project/bcm63xx/atf.git
6 years agoMerge pull request #1460 from robertovargas-arm/clang
Dimitris Papastamos [Wed, 11 Jul 2018 10:13:36 +0000 (11:13 +0100)]
Merge pull request #1460 from robertovargas-arm/clang

Make TF compatible with Clang assembler and linker

6 years agoUpdate documentation about how to use clang toolchain
Roberto Vargas [Mon, 23 Apr 2018 07:38:12 +0000 (08:38 +0100)]
Update documentation about how to use clang toolchain

Change-Id: Ie65eb779b048940cf32ed5744ff40610b3c5499d
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoUse clang as preprocessor when clang toolchain is selected
Roberto Vargas [Thu, 10 May 2018 12:38:19 +0000 (13:38 +0100)]
Use clang as preprocessor when clang toolchain is selected

Change-Id: I562c5de91e12fe384245df41225dfb9122a13a85
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoUse clang assembler when clang compiler is used
Roberto Vargas [Mon, 16 Apr 2018 10:29:45 +0000 (11:29 +0100)]
Use clang assembler when clang compiler is used

Change-Id: Ib90b767e46360ef07c1f22526e3f3eb5fe366d5d
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoRemove integrity check in declare_cpu_ops_base
Roberto Vargas [Fri, 4 May 2018 09:54:33 +0000 (10:54 +0100)]
Remove integrity check in declare_cpu_ops_base

This check was added to ensure the correct behaviour of fill_constants
macro. This macro has been verified and it is known his correct
behaviour. The check generates an error when the clang assembler is
used, so it is better to remove the check.

Change-Id: I3447ff9e9e5ee5cf0502f65e53c3d105d9396b8b
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoRemove .struct directive
Roberto Vargas [Tue, 1 May 2018 08:54:54 +0000 (09:54 +0100)]
Remove .struct directive

This directive is not implemented by clang assembler. The traditional
way to implement structs in assembly is using two macros for every field,
one for the offset, and another one for the size. For every field, the
offset can be calculated using the size and offset of the previous field.

Change-Id: Iacc6781e8f302fb925898737b8e85ab4e88a51cc
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoRemove string concatenation in assembler files
Roberto Vargas [Mon, 30 Apr 2018 16:58:58 +0000 (17:58 +0100)]
Remove string concatenation in assembler files

Clang assembler doesn't support concatenation of adjacent strings.

Change-Id: I092a1e30c137d431c6adcc13519fc2d595f92f14
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoAdd end_vector_entry assembler macro
Roberto Vargas [Tue, 17 Apr 2018 10:31:43 +0000 (11:31 +0100)]
Add end_vector_entry assembler macro

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>
6 years agoRemove .func and .endfunc assembler directives
Roberto Vargas [Tue, 17 Apr 2018 10:31:43 +0000 (11:31 +0100)]
Remove .func and .endfunc assembler directives

These directives are only used when stabs debugging information
is used, but we use ELF which uses DWARF debugging information.
Clang assembler doesn't support these directives, and removing
them makes the code more compatible with clang.

Change-Id: I2803f22ebd24c0fe248e04ef1b17de9cec5f89c4
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoAllow overriding the linker
Roberto Vargas [Fri, 13 Apr 2018 13:26:47 +0000 (14:26 +0100)]
Allow overriding the linker

This patch allows the user to override the value of the LD Makefile
variable. This feature can be used to force the use of the new Clang
linker.

Change-Id: I97ffeb18e48fa75346702a479d7dc1e8abcb3621
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoAdd .extab and .exidx sections
Roberto Vargas [Thu, 10 May 2018 10:01:16 +0000 (11:01 +0100)]
Add .extab and .exidx sections

These sections are required by clang when the code is compiled for
aarch32. These sections are related to the unwind of the stack in
exceptions, but in the way that clang defines and uses them, the
garbage collector cannot get rid of them.

Change-Id: I085efc0cf77eae961d522472f72c4b5bad2237ab
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoUse ALIGN instead of NEXT in linker scripts
Roberto Vargas [Wed, 11 Apr 2018 10:53:31 +0000 (11:53 +0100)]
Use ALIGN instead of NEXT in linker scripts

Clang linker doesn't support NEXT. As we are not using the MEMORY command
to define discontinuous memory for the output file in any of the linker
scripts, ALIGN and NEXT are equivalent.

Change-Id: I867ffb9c9a76d4e81c9ca7998280b2edf10efea0
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoAdd documentation about clang version supported
Roberto Vargas [Mon, 16 Apr 2018 14:43:26 +0000 (15:43 +0100)]
Add documentation about clang version supported

The user guide didn't contain any information about the
requirements of the clang version needed by TF, which is
at least 4.0.

Change-Id: I1ea120aca0fb2c0950fbeaf6978c654ec277afde
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoMerge pull request #1469 from satheesbalya-arm/dtc_doc
Dimitris Papastamos [Wed, 11 Jul 2018 08:18:38 +0000 (09:18 +0100)]
Merge pull request #1469 from satheesbalya-arm/dtc_doc

Add device-tree-compiler to the list of packages

6 years agoAdded device-tree-compiler to the list of packages
Sathees Balya [Tue, 10 Jul 2018 13:46:51 +0000 (14:46 +0100)]
Added device-tree-compiler to the list of packages

Change-Id: Ia7800dae52f152b2c3a3b41f1078ab7499d2f4b6
Signed-off-by: Sathees Balya <sathees.balya@arm.com>
6 years agoMerge pull request #1468 from grandpaul/paulliu-fix-cert-create-ftbfs
Dimitris Papastamos [Mon, 9 Jul 2018 12:14:42 +0000 (13:14 +0100)]
Merge pull request #1468 from grandpaul/paulliu-fix-cert-create-ftbfs

tools/cert_create: fix makefile to build build_msg.o by HOSTCC

6 years agotools/cert_create: fix makefile to build build_msg.o by HOSTCC
Ying-Chun Liu (PaulLiu) [Fri, 6 Jul 2018 15:20:08 +0000 (23:20 +0800)]
tools/cert_create: fix makefile to build build_msg.o by HOSTCC

Previously build_msg.o is built by CC. It causes FTBFS when CC
is not equal to HOSTCC.

