GitLab

The use of end addresses is preferred over the size of sections.
This was done for some AARCH64 files for PIE with commit [1],
and some extra explanations can be found in its commit message.
Align the missing AARCH64 files.

For AARCH32 files, this is required to prepare PIE support introduction.

 [1] f1722b69

 ("PIE: Use PC relative adrp/adr for symbol reference")

Change-Id: I8f1c06580182b10c680310850f72904e58a54d7d
Signed-off-by: Yann Gautier <yann.gautier@st.com>

This patch fixes the following compilation error
reported by aarch64-none-elf-gcc 11.0.0:

bl32/tsp/tsp_main.c: In function 'tsp_smc_handler':
bl32/tsp/tsp_main.c:393:9: error: 'tsp_get_magic'
 accessing 32 bytes in a region of size 16
 [-Werror=stringop-overflow=]
  393 |         tsp_get_magic(service_args);
      |         ^~~~~~~~~~~~~~~~~~~~~~~~~~~
bl32/tsp/tsp_main.c:393:9: note: referencing argument 1
 of type 'uint64_t *' {aka 'long long unsigned int *'}
In file included from bl32/tsp/tsp_main.c:19:
bl32/tsp/tsp_private.h:64:6: note: in a call to function 'tsp_get_magic'
   64 | void tsp_get_magic(uint64_t args[4]);
      |      ^~~~~~~~~~~~~

by changing declaration of tsp_get_magic function from
void tsp_get_magic(uint64_t args[4]);
to
uint128_t tsp_get_magic(void);
which returns arguments directly in x0 and x1 registers.

In bl32\tsp\tsp_main.c the current tsp_smc_handler()
implementation calls tsp_get_magic(service_args);
, where service_args array is declared as
uint64_t service_args[2];
and tsp_get_magic() in bl32\tsp\aarch64\tsp_request.S
copies only 2 registers in output buffer:
	/* Store returned arguments to the array */
	stp	x0, x1, [x4, #0]

Change-Id: Ib34759fc5d7bb803e6c734540d91ea278270b330
Signed-off-by: Alexei Fedorov <Alexei.Fedorov@arm.com>

Usually, C has no problem up-converting types to larger bit sizes. MISRA
rule 10.7 requires that you not do this, or be very explicit about this.
This resolves the following required rule:

    bl1/aarch64/bl1_context_mgmt.c:81:[MISRA C-2012 Rule 10.7 (required)]<None>
    The width of the composite expression "0U | ((mode & 3U) << 2U) | 1U |
    0x3c0U" (32 bits) is less that the right hand operand
    "18446744073709547519ULL" (64 bits).

This also resolves MISRA defects such as:

    bl2/aarch64/bl2arch_setup.c:18:[MISRA C-2012 Rule 12.2 (required)]
    In the expression "3U << 20", shifting more than 7 bits, the number
    of bits in the essential type of the left expression, "3U", is
    not allowed.

Further, MISRA requires that all shifts don't overflow. The definition of
PAGE_SIZE was (1U << 12), and 1U is 8 bits. This caused about 50 issues.
This fixes the violation by changing the definition to 1UL << 12. Since
this uses 32bits, it should not create any issues for aarch32.

This patch also contains a fix for a build failure in the sun50i_a64
platform. Specifically, these misra fixes removed a single and
instruction,

    92407e73        and     x19, x19, #0xffffffff

from the cm_setup_context function caused a relocation in
psci_cpus_on_start to require a linker-generated stub. This increased the
size of the .text section and caused an alignment later on to go over a
page boundary and round up to the end of RAM before placing the .data
section. This sectionn is of non-zero size and therefore causes a link
error.

The fix included in this reorders the functions during link time
without changing their ording with respect to alignment.

Change-Id: I76b4b662c3d262296728a8b9aab7a33b02087f16
Signed-off-by: Jimmy Brisson <jimmy.brisson@arm.com>

This implementation simply mimics that of BL31.

