33ea089ae2e221a933
To initiate a reset or reboot, the nonsecure OS invokes the PSCI
SYSTEM_RESET function from any one core. As per the PSCI specification,
it is the responsibility of firmware to implement the system view of
the reset or reboot operation. For the platforms supported by CSS,
trigger the reset/reboot operation by sending an SGI to rest all CPUs
which are online. The CPUs respond to this interrupt by initiating its
powerdown sequence.
In addition to these changes, fix coding style issues that are not
directly related to the code being introduced in this patch.
Change-Id: I547253ee28ef7eefa78180d016893671a406bbfa
Signed-off-by: Pranav Madhu <pranav.madhu@arm.com>
Add platform specific interrupt handler for handling the reboot of
all CPU's. On shutdown/reboot, only one CPU invoke PSCI and enter into
trusted firmware. The CPU which entered trusted firmware signals the
rest of the cores which are online using SGI to initiate power down
sequence. On receiving the SGI, the handler will power down the
GIC redistributor interface of the respective core, configure the power
control register and power down the CPU by executing wfi.
In addition to these changes, fix coding style issues that are not
directly related to the code being introduced in this patch.
Change-Id: I4917dfdc47be5ce7367bee629486a6344cdd706f
Signed-off-by: Pranav Madhu <pranav.madhu@arm.com>
Add a new function to setup a SGI interrupt that will be used to trigger
a request for per-cpu power down when executing the PSCI SYSTEM_RESET
request. This will be used on CSS platform that require all the CPUs to
execute the CPU specific power down sequence to complete a warm reboot
sequence in which only the CPUs are power cycled.
Change-Id: I80da0f6c3cd0c5c442c82239ba1e1f773821a7f5
Signed-off-by: Pranav Madhu <pranav.madhu@arm.com>
Turn ON/OFF GIC redistributor in sync with GIC CPU interface ON/OFF.
Issue :
The Linux prompt hangs when all the cores in a cluster are turned OFF
and we try to turn ON a core in that cluster. Previously when TF-A turns
ON a core, TF-A first turns ON the redistributor followed by the core.
This did not match the flow when turning OFF a core, as TF-A did not
turn OFF redistributor when the corresponding core[s] are disabled.
This hang is resolved by disabling redistributor as cores are disabled,
keeping them in sync.
Signed-off-by: Jagadeesh Ujja <jagadeesh.ujja@arm.com>
Change-Id: Ifd04fdcfd47b45e00f874f15b098471883d023f0
By writing 0 to CLUSTERPWRDN DSU register bit 0, we send an
advisory to the power controller that cluster power is not required
when all cores are powered down.
The AArch32 CLUSTERPWRDN register is architecturally mapped to the
AArch64 CLUSTERPWRDN_EL1 register
Change-Id: Ie6e67c1c7d811fa25c51e2e405ca7f59bd20c81b
Signed-off-by: Madhukar Pappireddy <madhukar.pappireddy@arm.com>
This patch invokes the new function gicv3_rdistif_probe() in the
ARM platform specific gicv3 driver. Since this API modifies the
shared GIC related data structure, it must be invoked coherently
by using the platform specific pwr_domain_on_finish_late hook.
Change-Id: I6efb17d5da61545a1c5a6641b8f58472b31e62a8
Signed-off-by: Madhukar Pappireddy <madhukar.pappireddy@arm.com>
Use full include paths like it is done for common includes.
This cleanup was started in commit d40e0e08283a ("Sanitise includes
across codebase"), but it only cleaned common files and drivers. This
patch does the same to Arm platforms.
Change-Id: If982e6450bbe84dceb56d464e282bcf5d6d9ab9b
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
From now on, platform_def.h must include any header with definitions that
are platform-specific (like arm_def.h) and the included headers mustn't
include back platform_def.h, and shouldn't be used by other files. Only
platform_def.h should be included in other files. This will ensure that all
needed definitions are present, rather than needing to include all the
headers in all the definitions' headers just in case.
This also prevents problems like cyclic dependencies.
Change-Id: I9d3cf4d1de4b956fa035c79545222697acdaf5ca
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
Enforce full include path for includes. Deprecate old paths.