Signed-off-by: Ying-Chun Liu (PaulLiu) <paulliu@debian.org>
6 years agoMerge pull request #1463 from grandpaul/paulliu-rpi3-tbb0
Dimitris Papastamos [Fri, 6 Jul 2018 12:03:57 +0000 (13:03 +0100)]
Merge pull request #1463 from grandpaul/paulliu-rpi3-tbb0

rpi3: Add support for Trusted Board Boot

6 years agoMerge pull request #1465 from Andre-ARM/allwinner/h6-support
Dimitris Papastamos [Thu, 5 Jul 2018 14:34:19 +0000 (15:34 +0100)]
Merge pull request #1465 from Andre-ARM/allwinner/h6-support

allwinner: Add H6 SoC support

6 years agodocs: rpi3: add description for Trusted Board Boot
Ying-Chun Liu (PaulLiu) [Thu, 5 Jul 2018 06:55:21 +0000 (14:55 +0800)]
docs: rpi3: add description for Trusted Board Boot

Add paragraph for how to enable Trusted Board Boot for rpi3

Tested-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
Signed-off-by: Ying-Chun Liu (PaulLiu) <paul.liu@linaro.org>
6 years agodocs: rpi3: fix the size of BL1
Ying-Chun Liu (PaulLiu) [Tue, 3 Jul 2018 18:32:27 +0000 (02:32 +0800)]
docs: rpi3: fix the size of BL1

For Trusted Board Boot we enlarge the BL1 size from 64k to 128k.

Tested-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
Signed-off-by: Ying-Chun Liu (PaulLiu) <paulliu@debian.org>
6 years agorpi3: Add support for Trusted Board Boot
Ying-Chun Liu (PaulLiu) [Tue, 3 Jul 2018 18:26:48 +0000 (02:26 +0800)]
rpi3: Add support for Trusted Board Boot

This patch adds support for TBB to rpi3. The ROTPK is generated at build
time and is included into BL1/BL2. The key and content certificates are
read over semihosting.

Tested-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
Signed-off-by: Ying-Chun Liu (PaulLiu) <paul.liu@linaro.org>
6 years agoMerge pull request #1461 from antonio-nino-diaz-arm/an/xlat-docs
Dimitris Papastamos [Wed, 4 Jul 2018 14:09:49 +0000 (15:09 +0100)]
Merge pull request #1461 from antonio-nino-diaz-arm/an/xlat-docs

xlat v2: Update documentation

6 years agoMerge pull request #1396 from ruchi393/multiple_fip
Dimitris Papastamos [Wed, 4 Jul 2018 10:46:51 +0000 (11:46 +0100)]
Merge pull request #1396 from ruchi393/multiple_fip

Extend FIP io driver to support multiple FIP devices

6 years agomaintainers: allwinner: add missing link to github page
Andre Przywara [Wed, 4 Jul 2018 10:36:31 +0000 (11:36 +0100)]
maintainers: allwinner: add missing link to github page

A proper link to Samuel's github page was missing. Add this to make
the link actually work.

Change-Id: I68b116935bf045df60b2e9309b5fd58a1c694d47
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoMerge pull request #1462 from sandrine-bailleux-arm/topics/sb/no-unaligned-access
Dimitris Papastamos [Wed, 4 Jul 2018 09:34:54 +0000 (10:34 +0100)]
Merge pull request #1462 from sandrine-bailleux-arm/topics/sb/no-unaligned-access

Fixes related to unaligned accesses

6 years agoallwinner: Add Allwinner H6 SoC support
Andre Przywara [Fri, 8 Dec 2017 01:27:02 +0000 (01:27 +0000)]
allwinner: Add Allwinner H6 SoC support

The H6 is Allwinner's most recent SoC. It shares most peripherals with the
other ARMv8 Allwinner SoCs (A64/H5), but has a completely different memory
map.

Introduce a separate platform target, which includes a different header
file to cater for the address differences. Also add the new build target
to the documentation.

The new ATF platform name is "sun50i_h6".

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoFix incorrect pointer conversion in SMC_UUID_RET()
Sandrine Bailleux [Mon, 2 Jul 2018 11:01:16 +0000 (13:01 +0200)]
Fix incorrect pointer conversion in SMC_UUID_RET()

Casting a pointer to a struct uuid into a pointer to uint32_t may
result in a pointer that is not correctly aligned, which constitutes
an undefined behaviour. In the case of TF, this also generates a data
abort because alignment fault checking is enabled (through the SCTLR.A
bit).

This patch modifies the SMC_UUID_RET() macro to read the uuid
structure without any pointer aliasing. A helper function then
combines every set of 4 bytes into a 32-bit value suitable to be
returned through the x0-x3 registers.

This fixes a violation of MISRA rule 11.3.

