The last CPU in a cluster is responsible for issuing the cluster power
down request to the FVP power controller. If another CPU in this
cluster wakes up before the last CPU enters WFI then the cluster power
down request remains pending. If this request is not cancelled and the
newly woken up CPU enters a simple WFI later, the power controller
powers the cluster down. This leads to unpredictable behaviour.
This patch fixes this issue by ensuring that the first CPU to wake up
in a cluster writes its MPIDR to the power controller's PPONR. This
cancels any pending cluster power down request.
Change-Id: I7e787adfd6c9a0bd7308390e3309d46f35c01086
The psci implementation does not track target affinity level requests
specified during cpu_suspend calls correctly as per the following
example.
1. cpu0.cluster0 calls cpu_suspend with the target affinity level as 0
2. Only the cpu0.cluster0 is powered down while cluster0 remains
powered up
3. cpu1.cluster0 calls cpu_off to power itself down to highest
possible affinity level
4. cluster0 will be powered off even though cpu0.cluster0 does not
allow cluster shutdown
This patch introduces reference counts at affinity levels > 0 to track
the number of cpus which want an affinity instance at level X to
remain powered up. This instance can be turned off only if its
reference count is 0. Cpus still undergo the normal state transitions
(ON, OFF, ON_PENDING, SUSPEND) but the higher levels can only be
either ON or OFF depending upon their reference count.
The above issue is thus fixed as follows:
1. cluster0's reference count is incremented by two when cpu0 and cpu1
are initially powered on.
2. cpu0.cluster0 calls cpu_suspend with the target affinity level as
0. This does not affect the cluster0 reference count.
3. Only the cpu0.cluster0 is powered down while cluster0 remains
powered up as it has a non-zero reference count.
4. cpu1.cluster0 call cpu_off to power itself down to highest possible
affinity level. This decrements the cluster0 reference count.
5. cluster0 is still not powered off since its reference count will at
least be 1 due to the restriction placed by cpu0.
Change-Id: I433dfe82b946f5f6985b1602c2de87800504f7a9
This patch adds support to save and restore the target affinity level
specified during a cpu_suspend psci call. This ensures that we
traverse only through the affinity levels that we originally intended
to after resuming from suspend.
Change-Id: I0900ae49a50b496da137cfec8f158da0397ec56c
The secure context saved and restored across a cpu_suspend operation
can be more than just the state of the secure system registers e.g. we
also need to save the affinity level till which the cpu is being
powered down. This patch creates a suspend_context data structure
which includes the system register context. This will allow other bits
to be saved and restored as well in subsequent patches.
Change-Id: I1c1f7d25497388b54b7d6ee4fab77e8c6a9992c4
The FVP platform has a few filenames that begin with fvp_. These are
renamed to plat_ to make it easier to use the FVP port as a template.
Change-Id: I601e6256d5ef3bae81a2e1f5df6de56db5b27069
Signed-off-by: Ryan Harkin <ryan.harkin@linaro.org>
Add the 'help' target to the Makefile to present a brief guide to the
various build options available.
Change-Id: Ic3a3489860b6362eb236470ea6b43a16a1b2fe3a
Signed-off-by: Ryan Harkin <ryan.harkin@linaro.org>
Tidy up the spacing of variable definitions within the makefiles to make
them more consistent, easier to read and amend.
Change-Id: Ic6d7c8489ca4330824abb5cd1ead8f1d449d1a85
Signed-off-by: Ryan Harkin <ryan.harkin@linaro.org>
Move all explicit platform or architecture specific references
into a new platform.mk file that is defined for each platform.
Change-Id: I9d6320d1ba957e0cc8d9b316b3578132331fa428
Signed-off-by: Ryan Harkin <ryan.harkin@linaro.org>
This patch replaces the empty definition of display_boot_progress() in
bl2_main.c with a weak definition. The former allowed bl2 to use the
early_exceptions(). It is possible to do that with a simple weak
definition as well.
Change-Id: Idb3f425a5e265f3579b638e3d26bd8c9bb78f80d
Ctags seem to have a problem with generating tags for assembler symbols
when a comment immediately follows an assembly label.
This patch inserts a single space character between the label
definition and the following comments to help ctags.