Change-Id: Ibbaa4ca012d38ac211c52b0b3e97449947160e07
Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>

Even though ERET always causes a jump to another address, aarch64 CPUs
speculatively execute following instructions as if the ERET
instruction was not a jump instruction.
The speculative execution does not cross privilege-levels (to the jump
target as one would expect), but it continues on the kernel privilege
level as if the ERET instruction did not change the control flow -
thus execution anything that is accidentally linked after the ERET
instruction. Later, the results of this speculative execution are
always architecturally discarded, however they can leak data using
microarchitectural side channels. This speculative execution is very
reliable (seems to be unconditional) and it manages to complete even
relatively performance-heavy operations (e.g. multiple dependent
fetches from uncached memory).

This was fixed in Linux, FreeBSD, OpenBSD and Optee OS:
https://github.com/torvalds/linux/commit/679db70801da9fda91d26caf13bf5b5ccc74e8e8
https://github.com/freebsd/freebsd/commit/29fb48ace4186a41c409fde52bcf4216e9e50b61
https://github.com/openbsd/src/commit/3a08873ece1cb28ace89fd65e8f3c1375cc98de2
https://github.com/OP-TEE/optee_os/commit/abfd092aa19f9c0251e3d5551e2d68a9ebcfec8a

It is demonstrated in a SafeSide example:
https://github.com/google/safeside/blob/master/demos/eret_hvc_smc_wrapper.cc
https://github.com/google/safeside/blob/master/kernel_modules/kmod_eret_hvc_smc/eret_hvc_smc_module.c

Signed-off-by: Anthony Steinhauser <asteinhauser@google.com>
Change-Id: Iead39b0b9fb4b8d8b5609daaa8be81497ba63a0f

This patch provides the following features and makes modifications
listed below:
- Individual APIAKey key generation for each CPU.
- New key generation on every BL31 warm boot and TSP CPU On event.
- Per-CPU storage of APIAKey added in percpu_data[]
  of cpu_data structure.
- `plat_init_apiakey()` function replaced with `plat_init_apkey()`
  which returns 128-bit value and uses Generic timer physical counter
  value to increase the randomness of the generated key.
  The new function can be used for generation of all ARMv8.3-PAuth keys
- ARMv8.3-PAuth specific code placed in `lib\extensions\pauth`.
- New `pauth_init_enable_el1()` and `pauth_init_enable_el3()` functions
  generate, program and enable APIAKey_EL1 for EL1 and EL3 respectively;
  pauth_disable_el1()` and `pauth_disable_el3()` functions disable
  PAuth for EL1 and EL3 respectively;
  `pauth_load_bl31_apiakey()` loads saved per-CPU APIAKey_EL1 from
  cpu-data structure.
- Combined `save_gp_pauth_registers()` function replaces calls to
  `save_gp_registers()` and `pauth_context_save()`;
  `restore_gp_pauth_registers()` replaces `pauth_context_restore()`
  and `restore_gp_registers()` calls.
- `restore_gp_registers_eret()` function removed with corresponding
  code placed in `el3_exit()`.
- Fixed the issue when `pauth_t pauth_ctx` structure allocated space
  for 12 uint64_t PAuth registers instead of 10 by removal of macro
  CTX_PACGAKEY_END from `include/lib/el3_runtime/aarch64/context.h`
  and assigning its value to CTX_PAUTH_REGS_END.
- Use of MODE_SP_ELX and MODE_SP_EL0 macro definitions
  in `msr	spsel`  instruction instead of hard-coded values.
- Changes in documentation related to ARMv8.3-PAuth and ARMv8.5-BTI.

Change-Id: Id18b81cc46f52a783a7e6a09b9f149b6ce803211
Signed-off-by: Alexei Fedorov <Alexei.Fedorov@arm.com>

This patch adds the functionality needed for platforms to provide
Branch Target Identification (BTI) extension, introduced to AArch64
in Armv8.5-A by adding BTI instruction used to mark valid targets
for indirect branches. The patch sets new GP bit [50] to the stage 1
Translation Table Block and Page entries to denote guarded EL3 code
pages which will cause processor to trap instructions in protected
pages trying to perform an indirect branch to any instruction other
than BTI.
BTI feature is selected by BRANCH_PROTECTION option which supersedes
the previous ENABLE_PAUTH used for Armv8.3-A Pointer Authentication
and is disabled by default. Enabling BTI requires compiler support
and was tested with GCC versions 9.0.0, 9.0.1 and 10.0.0.
The assembly macros and helpers are modified to accommodate the BTI
instruction.
This is an experimental feature.
Note. The previous ENABLE_PAUTH build option to enable PAuth in EL3
is now made as an internal flag and BRANCH_PROTECTION flag should be
used instead to enable Pointer Authentication.
Note. USE_LIBROM=1 option is currently not supported.