Change-Id: I53ee73bb4cb332f4d8286055ceceb6f347caa080
Signed-off-by: Sandrine Bailleux <sandrine.bailleux@arm.com>
6 years agoAArch32: Force compiler to align memory accesses
Sandrine Bailleux [Tue, 3 Jul 2018 07:14:45 +0000 (09:14 +0200)]
AArch32: Force compiler to align memory accesses

Alignment fault checking is always enabled in TF (by setting the
SCTLR.A bit). Thus, all instructions that load or store one or more
registers have an alignment check that the address being accessed is
aligned to the size of the data element(s) being accessed. If this
check fails it causes an Alignment fault, which is taken as a Data
Abort exception.

The compiler needs to be aware that it must not emit load and store
instructions resulting in unaligned accesses. It already is for
AArch64 builds (see commit fa1d37122c "Add -mstrict-align to the gcc
options"), this patch does the same for AArch32 builds.

Change-Id: Ic885796bc6ed0ff392aae2d49f3a13f517e0169f
Signed-off-by: Sandrine Bailleux <sandrine.bailleux@arm.com>
6 years agoxlat v2: Update documentation
Antonio Nino Diaz [Tue, 3 Jul 2018 13:56:31 +0000 (14:56 +0100)]
xlat v2: Update documentation

Update documentation to reflect the current state of the library.

Change-Id: Ic72f90ee322d2ebd6ea7f4296315700d6bc434e6
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
6 years agoMerge pull request #1459 from antonio-nino-diaz-arm/an/xlat-refactor
Dimitris Papastamos [Tue, 3 Jul 2018 13:38:01 +0000 (14:38 +0100)]
Merge pull request #1459 from antonio-nino-diaz-arm/an/xlat-refactor

Refactor of the xlat tables v2 library

6 years agoMerge pull request #1458 from Andre-ARM/allwinner/fixes
Dimitris Papastamos [Tue, 3 Jul 2018 13:36:55 +0000 (14:36 +0100)]
Merge pull request #1458 from Andre-ARM/allwinner/fixes

allwinner: various smaller fixes

6 years agoMerge pull request #1447 from Amit-Radur/bl32_v1
Dimitris Papastamos [Tue, 3 Jul 2018 13:36:46 +0000 (14:36 +0100)]
Merge pull request #1447 from Amit-Radur/bl32_v1

allwinner: Add BL32 (corresponds to Trusted OS) support

6 years agoxlat v2: Make get/set attrs functions less verbose
Antonio Nino Diaz [Mon, 2 Jul 2018 08:26:51 +0000 (09:26 +0100)]
xlat v2: Make get/set attrs functions less verbose

It is useful to have LOG_LEVEL_VERBOSE because it prints the memory map
of each image, but that also means that the change_mem_attributes and
get_mem_attributes functions have verbose prints, and generate a too
long text output that hides other useful information.

As they were mostly there for debug purposes, this patch removes them.

Change-Id: I2986537377d1f78be2b79cc8a6cf230c380bdb55
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
6 years agoxlat v2: Clean debug xlat tables descriptor print
Antonio Nino Diaz [Wed, 27 Jun 2018 13:59:22 +0000 (14:59 +0100)]
xlat v2: Clean debug xlat tables descriptor print

The previous debug output for EL1&0 translation regimes was too verbose,
which makes it hard to read and hides the intent behind the parameters
assigned to each region. This patch simplifies this output and makes the
outputs for EL3 and EL1&0 mostly the same. The difference is that in
EL1&0 it is specified whether the region is exclusively accessible from
EL1 (PRIV) or both EL0 and EL1 (USER).

For example:

    MEM-RW(PRIV)-NOACCESS(USER)-XN(PRIV)-XN(USER)-S
    MEM-RO(PRIV)-NOACCESS(USER)-EXEC(PRIV)-EXEC(USER)-S

After the change, it becomes this:

    MEM-RW-XN-PRIV-S
    MEM-RO-EXEC-PRIV-S

Change-Id: I15f4b99058429d42107fbf89e15f4838a9b559a5
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
6 years agoxlat v2: Split code into separate files
Antonio Nino Diaz [Tue, 3 Jul 2018 10:58:49 +0000 (11:58 +0100)]
xlat v2: Split code into separate files

Instead of having one big file with all the code, it's better to have
a few smaller files that are more manageable:

- xlat_tables_core.c: Code related to the core functionality of the
  library (map and unmap regions, initialize xlat context).
- xlat_tables_context.c: Instantiation of the active image context
  as well as APIs to manipulate it.
- xlat_tables_utils.c: Helper code that isn't part of the core
  functionality (change attributes, debug print messages).

Change-Id: I3ea956fc1afd7473c0bb5e7c6aab3b2e5d88c711
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
6 years agoMerge pull request #1456 from robertovargas-arm/make_cert
Dimitris Papastamos [Mon, 2 Jul 2018 10:39:22 +0000 (11:39 +0100)]
Merge pull request #1456 from robertovargas-arm/make_cert

make_cert: return error when invalid options are used

6 years agoMerge pull request #1452 from glneo/fixups-for-k3
Dimitris Papastamos [Mon, 2 Jul 2018 10:39:06 +0000 (11:39 +0100)]
Merge pull request #1452 from glneo/fixups-for-k3

Couple of fixups for K3 platform

6 years agoExtend FIP io driver to support multiple FIP devices
Ruchika Gupta [Wed, 27 Jun 2018 06:48:22 +0000 (12:18 +0530)]
Extend FIP io driver to support multiple FIP devices

Platform can define MAX_NUM_FIP_DEVICES in platform_def.h
to define the number of FIP io devices.

The FIP driver doesn't support muliple open file.
So only one single file can be open at a time across
multiple FIP devices.

For any FIP device, an image should be loaded
fully before moving on to the next image.