The patch is generated by the command:
git ls-files -- \*.S | xargs sed -i 's/^\([^:]\+\):;/\1: ;/1'
Change-Id: If7a3c9d0f51207ea033cc8b8e1b34acaa0926475
The GICv3 distributor can have more ports than CPUs are available in
the system. Probe all re-distributors and use the matching affinity
levels as specified by each core and re-distributor to decide which
re-distributor to use with which CPU core.
If a core cannot be matched with a re-distributor, the core panics and
is placed in an endless loop.
Change-Id: Ie393cfe07c7449a2383959e3c968664882e18afc
Traps when accessing architectural features are disabled by clearing bits
in CPTR_EL3 during early boot, including accesses to floating point
registers. The value of this register was previously undetermined, causing
unwanted traps to EL3. Future EL3 code (for example, context save/restore
code) may use floating point registers, although they are not used by current
code.
Also, the '-mgeneral-regs-only' flag is enabled in the GCC settings to
prevent generation of code that uses floating point registers.
Change-Id: I9a03675f6387bbbee81a6f2c9ccf81150db03747
GIC setup code which used to be in bl31_plat_setup.c is now in fvp_gic.c
to simplify future changes to other bootloader stages. This patch moves
code from bl31_plat_setup.c to fvp_gic.c, simplifies the include file
list for bl31_plat_setup.c, moves GIC declarations from the bl31.h header
file into the platform.h, and reworks files according to coding style
guide.
Change-Id: I48d82a4ba33e7114dcc88f9ca98767a06cf8f417
Tighten up ISO C standard checking. Fix 'CASSERT' implementation to
conform to C99 as opposed to GNU99 standard.
Change-Id: I58ddc61913617b66f11da5b6e3f7363136d5cf7d
In order to reduce the software dependency on the compiler, the project
is now compiled with the '-ffreestanding' flag. This is to prevent GCC
from replacing functions with more optimised versions. An example is
where GCC replaces a simple printf() with a puts().
Change-Id: I1973fe6957cd708e8432a0039af9d50e037bd884
SPSR is a 32-bit register and so its size should be reflected in
the gp_regs structure. This patch fixes the type of gp_regs.spsr
to use a 32-bit variable. It also makes the size of the other
register fields more explicit.
Change-Id: I27e0367df1a91cc501d5217c1b3856d4097c60ba
- Update porting guide to describe where files live and how to get
FreeBSD source code.
- Update change-log to describe relocation and new functions.
Change-Id: Id8f30cc7bafdd1064b3a5c5aae958c5aa3fb79f3
- This change is split into two separate patches in order to
simplify the history as interpreted by 'git'. The split is
between the move/rename and addition of new files.
- Remove dependency on toolchain C library headers and functions in
order to ensure behavioural compatibility between toolchains.
- Use FreeBSD as reference for C library implementation.
- Do not let GCC use default library include paths.
- Remove unused definitions in modified headers and implementations.
- Move C library files to 'lib/stdlib' and 'include/stdlib'.
- Break std.c functions out into separate files.
Change-Id: I3e3d8d992052264d2a02489034ae4c03bf0f5512
- This change is split into two separate patches in order to
simplify the history as interpreted by 'git'. The split is
between the move/rename and addition of new files.
- Remove dependency on toolchain C library headers and functions in
order to ensure behavioural compatibility between toolchains.
- Use FreeBSD as reference for C library implementation.
- Do not let GCC use default library include paths.
- Remove unused definitions in modified headers and implementations.
- Move C library files to 'lib/stdlib' and 'include/stdlib'.
- Break std.c functions out into separate files.
Change-Id: I91cddfb3229775f770ad781589670c57d347a154
- Add 'debug.h' with INFO, WARN and ERROR macros.
- This prints the specified message with the appropriate tag.
- INFO and WARN messages are only displayed when building with
the DEBUG flag set. Error messages are always printed.
Change-Id: I21835b6063fcc99649b30ac7489387cbd3705bc0
ns_entry_info used to be a per-cpu array. This is a waste of space
because it is only accessed by the primary CPU on the cold boot path.
This patch reduces ns_entry_info to a single-cpu area.
Change-Id: I647c70c4e76069560f1aaad37a1d5910f56fba4c
Platform setup code has to reserve some memory for storing the
memory layout information. It is populated in early platform setup
code.
blx_get_sec_mem_layout() functions used to return a copy of this
structure. This patch modifies blx_get_sec_mem_layout() functions
so that they now directly return a pointer to their memory layout
structure. It ensures that the memory layout returned by
blx_get_sec_mem_layout() is always up-to-date and also avoids a
useless copy of the meminfo structure.