Change-Id: Ifaf4438609b16647dc79468b70cd1f47a623362e
Signed-off-by: Alexei Fedorov <Alexei.Fedorov@arm.com>

The SCTLR.DSSBS bit is zero by default thus disabling speculative loads.
However, we also explicitly set it to zero for BL2 and TSP images when
each image initialises its context. This is done to ensure that the
image environment is initialised in a safe state, regardless of the
reset value of the bit.

Change-Id: If25a8396641edb640f7f298b8d3309d5cba3cd79
Signed-off-by: John Tsichritzis <john.tsichritzis@arm.com>

The size increase after enabling options related to ARMv8.3-PAuth is:

+----------------------------+-------+-------+-------+--------+
|                            |  text |  bss  |  data | rodata |
+----------------------------+-------+-------+-------+--------+
| CTX_INCLUDE_PAUTH_REGS = 1 |   +40 |   +0  |   +0  |   +0   |
|                            |  0.4% |       |       |        |
+----------------------------+-------+-------+-------+--------+
| ENABLE_PAUTH = 1           |  +352 |    +0 |  +16  |   +0   |
|                            |  3.1% |       | 15.8% |        |
+----------------------------+-------+-------+-------+--------+

Results calculated with the following build configuration:

    make PLAT=fvp SPD=tspd DEBUG=1 \
    SDEI_SUPPORT=1                 \
    EL3_EXCEPTION_HANDLING=1       \
    TSP_NS_INTR_ASYNC_PREEMPT=1    \
    CTX_INCLUDE_PAUTH_REGS=1       \
    ENABLE_PAUTH=1

Change-Id: I6cc1fe0b2345c547dcef66f98758c4eb55fe5ee4
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: e0ea0928 ("Fix gpio includes of mt8173 platform
to avoid collision."). More recently, this patch has had similar
problems: 46f9b2c3 ("drivers: add tzc380 support").

This problem was introduced in commit 4ecca339

 ("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>

Check_vector_size checks if the size of the vector fits
in the size reserved for it. This check creates problems in
the Clang assembler. A new macro, end_vector_entry, is added
and check_vector_size is deprecated.

This new macro fills the current exception vector until the next
exception vector. If the size of the current vector is bigger
than 32 instructions then it gives an error.

Change-Id: Ie8545cf1003a1e31656a1018dd6b4c28a4eaf671
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>

Previously, data caches were disabled while enabling MMU only because of
active stack. Now that we can enable MMU without using stack, we can
enable both MMU and data caches at the same time.

Change-Id: I73f3b8bae5178610e17e9ad06f81f8f6f97734a6
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>

Some error paths that lead to a crash dump will overwrite the value in
the x30 register by calling functions with the no_ret macro, which
resolves to a BL instruction. This is not very useful and not what the
reader would expect, since a crash dump should usually show all
registers in the state they were in when the exception happened. This
patch replaces the offending function calls with a B instruction to
preserve the value in x30.

Change-Id: I2a3636f2943f79bab0cd911f89d070012e697c2a
Signed-off-by: Julius Werner <jwerner@chromium.org>

Assembler programmers are used to being able to define functions with a
specific aligment with a pattern like this:

    .align X
  myfunction:

However, this pattern is subtly broken when instead of a direct label
like 'myfunction:', you use the 'func myfunction' macro that's standard
in Trusted Firmware. Since the func macro declares a new section for the
function, the .align directive written above it actually applies to the
*previous* section in the assembly file, and the function it was
supposed to apply to is linked with default alignment.

An extreme case can be seen in Rockchip's plat_helpers.S which contains
this code:

  [...]
  endfunc plat_crash_console_putc

  .align 16
  func platform_cpu_warmboot
  [...]