The following folders inside include/lib have been left unchanged:
- include/lib/cpus/${ARCH}
- include/lib/el3_runtime/${ARCH}
The reason for this change is that having a global namespace for
includes isn't a good idea. It defeats one of the advantages of having
folders and it introduces problems that are sometimes subtle (because
you may not know the header you are actually including if there are two
of them).
For example, this patch had to be created because two headers were
called the same way: e0ea0928d5 ("Fix gpio includes of mt8173 platform
to avoid collision."). More recently, this patch has had similar
problems: 46f9b2c3a2 ("drivers: add tzc380 support").
This problem was introduced in commit 4ecca33988 ("Move include and
source files to logical locations"). At that time, there weren't too
many headers so it wasn't a real issue. However, time has shown that
this creates problems.
Platforms that want to preserve the way they include headers may add the
removed paths to PLAT_INCLUDES, but this is discouraged.
Change-Id: I39dc53ed98f9e297a5966e723d1936d6ccf2fc8f
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
This patch makes use of mov_imm macro where possible to load
immediate values within ARM platform layer.
Change-Id: I02bc7fbc1fa334c9fccf76fbddf515952f9a1298
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
This lets any future CSS platforms to use RESET_TO_BL31 flag.
Change-Id: I32a90fce43cb0c6f4d33589653a0fd6a7ecc9577
Signed-off-by: Deepak Pandey <Deepak.Pandey@arm.com>
- Remove references to removed build options.
- Remove support for legacy GIC driver.
- Remove support for LOAD_IMAGE_V2=0.
Change-Id: I72f8c05620bdf4a682765e6e53e2c04ca749a3d5
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
Adds an undocumented build option that enables non-secure access to
the PL011 UART1.
This allows a custom build where the UART can be used as a serial debug
port for WinDbg (or other debugger) connection.
This option is not documented in the user guide, as it is provided as a
convenience for Windows debugging, and not intended for general use.
In particular, enabling non-secure access to the UART might allow
a denial of service attack!
Change-Id: I4cd7d59c2cac897cc654ab5e1188ff031114ed3c
Signed-off-by: Alexei Fedorov <Alexei.Fedorov@arm.com>
Signed-off-by: Evan Lloyd <evan.lloyd@arm.com>
In the case of the platform max power level being less than the system
power level, make sure to not overrun the array of power states.
This fixes Coverity Scan OVERRUN defect CID 267021.
Change-Id: I52646ab9be2fceeb5c331b5dad7a6267991f4197
Signed-off-by: Nariman Poushin <nariman.poushin@linaro.org>
This is to fix a number of Coverity Scan DEADCODE defects, CID numbers
listed below, as reported from
https://scan.coverity.com/projects/arm-software-arm-trusted-firmware
CID 267023
CID 267022
CID 267020
Change-Id: I2963a799b210149e84ccab5c5b9082267ddfe337
Signed-off-by: Nariman Poushin <nariman.poushin@linaro.org>
The MT bit in MPIDR is always set for SGI platforms and so the
core position calculation code is updated to take into account
the thread affinity value as well.
Change-Id: I7b2a52707f607dc3859c6bbcd2b145b7987cb4ed
Signed-off-by: Chandni Cherukuri <chandni.cherukuri@arm.com>
Signed-off-by: Vishwanatha HG <vishwanatha.hg@arm.com>
Previously mem_protect used to be only supported from BL2. This is not
helpful in the case when ARM TF-A BL2 is not used. This patch demonstrates
mem_protect from el3_runtime firmware on ARM Platforms specifically
when RESET_TO_BL31 or RESET_TO_SP_MIN flag is set as BL2 may be absent
in these cases. The Non secure DRAM is dynamically mapped into EL3 mmap
tables temporarily and then the protected regions are then cleared. This
avoids the need to map the non secure DRAM permanently to BL31/sp_min.
The stack size is also increased, because DYNAMIC_XLAT_TABLES require
a bigger stack.