Signed-off-by: Ruchika Gupta <ruchika.gupta@nxp.com>
6 years agoti: k3: common: Enable interrupts before entering standby state
Andrew F. Davis [Mon, 25 Jun 2018 17:36:25 +0000 (12:36 -0500)]
ti: k3: common: Enable interrupts before entering standby state

To wake a core from wfi interrupts must be enabled, in some cases they
may not be and so we can lock up here. Unconditionally enable interrupts
before wfi and then restore interrupt state.

Signed-off-by: Andrew F. Davis <afd@ti.com>
6 years agoti: k3: common: Add root domain node to platform topology
Andrew F. Davis [Mon, 25 Jun 2018 17:10:53 +0000 (12:10 -0500)]
ti: k3: common: Add root domain node to platform topology

Actions may need to be taken by the last core when all clusters
have been shutdown. Add a top level root domain node to coordinate
this between clusters.

Signed-off-by: Andrew F. Davis <afd@ti.com>
6 years agoallwinner: Add Samuel as second maintainer
Andre Przywara [Thu, 28 Jun 2018 22:49:39 +0000 (23:49 +0100)]
allwinner: Add Samuel as second maintainer

As Samuel contributed most of the original code and he has an interest
in that, add him as a second maintainer for the Allwinner port.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: Detect and output current SoC
Andre Przywara [Thu, 21 Jun 2018 23:47:08 +0000 (00:47 +0100)]
allwinner: Detect and output current SoC

So far we already support booting on two different SoCs, and we will
shortly add a third, so add some code to determine the current SoC type.
This can be later used to runtime detect certain properties.

Also print the SoC name to the console, to give valuable debug information.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: Don't map boot ROM
Andre Przywara [Thu, 21 Jun 2018 23:53:44 +0000 (00:53 +0100)]
allwinner: Don't map boot ROM

There is nothing we need from the BootROM area, so we also don't need
to map it in EL3.

Remove the mapping and reduce the number of MMAP regions by one.

Reported-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: Relax PSCI entry point check
Andre Przywara [Thu, 21 Jun 2018 23:48:15 +0000 (00:48 +0100)]
allwinner: Relax PSCI entry point check

The DRAM controller supports up to 4GB of DRAM, and there are actually
boards out there where we can use at least 3GB of this.

Relax the PSCI entry point check, to be not restricted to 2GB of DRAM.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: security: Fix SPC guard
Andre Przywara [Thu, 21 Jun 2018 23:28:31 +0000 (00:28 +0100)]
allwinner: security: Fix SPC guard

The "#ifdef SUNXI_SPC_BASE" guard was meant to allow the build on SoCs
without a Secure Peripherals Controller, so that we skip that part of
the security setup. But in the current position this will trigger a
warning about an unused variable.

Simply move the guard one line up to cover the variable as well.

Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: cpu-ops: avoid PSCI on/off output
Andre Przywara [Fri, 22 Jun 2018 00:33:34 +0000 (01:33 +0100)]
allwinner: cpu-ops: avoid PSCI on/off output

The "INFO" output in sunxi_cpu_ops.c is quite verbose, so make this more
obvious by changing the log level to "VERBOSE" and so avoiding it to
be printed in a normal (even debug) build.

Reported-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: sun50i_a64: remove unneeded VER_REG offset
Andre Przywara [Thu, 21 Jun 2018 23:34:16 +0000 (00:34 +0100)]
allwinner: sun50i_a64: remove unneeded VER_REG offset

The relative VER_REG *offset* is the same across all known SoCs, so we
can define this offset near it's user.

Remove it from the memory map.

Reported-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: doc: add missing paragraphs
Andre Przywara [Thu, 21 Jun 2018 23:33:28 +0000 (00:33 +0100)]
allwinner: doc: add missing paragraphs

Add two empty lines to denote the paragraphs properly and improve
readability.

Reported-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: common: add missing header
Andre Przywara [Thu, 21 Jun 2018 23:32:18 +0000 (00:32 +0100)]
allwinner: common: add missing header

Some code in sunxi_common.c requires symbols defined in sunxi_private.h,
so add the header to that file.
It was included via another header before, but let's make this explicit.

Reported-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Andre Przywara <andre.przywara@arm.com>
6 years agoallwinner: Add BL32 (corresponds to Trusted OS) support
Amit Singh Tomar [Tue, 19 Jun 2018 19:14:50 +0000 (00:44 +0530)]
allwinner: Add BL32 (corresponds to Trusted OS) support

This patch is an attempt to run Trusted OS (OP-TEE OS being one of them) along
side BL31 image.

ATF supports multiple SPD's that can take dispatcher name (opteed for OP-TEE OS)
as an input using the 'SPD=<dispatcher name>' option during bl31 build.

Signed-off-by: Amit Singh Tomar <amittomer25@gmail.com>
6 years agoMerge pull request #1457 from antonio-nino-diaz-arm/an/maintainers
Dimitris Papastamos [Thu, 28 Jun 2018 13:49:33 +0000 (14:49 +0100)]
Merge pull request #1457 from antonio-nino-diaz-arm/an/maintainers

maintainers: Modify format of file

6 years agomaintainers: Modify format of file
Antonio Nino Diaz [Mon, 25 Jun 2018 18:06:49 +0000 (19:06 +0100)]
maintainers: Modify format of file

The previous format was not very practical and hard to parse by
scripts. The current format is easier as it uses more tokens that can be
identified by scripts, while maintaining a reasonably good result when
it is rendered.

Some maintainers maintain more than one subsystem but they were all
part of the same entry. In cases like this the entry has been split
into two to clarify what file belongs to which subsystem.