This assembles into the following plat_helpers.o:

  Sections:
  Idx Name                             Size  [...]  Algn
   9 .text.plat_crash_console_putc 00010000  [...]  2**16
  10 .text.platform_cpu_warmboot   00000080  [...]  2**3

As can be seen, the *previous* function actually got the alignment
constraint, and it is also 64KB big even though it contains only two
instructions, because the .align directive at the end of its section
forces the assembler to insert a giant sled of NOPs. The function we
actually wanted to align has the default constraint. This code only
works at all because the linker just happens to put the two functions
right behind each other when linking the final image, and since the end
of plat_crash_console_putc is aligned the start of platform_cpu_warmboot
will also be. But it still wastes almost 64KB of image space
unnecessarily, and it will break under certain circumstances (e.g. if
the plat_crash_console_putc function becomes unused and its section gets
garbage-collected out).

There's no real way to fix this with the existing func macro. Code like

 func myfunc
 .align X

happens to do the right thing, but is still not really correct code
(because the function label is inserted before the .align directive, so
the assembler is technically allowed to insert padding at the beginning
of the function which would then get executed as instructions if the
function was called). Therefore, this patch adds a new parameter with a
default value to the func macro that allows overriding its alignment.

Also fix up all existing instances of this dangerous antipattern.

Change-Id: I5696a07e2fde896f21e0e83644c95b7b6ac79a10
Signed-off-by: Julius Werner <jwerner@chromium.org>

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>

Since Issue B (November 2016) of the SMC Calling Convention document
standard SMC calls are renamed to yielding SMC calls to help avoid
confusion with the standard service SMC range, which remains unchanged.

http://infocenter.arm.com/help/topic/com.arm.doc.den0028b/ARM_DEN0028B_SMC_Calling_Convention.pdf



This patch adds a new define for yielding SMC call type and deprecates
the current standard SMC call type. The tsp is migrated to use this new
terminology and, additionally, the documentation and code comments are
updated to use this new terminology.

Change-Id: I0d7cc0224667ee6c050af976745f18c55906a793
Signed-off-by: David Cunado <david.cunado@arm.com>

Introduce new build option ENABLE_STACK_PROTECTOR. It enables
compilation of all BL images with one of the GCC -fstack-protector-*
options.

A new platform function plat_get_stack_protector_canary() is introduced.
It returns a value that is used to initialize the canary for stack
corruption detection. Returning a random value will prevent an attacker
from predicting the value and greatly increase the effectiveness of the
protection.

A message is printed at the ERROR level when a stack corruption is
detected.

To be effective, the global data must be stored at an address
lower than the base of the stacks. Failure to do so would allow an
attacker to overwrite the canary as part of an attack which would void
the protection.

FVP implementation of plat_get_stack_protector_canary is weak as
there is no real source of entropy on the FVP. It therefore relies on a
timer's value, which could be predictable.

Change-Id: Icaaee96392733b721fa7c86a81d03660d3c1bc06
Signed-off-by: Douglas Raillard <douglas.raillard@arm.com>

The files affected by this patch don't really depend on `xlat_tables.h`.
By changing the included file it becomes easier to switch between the
two versions of the translation tables library.

Change-Id: Idae9171c490e0865cb55883b19eaf942457c4ccc
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>

Introduce zeromem_dczva function on AArch64 that can handle unaligned
addresses and make use of DC ZVA instruction to zero a whole block at a
time. This zeroing takes place directly in the cache to speed it up
without doing external memory access.

Remove the zeromem16 function on AArch64 and replace it with an alias to
zeromem. This zeromem16 function is now deprecated.

Remove the 16-bytes alignment constraint on __BSS_START__ in
firmware-design.md as it is now not mandatory anymore (it used to comply
with zeromem16 requirements).

Change the 16-bytes alignment constraints in SP min's linker script to a
8-bytes alignment constraint as the AArch32 zeromem implementation is now
more efficient on 8-bytes aligned addresses.

Introduce zero_normalmem and zeromem helpers in platform agnostic header
that are implemented this way:
* AArch32:
	* zero_normalmem: zero using usual data access
	* zeromem: alias for zero_normalmem
* AArch64:
	* zero_normalmem: zero normal memory  using DC ZVA instruction
	                  (needs MMU enabled)
	* zeromem: zero using usual data access

Usage guidelines: in most cases, zero_normalmem should be preferred.