Change-Id: Ia44c594192ed5c5adc596c0cff2c7cc18c001fde
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
Currently the SCMI driver supports MHUv1, but Arm platforms may have
varied versions of MHU driver, with MHUv2 controllers being in the
latest Arm platforms.
This patch updates the SCMI driver to support MHUv2, specifically that
the sender must send the wake-up to the receiver before initiating any
data transfer.
Also, the existing mhu driver files, css_mhu.c and css_mhu.h, have been
moved from the scpi directory to a new directory, css/drivers/mhu.
Change-Id: I9b46b492a3e1d9e26db12d83a9773958a8c8402f
Signed-off-by: Samarth Parikh <samarth.parikh@arm.com>
Rule 8.4: A compatible declaration shall be visible when
an object or function with external linkage is defined
Fixed for:
make DEBUG=1 PLAT=juno LOG_LEVEL=50 all
Change-Id: Ic8f611da734f356566e8208053296e6c62b54709
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
This patch migrates the ARM Standard platforms to the new BL
handover interface. The arm_blx_early_platform_setup() functions
are also modified to take in 4 arguments. The `ARM_BL31_PLAT_PARAM_VAL`
value passed to BL31 from BL2 is now in arg3 in preparation of dynamic
configuration arguments.
Change-Id: I33e8e61325a19e7a7127b1ff203c3b86921bf153
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
In the next commit, I need the image name in lowercase because
output files are generally named in lowercase.
Unfortunately, TOOL_ADD_IMG takes the first argument in uppercase
since we generally use uppercase Make variables.
make_helpers/build_macros.mk provides 'uppercase' macro to convert
a string into uppercase, but 'lowercase' does not exist. We can
implement it if we like, but it would be more straightforward to
change the argument of TOOL_ADD_IMG.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Now FIP_ADD_IMG takes care of both fiptool and cert_create
symmetrically. Rename it so that it matches the behavior.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
The build system supports generating two FIP images, fip and fwu_fip.
Accordingly, we have similar build macros.
FIP_ADD_PAYLOAD <--> FWU_FIP_ADD_PAYLOAD
CERT_ADD_CMD_OPT <--> FWU_CERT_ADD_CMD_OPT
FIP_ADD_IMG <--> FWU_FIP_ADD_IMG
The duplicated code increases the maintenance burden. Also, the build
rule of BL2U looks clumsy - we want to call MAKE_BL to compile it from
source files, but we want to put it in fwu_fip. We can not do it in a
single macro call since the current MAKE_BL does not support fwu_fip.
To refactor those in a clean way is to support one more argument to
specify the FIP prefix. If it is empty, the images are targeted to
fip, whereas if the argument is "FWU_", targeted to fwu_fip.
The build macros prefixed with FWU_ go away.
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
The bl2_early_platform_setup() and bl2_platform_setup() were
redefined for Juno AArch32 eventhough CSS platform layer had
same definition for them. The CSS definitions definitions were
previously restricted to EL3_PAYLOAD_BASE builds and this is now
modified to include the Juno AArch32 builds as well thus
allowing us to remove the duplicate definitions in Juno platform
layer.
Change-Id: Ibd1d8c1428cc1d51ac0ba90f19f5208ff3278ab5
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
The function arm_validate_ns_entrypoint() validates a given non-secure
physical address. This function however specifically returns PSCI error
codes.
Non-secure physical address validation is potentially useful across ARM
platforms, even for non-PSCI use cases. Therefore make this function
common by returning 0 for success or -1 otherwise.
Having made the function common, make arm_validate_psci_entrypoint() a
wrapper around arm_validate_ns_entrypoint() which only translates return
value into PSCI error codes. This wrapper is now used where
arm_validate_ns_entrypoint() was currently used for PSCI entry point
validation.
Change-Id: Ic781fc3105d6d199fd8f53f01aba5baea0ebc310
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
This function implements the platform dependant part of PSCI system
reset2 for CSS platforms using SCMI.
Change-Id: I724389decab484043cadf577aeed96b349c1466d
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
Provides GICv3 save/restore feature to arm_system_pwr_domain_resume and
arm_system_pwr_domain_save functions.