The list of maintainers of the Trusted Firmware has been updated.

Change-Id: I4be2d527c5171e8d2d86fb49e45e1d9dbcbd2d80
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
6 years agoMerge pull request #1429 from jeenu-arm/mmu-direct
Dimitris Papastamos [Wed, 27 Jun 2018 11:59:47 +0000 (12:59 +0100)]
Merge pull request #1429 from jeenu-arm/mmu-direct

Enable MMU without stack for xlat v2/DynamIQ

6 years agoTSP: Enable cache along with MMU
Jeenu Viswambharan [Fri, 27 Apr 2018 15:28:12 +0000 (16:28 +0100)]
TSP: Enable cache along with MMU

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>
6 years agoDynamIQ: Enable MMU without using stack
Jeenu Viswambharan [Fri, 27 Apr 2018 14:17:03 +0000 (15:17 +0100)]
DynamIQ: Enable MMU without using stack

Having an active stack while enabling MMU has shown coherency problems.
This patch builds on top of translation library changes that introduces
MMU-enabling without using stacks.

Previously, with HW_ASSISTED_COHERENCY, 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.

NOTE: Since this feature depends on using translation table library v2,
disallow using translation table library v1 with HW_ASSISTED_COHERENCY.

Fixes ARM-software/tf-issues#566

Change-Id: Ie55aba0c23ee9c5109eb3454cb8fa45d74f8bbb2
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoxlat v1: Provide direct MMU-enabling stubs
Jeenu Viswambharan [Fri, 27 Apr 2018 14:06:57 +0000 (15:06 +0100)]
xlat v1: Provide direct MMU-enabling stubs

An earlier patch split MMU-enabling function for translation library v2.
Although we don't intend to introduce the exact same functionality for
xlat v1, this patch introduces stubs for directly enabling MMU to
maintain API-compatibility.

Change-Id: Id7d56e124c80af71de999fcda10f1734b50bca97
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoxlat v2: Split MMU setup and enable
Jeenu Viswambharan [Fri, 27 Apr 2018 14:06:57 +0000 (15:06 +0100)]
xlat v2: Split MMU setup and enable

At present, the function provided by the translation library to enable
MMU constructs appropriate values for translation library, and programs
them to the right registers. The construction of initial values,
however, is only required once as both the primary and secondaries
program the same values.

Additionally, the MMU-enabling function is written in C, which means
there's an active stack at the time of enabling MMU. On some systems,
like Arm DynamIQ, having active stack while enabling MMU during warm
boot might lead to coherency problems.

This patch addresses both the above problems by:

  - Splitting the MMU-enabling function into two: one that sets up
    values to be programmed into the registers, and another one that
    takes the pre-computed values and writes to the appropriate
    registers. With this, the primary effectively calls both functions
    to have the MMU enabled, but secondaries only need to call the
    latter.

  - Rewriting the function that enables MMU in assembly so that it
    doesn't use stack.

This patch fixes a bunch of MISRA issues on the way.

Change-Id: I0faca97263a970ffe765f0e731a1417e43fbfc45
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoMerge pull request #1453 from soby-mathew/sm/set_cntfrq_bl1
Dimitris Papastamos [Wed, 27 Jun 2018 08:25:52 +0000 (09:25 +0100)]
Merge pull request #1453 from soby-mathew/sm/set_cntfrq_bl1

ARM platforms: Initialize cntfrq for BL1 Firmware update

6 years agoMerge pull request #1455 from antonio-nino-diaz-arm/an/ep-doc
Dimitris Papastamos [Wed, 27 Jun 2018 08:25:38 +0000 (09:25 +0100)]
Merge pull request #1455 from antonio-nino-diaz-arm/an/ep-doc

Document ep_info attribute flags

6 years agomake_cert: return error when invalid options are used
Roberto Vargas [Wed, 27 Jun 2018 07:23:22 +0000 (08:23 +0100)]
make_cert: return error when invalid options are used

Print_help was used in different contexts and returning no
error in that function was hiding the error when incorrect
options were used.

Change-Id: Ic3f71748be7ff8440c9d54810b986e9f177f4439
Signed-off-by: Roberto Vargas <roberto.vargas@arm.com>
6 years agoDocument ep_info attribute flags
Antonio Nino Diaz [Tue, 26 Jun 2018 20:42:51 +0000 (21:42 +0100)]
Document ep_info attribute flags

The meaning of them wasn't clear in the header file, it was needed to
search the code for the places where they are used.

Add parentheses to macros in the same header.

Change-Id: I700bc0fd75ccfc0a34f439cb3b2692861bb73b9b
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
6 years agoARM platforms: Initialize cntfrq for BL1 Firmware update
Soby Mathew [Mon, 11 Jun 2018 15:40:36 +0000 (16:40 +0100)]
ARM platforms: Initialize cntfrq for BL1 Firmware update

Currenly the CNTFRQ register and system timer is initialized in BL31 for
use by the normal world. During firmware update, the NS-BL1 or NS-BL2U
may need to access the system timer. Hence this patch duplicates the
CNTFRQ and system timer initialization in BL1 as well.