There are 2 scenarios where zeromem (or memset) must be used instead:
* Code that must run with MMU disabled (which means all memory is
  considered device memory for data accesses).
* Code that fills device memory with null bytes.

Optionally, the following rule can be applied if performance is
important:
* Code zeroing small areas (few bytes) that are not secrets should use
  memset to take advantage of compiler optimizations.

  Note: Code zeroing security-related critical information should use
  zero_normalmem/zeromem instead of memset to avoid removal by
  compilers' optimizations in some cases or misbehaving versions of GCC.

Fixes ARM-software/tf-issues#408

Change-Id: Iafd9663fc1070413c3e1904e54091cf60effaa82
Signed-off-by: Douglas Raillard <douglas.raillard@arm.com>

Standard SMC requests that are handled in the secure-world by the Secure
Payload can be preempted by interrupts that must be handled in the
normal world. When the TSP is preempted the secure context is stored and
control is passed to the normal world to handle the non-secure
interrupt. Once completed the preempted secure context is restored. When
restoring the preempted context, the dispatcher assumes that the TSP
preempted context is still stored as the SECURE context by the context
management library.

However, PSCI power management operations causes synchronous entry into
TSP. This overwrites the preempted SECURE context in the context
management library. When restoring back the SECURE context, the Secure
Payload crashes because this context is not the preempted context
anymore.

This patch avoids corruption of the preempted SECURE context by aborting
any preempted SMC during PSCI power management calls. The
abort_std_smc_entry hook of the TSP is called when aborting the SMC
request.

It also exposes this feature as a FAST SMC callable from normal world to
abort preempted SMC with FID TSP_FID_ABORT.

Change-Id: I7a70347e9293f47d87b5de20484b4ffefb56b770
Signed-off-by: Douglas Raillard <douglas.raillard@arm.com>

There are many instances in ARM Trusted Firmware where control is
transferred to functions from which return isn't expected. Such jumps
are made using 'bl' instruction to provide the callee with the location
from which it was jumped to. Additionally, debuggers infer the caller by
examining where 'lr' register points to. If a 'bl' of the nature
described above falls at the end of an assembly function, 'lr' will be
left pointing to a location outside of the function range. This misleads
the debugger back trace.

This patch defines a 'no_ret' macro to be used when jumping to functions
from which return isn't expected. The macro ensures to use 'bl'
instruction for the jump, and also, for debug builds, places a 'nop'
instruction immediately thereafter (unless instructed otherwise) so as
to leave 'lr' pointing within the function range.

Change-Id: Ib34c69fc09197cfd57bc06e147cc8252910e01b0
Co-authored-by: Douglas Raillard <douglas.raillard@arm.com>
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>

This patch introduces some assembler macros to simplify the
declaration of the exception vectors. It abstracts the section
the exception code is put into as well as the alignments
constraints mandated by the ARMv8 architecture. For all TF images,
the exception code has been updated to make use of these macros.

This patch also updates some invalid comments in the exception
vector code.

Change-Id: I35737b8f1c8c24b6da89b0a954c8152a4096fa95

Added a new platform porting function plat_panic_handler, to allow
platforms to handle unexpected error situations. It must be
implemented in assembly as it may be called before the C environment
is initialized. A default implementation is provided, which simply
spins.

Corrected all dead loops in generic code to call this function
instead. This includes the dead loop that occurs at the end of the
call to panic().

All unnecesary wfis from bl32/tsp/aarch64/tsp_exceptions.S have
been removed.

Change-Id: I67cb85f6112fa8e77bd62f5718efcef4173d8134

Earlier the TSP only ever expected to be preempted during Standard SMC
processing. If a S-EL1 interrupt triggered while in the normal world, it
will routed to S-EL1 `synchronously` for handling. The `synchronous` S-EL1
interrupt handler `tsp_sel1_intr_entry` used to panic if this S-EL1 interrupt
was preempted by another higher priority pending interrupt which should be
handled in EL3 e.g. Group0 interrupt in GICv3.