Also rename blx_get_sec_mem_layout() to blx_plat_sec_mem_layout()
to make it clear those functions are platform specific.
Change-Id: Ic7a6f9d6b6236b14865ab48a9f5eff545ce56551
This patch makes changes to the documents to reflect the current
state of play of the psci cpu_suspend function.
Change-Id: I086509fb75111b6e9f93b7f6dbcd33cc4591b9f3
This patch performs a major rework of the psci generic implementation
to achieve the following:
1. replace recursion with iteration where possible to aid code
readability e.g. affinity instance states are changed iteratively
instead of recursively.
2. acquire pointers to affinity instance nodes at the beginning of a
psci operation. All subsequent actions use these pointers instead
of calling psci_get_aff_map_node() repeatedly e.g. management of
locks has been abstracted under functions which use these pointers
to ensure correct ordering. Helper functions have been added to
create these abstractions.
3. assertions have been added to cpu level handlers to ensure correct
state transition
4. the affinity level extents specified to various functions have the
same meaning i.e. start level is always less than the end level.
Change-Id: If0508c3a7b20ea3ddda2a66128429382afc3dfc8
This patch:
1. removes a duplicate assertion to check that the only error
condition that can be returned while turning a cpu off is
PSCI_E_DENIED. Having this assertion after calling
psci_afflvl_off() is sufficient.
2. corrects some incorrect usage of 'its' vs 'it is'
3. removes some unwanted white spaces
Change-Id: Icf014e269b54f5be5ce0b9fbe6b41258e4ebf403
In fvp_affinst_on/suspend, the non-secure entrypoint is always
expected to lie in the DRAM. This check will not be valid if
non-secure code executes directly out of flash e.g. a baremetal
test. This patch removes this check.
Change-Id: I0436e1138fc394aae8ff1ea59ebe38b46a440b61
This patch removes the duplicate declaration of psci_cpu_on in psci.h
and moves the constants for the system level implementation of the
generic timer from arch_helpers.h to arch.h. All other architectural
constants are defined in arch.h so there is no need to add them to
arch_helpers.h
Change-Id: Ia8ad3f91854f7e57fce31873773eede55c384ff1
In the previous psci implementation, the psci_afflvl_power_on_finish()
function would run into an error condition if the value of the context
id parameter in the cpu_on and cpu_suspend psci calls was != 0. The
parameter was being restored as the return value of the affinity level
0 finisher function. A non zero context id would be treated as an
error condition. This would prevent successful wake up of the cpu from
a power down state. Also, the contents of the general purpose
registers were not being cleared upon return to the non-secure world
after a cpu power up. This could potentially allow the non-secure
world to view secure data.
This patch ensures that all general purpose registers are set to ~0
prior to the final eret that drops the execution to the non-secure
world. The context id is used to initialize the general purpose
register x0 prior to re-entry into the non-secure world and is no
longer restored as a function return value. A platform helper
(platform_get_stack()) has been introduced to facilitate this change.
Change-Id: I2454911ffd75705d6aa8609a5d250d9b26fa097c
This patch fixes the following constant values in the psci.h:
1. The affinity level shift value in the power_state parameter of the
cpu_suspend psci call. The previous value was preventing shutdown
of the affinity level 1.
2. The values used for affinity state constants (ON, OFF,
ON_PENDING). They did not match the values expected to be returned
by the affinity_info psci api as mentioned in the spec.
3. The state id shift value in the power_state parameter of the
cpu_suspend psci call.
Change-Id: I62ed5eb0e9640b4aa97b93923d6630e6b877a097
The FVP specific code that gets called after a cpu has been physically
powered on after having been turned off or suspended earlier does not
clear the PWRC.PWKUPR.WEN bit. Not doing so causes problems if: a cpu
is suspended, woken from suspend, powered down through a cpu_off call
& receives a spurious interrupt. Since the WEN bit is not cleared
after the cpu woke up from suspend, the spurious wakeup will power the
cpu on. Since the cpu_off call clears the jump address in the mailbox
this spurious wakeup will cause the cpu to crash.
This patch fixes this issue by clearing the WEN bit whenever a cpu is
powered up.