Signed-off-by: Soby Mathew <soby.mathew@arm.com>
Change-Id: I1ede78b4ae64080fb418cb93f3e48b26d7b724dc

6 years agoMerge pull request #1445 from danielboulby-arm/db/DynamicAuthentication
Dimitris Papastamos [Mon, 25 Jun 2018 14:12:47 +0000 (15:12 +0100)]
Merge pull request #1445 from danielboulby-arm/db/DynamicAuthentication

Set FVP DYN_DISABLE_AUTH authentication default to 0

6 years agoSet FVP DYN_DISABLE_AUTH authentication default to 0
Daniel Boulby [Fri, 22 Jun 2018 15:44:57 +0000 (16:44 +0100)]
Set FVP DYN_DISABLE_AUTH authentication default to 0

Set the ability to dynamically disable Trusted Boot Board
authentication to be off by default

Change-Id: Ibd2aa179179f7d9b0e7731c6e450f200a8c67529
Signed-off-by: Daniel Boulby <daniel.boulby@arm.com>
6 years agoMerge pull request #1448 from npoushin/npoushin/sgi575-coverity-fixes
Dimitris Papastamos [Mon, 25 Jun 2018 12:38:29 +0000 (13:38 +0100)]
Merge pull request #1448 from npoushin/npoushin/sgi575-coverity-fixes

npoushin/sgi575 coverity fixes

6 years agoMerge pull request #1446 from masahir0y/uniphier
Dimitris Papastamos [Mon, 25 Jun 2018 08:32:33 +0000 (09:32 +0100)]
Merge pull request #1446 from masahir0y/uniphier

maintainers.rst: separate maintainership of uniphier and synquacer

6 years agoplat/arm: css: Sanitize the input to css_validate_power_state
Nariman Poushin [Tue, 1 May 2018 12:07:47 +0000 (13:07 +0100)]
plat/arm: css: Sanitize the input to css_validate_power_state

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>
6 years agoplat/arm: css: Convert CSS_SYSTEM_PWR_STATE from macro to inline function
Nariman Poushin [Tue, 1 May 2018 08:28:40 +0000 (09:28 +0100)]
plat/arm: css: Convert CSS_SYSTEM_PWR_STATE from macro to inline function

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>
6 years agoallwinner: Add BL32 (corresponds to Trusted OS) support
Amit Singh Tomar [Tue, 19 Jun 2018 19:14:50 +0000 (00:44 +0530)]
allwinner: Add BL32 (corresponds to Trusted OS) support

This patch is an attempt to run Trusted OS (OP-TEE OS being one of them) along
side BL31 image.

ATF supports multiple SPD's that can take dispatcher name (opteed for OP-TEE OS)
as an input using the 'SPD=<dispatcher name>' option during bl31 build.

Signed-off-by: Amit Singh Tomar <amittomer25@gmail.com>
6 years agomaintainers.rst: separate maintainership of uniphier and synquacer
Masahiro Yamada [Fri, 22 Jun 2018 16:09:51 +0000 (01:09 +0900)]
maintainers.rst: separate maintainership of uniphier and synquacer

Socionext has multiple product lines.  The UniPhier is not the only
platform any more.  Correct the maintainership.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
6 years agoMerge pull request #1441 from robertovargas-arm/mem_protect_board
Dimitris Papastamos [Fri, 22 Jun 2018 14:58:11 +0000 (15:58 +0100)]
Merge pull request #1441 from robertovargas-arm/mem_protect_board

Move mem-protect definitions to board specific files

6 years agoMerge pull request #1444 from jts-arm/docs
Dimitris Papastamos [Fri, 22 Jun 2018 14:57:40 +0000 (15:57 +0100)]
Merge pull request #1444 from jts-arm/docs

Fix typos in documentation

6 years agoMerge pull request #1406 from robertovargas-arm/uuid
Dimitris Papastamos [Fri, 22 Jun 2018 14:56:42 +0000 (15:56 +0100)]
Merge pull request #1406 from robertovargas-arm/uuid

Make TF UUID RFC 4122 compliant

6 years agoFix typos in documentation
John Tsichritzis [Wed, 6 Jun 2018 08:38:10 +0000 (09:38 +0100)]
Fix typos in documentation

In the porting guide, there are blue boxes that describe the argument
types and the return type of each function.  A small typo caused some of
these boxes not being properly rendered.

In the user guide, small typos were fixed that caused random text being
unintentionally rendered as bold.  Also, a slight rewording was done in
the section describing the DYN_DISABLE_AUTH flag.

Change-Id: I57303ca609436a82162fa9b42c06b0d5a63da6df
Signed-off-by: John Tsichritzis <john.tsichritzis@arm.com>
6 years agoMerge pull request #1443 from jeenu-arm/sdei-fixes
Dimitris Papastamos [Fri, 22 Jun 2018 11:42:17 +0000 (12:42 +0100)]
Merge pull request #1443 from jeenu-arm/sdei-fixes

SDEI client EL determination fix

6 years agoSDEI: Fix name of internal function
Jeenu Viswambharan [Fri, 22 Jun 2018 11:03:44 +0000 (12:03 +0100)]
SDEI: Fix name of internal function

The function end_sdei_explicit_dispatch() was intended to be
end_sdei_synchronous_dispatch() which does the opposite of
begin_sdei_synchronous_dispatch(). This patch fixes that.

No functional changes.

Change-Id: I141bd91eb342ecf4ddfd05b49513eee4549e7a56
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoSDEI: Fix determining client EL
Jeenu Viswambharan [Fri, 22 Jun 2018 11:00:20 +0000 (12:00 +0100)]
SDEI: Fix determining client EL

commit 2ccfcb2ea555eb86122e7780010cc50fcee08f54 ("SDEI: Determine client
EL from NS context's SCR_EL3") intended to read from SCR_EL3, but
wrongly read from SPSR_EL3 instead. This patch fixes that.