With this patch, the `tsp_sel1_intr_entry` now expects `TSP_PREEMPTED` as the
return code from the `tsp_common_int_handler` in addition to 0 (interrupt
successfully handled) and in both cases it issues an SMC with id
`TSP_HANDLED_S_EL1_INTR`. The TSPD switches the context and returns back
to normal world. In case a higher priority EL3 interrupt was pending, the
execution will be routed to EL3 where interrupt will be handled. On return
back to normal world, the pending S-EL1 interrupt which was preempted will
get routed to S-EL1 to be handled `synchronously` via `tsp_sel1_intr_entry`.

Change-Id: I2087c7fedb37746fbd9200cdda9b6dba93e16201

On a GICv2 system, interrupts that should be handled in the secure world are
typically signalled as FIQs. On a GICv3 system, these interrupts are signalled
as IRQs instead. The mechanism for handling both types of interrupts is the same
in both cases. This patch enables the TSP to run on a GICv3 system by:

1. adding support for handling IRQs in the exception handling code.
2. removing use of "fiq" in the names of data structures, macros and functions.

The build option TSPD_ROUTE_IRQ_TO_EL3 is deprecated and is replaced with a
new build flag TSP_NS_INTR_ASYNC_PREEMPT. For compatibility reasons, if the
former build flag is defined, it will be used to define the value for the
new build flag. The documentation is also updated accordingly.

Change-Id: I1807d371f41c3656322dd259340a57649833065e

The TSP is expected to pass control back to EL3 if it gets preempted due to
an interrupt while handling a Standard SMC in the following scenarios:

1. An FIQ preempts Standard SMC execution and that FIQ is not a TSP Secure
   timer interrupt or is preempted by a higher priority interrupt by the time
   the TSP acknowledges it. In this case, the TSP issues an SMC with the ID
   as `TSP_EL3_FIQ`. Currently this case is never expected to happen as only
   the TSP Secure Timer is expected to generate FIQ.

2. An IRQ preempts Standard SMC execution and in this case the TSP issues
   an SMC with the ID as `TSP_PREEMPTED`.

In both the cases, the TSPD hands control back to the normal world and returns
returns an error code to the normal world to indicate that the standard SMC it
had issued has been preempted but not completed.

This patch unifies the handling of these two cases in the TSPD and ensures that
the TSP only uses TSP_PREEMPTED instead of separate SMC IDs. Also instead of 2
separate error codes, SMC_PREEMPTED and TSP_EL3_FIQ, only SMC_PREEMPTED is
returned as error code back to the normal world.

Background information: On a GICv3 system, when the secure world has affinity
routing enabled, in 2. an FIQ will preempt TSP execution instead of an IRQ. The
FIQ could be a result of a Group 0 or a Group 1 NS interrupt. In both case, the
TSPD passes control back to the normal world upon receipt of the TSP_PREEMPTED
SMC. A Group 0 interrupt will immediately preempt execution to EL3 where it
will be handled. This allows for unified interrupt handling in TSP for both
GICv3 and GICv2 systems.

Change-Id: I9895344db74b188021e3f6a694701ad272fb40d4

On the ARMv8 architecture, cache maintenance operations by set/way on the last
level of integrated cache do not affect the system cache. This means that such a
flush or clean operation could result in the data being pushed out to the system
cache rather than main memory. Another CPU could access this data before it
enables its data cache or MMU. Such accesses could be serviced from the main
memory instead of the system cache. If the data in the sysem cache has not yet
been flushed or evicted to main memory then there could be a loss of
coherency. The only mechanism to guarantee that the main memory will be updated
is to use cache maintenance operations to the PoC by MVA(See section D3.4.11
(System level caches) of ARMv8-A Reference Manual (Issue A.g/ARM DDI0487A.G).

This patch removes the reliance of Trusted Firmware on the flush by set/way
operation to ensure visibility of data in the main memory. Cache maintenance
operations by MVA are now used instead. The following are the broad category of
changes:

1. The RW areas of BL2/BL31/BL32 are invalidated by MVA before the C runtime is
   initialised. This ensures that any stale cache lines at any level of cache
   are removed.

2. Updates to global data in runtime firmware (BL31) by the primary CPU are made
   visible to secondary CPUs using a cache clean operation by MVA.

3. Cache maintenance by set/way operations are only used prior to power down.

NOTE: NON-UPSTREAM TRUSTED FIRMWARE CODE SHOULD MAKE EQUIVALENT CHANGES IN
ORDER TO FUNCTION CORRECTLY ON PLATFORMS WITH SUPPORT FOR SYSTEM CACHES.