Change-Id: Ic91f5dffe1ed01d76bc7fc807acf0ecd3e38ce5b
The runtime exception handling assembler code used magic numbers for
saving and restoring the general purpose register context on stack
memory. The memory is interpreted as a 'gp_regs' structure and the
magic numbers are offsets to members of this structure. This patch
replaces the magic number offsets with constants. It also adds compile
time assertions to prevent an incorrect assembler view of this
structure.
Change-Id: Ibf125bfdd62ba3a33e58c5f1d71f8c229720781c
- Add instructions for contributing to ARM Trusted Firmware.
- Update copyright text in all files to acknowledge contributors.
Change-Id: I9311aac81b00c6c167d2f8c889aea403b84450e5
This patch updates the user guide section about the memory layout.
- Explain the verifications that the linker scripts does on the
global memory layout.
- Refer to the new linker symbols.
- Describe the linker symbols exported to the trusted firmware code.
Change-Id: I033ab2b867e8b9776deb4185b9986bcb8218f286
This patch makes sure the C runtime environment is properly
initialised before executing any C code.
- Zero-initialise NOBITS sections (e.g. the bss section).
- Relocate BL1 data from ROM to RAM.
Change-Id: I0da81b417b2f0d1f7ef667cc5131b1e47e22571f
- Check at link-time that bootloader images will fit in memory
at run time and that they won't overlap each other.
- Remove text and rodata orphan sections.
- Define new linker symbols to remove the need for platform setup
code to know the order of sections.
- Reduce the size of the raw binary images by cutting some sections
out of the disk image and allocating them at load time, whenever
possible.
- Rework alignment constraints on sections.
- Remove unused linker symbols.
- Homogenize linker symbols names across all BLs.
- Add some comments in the linker scripts.
Change-Id: I47a328af0ccc7c8ab47fcc0dc6e7dd26160610b9
A single binary can be compiled using a command such as:
make CROSS_COMPILE=aarch64-none-elf- bl1
Also make use of brackets consistent in the Makefile.
Change-Id: I2180fdb473411ef7cffe39670a7b2de82def812e
- Large RAM-disks may have trouble starting with 2GB of memory.
- Increase from 2GB to 4GB in FDT.
Change-Id: I12c1b8e5db41114b88c69c48621cb21247a6a6a7
On FVP platforms, for now it is assumed that the normal-world
bootloader is already sitting in its final memory location.
Therefore, BL2 doesn't need to load it and so it doesn't need
to know the extents of the non-trusted DRAM.
Change-Id: I33177ab43ca242edc8958f2fa8d994e7cf3e0843
Also, don't invalidate the TLBs in disable_mmu() function, it's better
to do it in enable_mmu() function just before actually enabling the
MMU.
Change-Id: Ib32d6660019b0b2c17254156aad4be67ab4970e1
Any asynchronous exception caused by the firmware should be handled
in the firmware itself. For this reason, unmask SError exceptions
(and Debug ones as well) on all boot paths. Also route external
abort and SError interrupts to EL3, otherwise they will target EL1.
Change-Id: I9c191d2d0dcfef85f265641c8460dfbb4d112092
Global and static variables are expected to be initialised to zero
by default. This is specified by the C99 standard. This patch
removes some unnecessary initialisations of such variables.
It fixes a compilation warning at the same time:
plat/fvp/bl31_plat_setup.c:82:3: warning: missing braces around
initializer [-Wmissing-braces]
section("tzfw_coherent_mem"))) = {0};
^
plat/fvp/bl31_plat_setup.c:82:3: warning: (near initialization for
‘ns_entry_info[0]’) [-Wmissing-braces]
Note that GCC should not have emitted this warning message in the
first place. The C Standard permits braces to be elided around
subaggregate initializers. See this GCC bug report:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53119
Change-Id: I13cb0c344feb9803bca8819f976377741fa6bc35
blx_plat_arch_setup() should only perform platform-specific
architectural setup, e.g. enabling the MMU. This patch moves
generic architectural setup code out of blx_plat_arch_setup().
Change-Id: I4ccf56b8c4a2fa84909817779a2d97a14aaafab6
Joakim Bech
Achin Gupta
Ryan Harkin
Jeenu Viswambharan
Harry Liebel
Dan Handley
Ian Spray
Sandrine Bailleux
James Morrissey