Introduce FVP PSCI power level 3 (System level) support. This is solely
done to provide example code on how to use the GICv3 save and restore
helpers.
Also make CSS GICv3 platforms power off the Redistributor on SYSTEM
SUSPEND as its state is saved and restored.
Change-Id: I0d852f3af8824edee1a17c085cf593ddd33a4e77
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
Co-Authored-by: Douglas Raillard <douglas.raillard@arm.com>
mem_protect needs some kind of non-volatile memory because it has
to remember its state across reset and power down events.
The most suitable electronic part for this feature is a NVRAM
which should be only accesible from the secure world. Juno and
FVP lack such hardware and for this reason the MEM_PROTECT
functionality is implemented with Flash EEPROM memory on both
boards, even though this memory is accesible from the non-secure
world. This is done only to show a full implementation of
these PSCI features, but an actual system shouldn't use a
non-secure NVRAM to implement it.
The EL3 runtime software will write the mem_protect flag and BL2
will read and clear the memory ranges if enabled. It is done in
BL2 because it reduces the time that TF needs access to the full
non-secure memory.
The memory layout of both boards is defined using macros which
take different values in Juno and FVP platforms. Generic platform
helpers are added that use the platform specific macros to generate
a mem_region_t that is valid for the platform.
Change-Id: I2c6818ac091a2966fa07a52c5ddf8f6fde4941e9
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
This patch does the required changes to enable CSS platforms
to build and use the SDS framework. Since SDS is always coupled with
SCMI protocol, the preexisting SCMI build flag is now renamed to
`CSS_USE_SCMI_SDS_DRIVER` which will enable both SCMI and SDS on
CSS platforms. Also some of the workarounds applied for SCMI are
now removed with SDS in place.
Change-Id: I94e8b93f05e3fe95e475c5501c25bec052588a9c
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
This patch factors out common files required for sp_min for all CSS
platforms from the JUNO specific makefile to a the new `css_sp_min.mk`
makefile. This also allows the common build options that affect CSS
platforms to be configured in a central makefile for sp_min.
Change-Id: Ida952d8833b1aa5eda77ae0a6664a4632aeab24c
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
zero_normalmem only can zero memory when caches are enabled
and the target memory is a normal memory, otherwise an abort is
generated. In the case of EL3_PAYLOAD_BASE bl2_platform_setup was
calling zero_normalmem with device memory and it generated an abort.
Change-Id: If013603f209a12af488a9c54481f97a8f395b26a
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
The SCP_BL2 is transferred to SCP during BL2 image load and authenticate
sequence. The Boot-Over-MHU (BOM) protocol is used as transport for this. After
the SCP boots using the transferred image, the AP CPU waits till the `READY`
message is received from SCP. This patch separates the API for transport of
image from the wait for `READY` message and also moves the related files to
the `css/drivers` folder. The previous API `scp_bootloader_transfer` is
renamed to `css_scp_boot_image_xfer` to reflect the css naming convention.
This reorganisation also allows easier switch to a different transport
(eg: Shared Data Structure based transfer) in future
Change-Id: I8a96f9c4616ffde6dbfdf7c18f6f6f8bfa40bbf0
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
This patch adds the SCMI driver for communicating with SCP. The power
domain management and system power management protocol of the SCMI
specification[1] is implemented in the driver. The SCP power management
abstraction layer for SCMI for CSS power management is also added.
A new buid option `CSS_USE_SCMI_DRIVER` is introduced to select SCMI
driver over SCPI.
[1] ARM System Control and Management Interface v1.0 (SCMI)
Document number: ARM DEN 0056A
Change-Id: I67265615a17e679a2afe810b9b0043711ba09dbb
Signed-off-by: Soby Mathew <soby.mathew@arm.com>
To make software license auditing simpler, use SPDX[0] license
identifiers instead of duplicating the license text in every file.
NOTE: Files that have been imported by FreeBSD have not been modified.
[0]: https://spdx.org/
Change-Id: I80a00e1f641b8cc075ca5a95b10607ed9ed8761a
Signed-off-by: dp-arm <dimitris.papastamos@arm.com>