Fixes ARM-software/tf-issues#205

Change-Id: I64f1b398de0432813a0e0881d70f8337681f6e9a

The new PSCI frameworks mandates that the platform APIs and the various
frameworks in Trusted Firmware migrate away from MPIDR based core
identification to one based on core index. Deprecated versions of the old
APIs are still present to provide compatibility but their implementations
are not optimal. This patch migrates the various SPDs exisiting within
Trusted Firmware tree and TSP to the new APIs.

Change-Id: Ifc37e7071c5769b5ded21d0b6a071c8c4cab7836

In order for the symbol table in the ELF file to contain the size of
functions written in assembly, it is necessary to report it to the
assembler using the .size directive.

To fulfil the above requirements, this patch introduces an 'endfunc'
macro which contains the .endfunc and .size directives. It also adds
a .func directive to the 'func' assembler macro.

The .func/.endfunc have been used so the assembler can fail if
endfunc is omitted.

Fixes ARM-Software/tf-issues#295

Change-Id: If8cb331b03d7f38fe7e3694d4de26f1075b278fc
Signed-off-by: Kévin Petit <kevin.petit@arm.com>

This patch extends the build option `USE_COHERENT_MEMORY` to
conditionally remove coherent memory from the memory maps of
all boot loader stages. The patch also adds necessary
documentation for coherent memory removal in firmware-design,
porting and user guides.

Fixes ARM-Software/tf-issues#106

Change-Id: I260e8768c6a5c2efc402f5804a80657d8ce38773

This patch adds support for SYSTEM_OFF and SYSTEM_RESET PSCI
operations. A platform should export handlers to complete the
requested operation. The FVP port exports fvp_system_off() and
fvp_system_reset() as an example.

If the SPD provides a power management hook for system off and
system reset, then the SPD is notified about the corresponding
operation so it can do some bookkeeping. The TSPD exports
tspd_system_off() and tspd_system_reset() for that purpose.

Versatile Express shutdown and reset methods have been removed
from the FDT as new PSCI sys_poweroff and sys_reset services
have been added. For those kernels that do not support yet these
PSCI services (i.e. GICv3 kernel), the original dtsi files have
been renamed to *-no_psci.dtsi.

Fixes ARM-software/tf-issues#218

Change-Id: Ic8a3bf801db979099ab7029162af041c4e8330c8

* Move TSP platform porting functions to new file:
  include/bl32/tsp/platform_tsp.h.

* Create new TSP_IRQ_SEC_PHY_TIMER definition for use by the generic
  TSP interrupt handling code, instead of depending on the FVP
  specific definition IRQ_SEC_PHY_TIMER.

* Rename TSP platform porting functions from bl32_* to tsp_*, and
  definitions from BL32_* to TSP_*.

* Update generic TSP code to use new platform porting function names
  and definitions.

* Update FVP port accordingly and move all TSP source files to:
  plat/fvp/tsp/.

* Update porting guide with above changes.

Note: THIS CHANGE REQUIRES ALL PLATFORM PORTS OF THE TSP TO
      BE UPDATED

Fixes ARM-software/tf-issues#167

Change-Id: Ic0ff8caf72aebb378d378193d2f017599fc6b78f

This patch disables routing of external aborts from lower exception levels to
EL3 and ensures that a SError interrupt generated as a result of execution in
EL3 is taken locally instead of a lower exception level.

The SError interrupt is enabled in the TSP code only when the operation has not
been directly initiated by the normal world. This is to prevent the possibility
of an asynchronous external abort which originated in normal world from being
taken when execution is in S-EL1.

Fixes ARM-software/tf-issues#153

Change-Id: I157b996c75996d12fd86d27e98bc73dd8bce6cd5

Move the TSP private declarations out of tsp.h and into a new
header, tsp_private.h. This clarifies the TSP interface to the TSPD.

Change-Id: I39af346eeba3350cadcac56c02d97a5cb978c28b

The purpose of platform_is_primary_cpu() is to determine after reset
(BL1 or BL3-1 with reset handler) if the current CPU must follow the
cold boot path (primary CPU), or wait in a safe state (secondary CPU)
until the primary CPU has finished the system initialization.