Change-Id: I8ffea39cc98ef59cb8e7a4c6ef4cb12011c58536
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoMerge pull request #1442 from jeenu-arm/sdei-dispatch-fix
Dimitris Papastamos [Fri, 22 Jun 2018 10:38:12 +0000 (11:38 +0100)]
Merge pull request #1442 from jeenu-arm/sdei-dispatch-fix

SDEI: Fix dispatch bug

6 years agoSDEI: Fix dispatch bug
Jeenu Viswambharan [Fri, 22 Jun 2018 10:21:35 +0000 (11:21 +0100)]
SDEI: Fix dispatch bug

The Commit cdb6ac94ecb3c4caa784cd4d7580cf6252146196 introduced a bug
because of which the SDEI dispatcher wrongly panic when an SDEI event
dispatched earlier as a result of interrupt. This patch fixes the check
for a bound interrupt.

Change-Id: If55c8f0422ff781731248bbbc9c1b59fa0d3c4b0
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoMerge pull request #1437 from jeenu-arm/ras-remaining
Dimitris Papastamos [Fri, 22 Jun 2018 08:36:59 +0000 (09:36 +0100)]
Merge pull request #1437 from jeenu-arm/ras-remaining

SDEI dispatch changes to enable RAS use cases

6 years agoMerge pull request #1440 from antonio-nino-diaz-arm/an/xlat-enums
Dimitris Papastamos [Fri, 22 Jun 2018 08:36:45 +0000 (09:36 +0100)]
Merge pull request #1440 from antonio-nino-diaz-arm/an/xlat-enums

xlat: Remove mmap_attr_t enum type

6 years agoMerge pull request #1427 from b49020/integration
Dimitris Papastamos [Fri, 22 Jun 2018 08:36:15 +0000 (09:36 +0100)]
Merge pull request #1427 from b49020/integration

Add support for Socionext Synquacer SC2A11 SoC based Developerbox platform.

6 years agoxlat: Remove mmap_attr_t enum type
Antonio Nino Diaz [Thu, 21 Jun 2018 13:39:16 +0000 (14:39 +0100)]
xlat: Remove mmap_attr_t enum type

The values defined in this type are used in logical operations, which
goes against MISRA Rule 10.1: "Operands shall not be of an inappropriate
essential type".

Now, `unsigned int` is used instead. This also allows us to move the
dynamic mapping bit from 30 to 31. It was an undefined behaviour in the
past because an enum is signed by default, and bit 31 corresponds to the
sign bit. It is undefined behaviour to modify the sign bit. Now, bit 31
is free to use as it was originally meant to be.

mmap_attr_t is now defined as an `unsigned int` for backwards
compatibility.

Change-Id: I6b31218c14b9c7fdabebe432de7fae6e90a97f34
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
6 years agoMerge pull request #1436 from antonio-nino-diaz-arm/an/spm-sync
Dimitris Papastamos [Thu, 21 Jun 2018 15:34:42 +0000 (16:34 +0100)]
Merge pull request #1436 from antonio-nino-diaz-arm/an/spm-sync

SPM: Allow entering the SP without needing a SMC

6 years agoArm platforms: Remove common RAS configuration source
Jeenu Viswambharan [Fri, 8 Jun 2018 07:44:36 +0000 (08:44 +0100)]
Arm platforms: Remove common RAS configuration source

The file arm_ras.c intended to provide common platform-specific RAS
configuration for Arm platforms. Because this file has symbol
definitions, it's proving difficult to provide a common definition.

This patch therefore renames and makes the file specific to FVP. Other
platforms shall provide their own configuration in similar fashion.

Change-Id: I766fd238946e3e49cdb659680e1b45f41b237901
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoSDEI: Make dispatches synchronous
Jeenu Viswambharan [Fri, 16 Feb 2018 12:07:48 +0000 (12:07 +0000)]
SDEI: Make dispatches synchronous

SDEI event dispatches currently only sets up the Non-secure context
before returning to the caller. The actual dispatch only happens upon
exiting EL3 next time.

However, for various error handling scenarios, it's beneficial to have
the dispatch happen synchronously. I.e. when receiving SDEI interrupt,
or for a successful sdei_dispatch_event() call, the event handler is
executed; and upon the event completion, dispatcher execution resumes
after the point of dispatch. The jump primitives introduced in the
earlier patch facilitates this feature.

With this patch:

  - SDEI interrupts and calls to sdei_dispatch_event prepares the NS
    context for event dispatch, then sets a jump point, and immediately
    exits EL3. This results in the client handler executing in
    Non-secure.

  - When the SDEI client completes the dispatched event, the SDEI
    dispatcher does a longjmp to the jump pointer created earlier. For
    the caller of the sdei_dispatch_event() in particular, this would
    appear as if call returned successfully.

The dynamic workaround for CVE_2018_3639 is slightly shifted around as
part of related minor refactoring. It doesn't affect the workaround
functionality.

Documentation updated.

NOTE: This breaks the semantics of the explicit dispatch API, and any
exiting usages should be carefully reviewed.

Change-Id: Ib9c876d27ea2af7fb22de49832e55a0da83da3f9
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoBL31: Introduce jump primitives
Jeenu Viswambharan [Fri, 16 Feb 2018 11:54:24 +0000 (11:54 +0000)]
BL31: Introduce jump primitives

This patch introduces setjmp() and ongjmp() primitives to enable
standard setjmp/longjmp style execution. Both APIs parameters take a
pointer to struct jmpbuf type, which hosts CPU registers saved/restored
during jump.

As per the standard usage:

  - setjmp() return 0 when a jump is setup; and a non-zero value when
    returning from jump.

  - The caller of setjmp() must not return, or otherwise update stack
    pointer since.