This patch removes redundant calls to platform_is_primary_cpu() in
subsequent bootloader entrypoints since the reset handler already
guarantees that code is executed exclusively on the primary CPU.

Additionally, this patch removes the weak definition of
platform_is_primary_cpu(), so the implementation of this function
becomes mandatory. Removing the weak symbol avoids other
bootloaders accidentally picking up an invalid definition in case the
porting layer makes the real function available only to BL1.

The define PRIMARY_CPU is no longer mandatory in the platform porting
because platform_is_primary_cpu() hides the implementation details
(for instance, there may be platforms that report the primary CPU in
a system register). The primary CPU definition in FVP has been moved
to fvp_def.h.

The porting guide has been updated accordingly.

Fixes ARM-software/tf-issues#219

Change-Id: If675a1de8e8d25122b7fef147cb238d939f90b5e

This patch reworks the manner in which the M,A, C, SA, I, WXN & EE bits of
SCTLR_EL3 & SCTLR_EL1 are managed. The EE bit is cleared immediately after reset
in EL3. The I, A and SA bits are set next in EL3 and immediately upon entry in
S-EL1. These bits are no longer managed in the blX_arch_setup() functions. They
do not have to be saved and restored either. The M, WXN and optionally the C
bit are set in the enable_mmu_elX() function. This is done during both the warm
and cold boot paths.

Fixes ARM-software/tf-issues#226

Change-Id: Ie894d1a07b8697c116960d858cd138c50bc7a069

This patch uses stacks allocated in normal memory to enable the MMU early in the
warm boot path thus removing the dependency on stacks allocated in coherent
memory. Necessary cache and stack maintenance is performed when a cpu is being
powered down and up. This avoids any coherency issues that can arise from
reading speculatively fetched stale stack memory from another CPUs cache. These
changes affect the warm boot path in both BL3-1 and BL3-2.

The EL3 system registers responsible for preserving the MMU state are not saved
and restored any longer. Static values are used to program these system
registers when a cpu is powered on or resumed from suspend.

Change-Id: I8357e2eb5eb6c5f448492c5094b82b8927603784

This patch reworks the cold boot path across the BL1, BL2, BL3-1 and BL3-2 boot
loader stages to not use stacks allocated in coherent memory for early platform
setup and enabling the MMU. Stacks allocated in normal memory are used instead.

Attributes for stack memory change from nGnRnE when the MMU is disabled to
Normal WBWA Inner-shareable when the MMU and data cache are enabled. It is
possible for the CPU to read stale stack memory after the MMU is enabled from
another CPUs cache. Hence, it is unsafe to turn on the MMU and data cache while
using normal stacks when multiple CPUs are a part of the same coherency
domain. It is safe to do so in the cold boot path as only the primary cpu
executes it. The secondary cpus are in a quiescent state.

This patch does not remove the allocation of coherent stack memory. That is done
in a subsequent patch.

Change-Id: I12c80b7c7ab23506d425c5b3a8a7de693498f830

Previously, the enable_mmu_elX() functions were implicitly part of
the platform porting layer since they were included by generic
code. These functions have been placed behind 2 new platform
functions, bl31_plat_enable_mmu() and bl32_plat_enable_mmu().
These are weakly defined so that they can be optionally overridden
by platform ports.

Also, the enable_mmu_elX() functions have been moved to
lib/aarch64/xlat_tables.c for optional re-use by platform ports.
These functions are tightly coupled with the translation table
initialization code.

Fixes ARM-software/tf-issues#152

Change-Id: I0a2251ce76acfa3c27541f832a9efaa49135cc1c

The TSP has a number of entrypoints used by the TSP on different
occasions. These were provided to the TSPD as a table of function
pointers, and required the TSPD to read the entry in the table,
which is in TSP memory, in order to program the exception return
address.

Ideally, the TSPD has no access to the TSP memory.

This patch changes the table of function pointers into a vector
table of single instruction entrypoints. This allows the TSPD to
calculate the entrypoint address instead of read it.

Fixes ARM-software/tf-issues#160

Change-Id: Iec6e055d537ade78a45799fbc6f43765a4725ad3