Change-Id: I4af1d32e490cfa547979631b762b4cba188d0551
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoSDEI: Determine client EL from NS context's SCR_EL3
Jeenu Viswambharan [Thu, 21 Jun 2018 07:47:42 +0000 (08:47 +0100)]
SDEI: Determine client EL from NS context's SCR_EL3

Currently, the dispatcher reads from SCR_EL3 register directly to
determine the EL of SDEI client. This is with the assumption that
SCR_EL3 is not modified throughout. However, with RAS work flows, it's
possible that SCR_EL3 register contains values corresponding to Secure
world, and therefore EL determination can go wrong. To mitigate this,
always read the register from the saved Non-secure context.

Change-Id: Ic85e4021deb18eb58757f676f9a001174998543a
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoSDEI: Allow platforms to define explicit events
Jeenu Viswambharan [Wed, 17 Jan 2018 12:30:11 +0000 (12:30 +0000)]
SDEI: Allow platforms to define explicit events

The current macros only allow to define dynamic and statically-bound
SDEI events. However, there ought be a mechanism to define SDEI events
that are explicitly dispatched; i.e., events that are dispatched as a
result of a previous secure interrupt or other exception

This patch introduces SDEI_EXPLICIT_EVENT() macro to define an explicit
event. They must be placed under private mappings. Only the priority
flags are allowed to be additionally specified.

Documentation updated.

Change-Id: I2e12f5571381195d6234c9dfbd5904608ad41db3
Signed-off-by: Jeenu Viswambharan <jeenu.viswambharan@arm.com>
6 years agoMerge pull request #1439 from dp-arm/dp/bl2_size
Dimitris Papastamos [Thu, 21 Jun 2018 15:03:05 +0000 (16:03 +0100)]
Merge pull request #1439 from dp-arm/dp/bl2_size

fvp: Increase BL2 size for TBBR builds

6 years agofvp: Increase BL2 size for TBBR builds
Dimitris Papastamos [Thu, 21 Jun 2018 14:49:08 +0000 (15:49 +0100)]
fvp: Increase BL2 size for TBBR builds

Change-Id: I67e64bb79cc984ea3263f069e22738a42321c46d
Signed-off-by: Dimitris Papastamos <dimitris.papastamos@arm.com>
6 years agoMerge pull request #1428 from jeenu-arm/mbedtls
Dimitris Papastamos [Thu, 21 Jun 2018 13:35:54 +0000 (14:35 +0100)]
Merge pull request #1428 from jeenu-arm/mbedtls

Move to mbedtls-2.10.0 tag

6 years agoMerge pull request #1434 from soby-mathew/sm/fix_cntfrq
Dimitris Papastamos [Thu, 21 Jun 2018 12:42:45 +0000 (13:42 +0100)]
Merge pull request #1434 from soby-mathew/sm/fix_cntfrq

ARM Platforms: Update CNTFRQ register in CNTCTLBase frame

6 years agoMerge pull request #1435 from antonio-nino-diaz-arm/an/xlat-fix
Dimitris Papastamos [Thu, 21 Jun 2018 12:35:48 +0000 (13:35 +0100)]
Merge pull request #1435 from antonio-nino-diaz-arm/an/xlat-fix

xlat_v2: Fix descriptor debug print

6 years agoMerge pull request #1433 from sivadur/integration
Dimitris Papastamos [Thu, 21 Jun 2018 11:55:27 +0000 (12:55 +0100)]
Merge pull request #1433 from sivadur/integration

xilinx: fix zynqmp build when tsp is enabled

6 years agoxlat_v2: Fix descriptor debug print
Antonio Nino Diaz [Thu, 21 Jun 2018 09:52:44 +0000 (10:52 +0100)]
xlat_v2: Fix descriptor debug print

The XN, PXN and UXN bits are part of the upper attributes, not the
lower attributes.

Change-Id: Ia5e83f06f2a8de88b551f55f1d36d694918ccbc0
Signed-off-by: Antonio Nino Diaz <antonio.ninodiaz@arm.com>
6 years agosynquacer: Add platform makefile and documentation
Sumit Garg [Thu, 21 Jun 2018 05:58:18 +0000 (11:28 +0530)]
synquacer: Add platform makefile and documentation

Add Makefile and plaform definations file.
My thanks to Daniel Thompson and Ard Biesheuvel for the bits and pieces
I've taken from their earlier work regarding build and deploy steps for
Developerbox based on Synquacer SoCs. They deserve much of the credit
for this work although, since I assembled and tested things, any blame
is probably mine.

Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
Tested-by: Daniel Thompson <daniel.thompson@linaro.org>
Tested-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
6 years agosynquacer: Enable PSCI framework
Sumit Garg [Fri, 15 Jun 2018 09:59:02 +0000 (15:29 +0530)]
synquacer: Enable PSCI framework

PSCI framework uses SCPI driver to communicate to SCP firmware for
various power management operations. Following PSCI operations are
supported:
- CPU ON
- CPU OFF
- CPU STANDBY
- SYSTEM RESET
- SYSTEM OFF

Signed-off-by: Masahisa Kojima <masahisa.kojima@linaro.org>
Signed-off-by: Sumit Garg <sumit.garg@linaro.org>
6 years agosynquacer: Retrieve DRAM info from SCP firmware
Ard Biesheuvel [Fri, 15 Jun 2018 09:55:42 +0000 (15:25 +0530)]
synquacer: Retrieve DRAM info from SCP firmware

Retrieve DRAM info from SCP firmware using SCPI driver. Board supports
multiple DRAM slots so its required to fetch DRAM info from SCP firmware
and pass this info to UEFI via non-secure SRAM.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Sumit Garg <sumit.garg@linaro.org>