Fix the following build error:

 $ make GDBSERVER=y
 ...
 Makefile.legacy:9: *** "You have legacy configuration in your .config! Please check your configuration.".  Stop.

The error was introduced by the upgrade to Buildroot 2021.08 [1].

The deprecated option can be found by running "make menuconfig" in
out-br/ and checking the "Legacy options" submenu. It shows that
"gcc 8.x support removed" is selected. We can also see that
BR2_GCC_VERSION_8_X=y appears just below the comment "Legacy options
removed in 2021.08" in out-br/.config. This options comes from file
br-ext/configs/toolchain-br.

Upgrade the toolchain to 10.x which is available in Buildroot 2021.08
and corresponds to the pre-built version we use by default ("make
toolchains" currently downloads 10.2).

Link: https://github.com/OP-TEE/manifest/commit/e8d2210


Signed-off-by: Jerome Forissier <jerome@forissier.org>
Reviewed-by: Joakim Bech <joakim.bech@linaro.org>

Following error occurs when running make buildroot QEMU_RUST_ENABLE=y:

$ make buildroot QEMU_RUST_ENABLE=y
...
error: linker  not found
  |
  = note: No such file or directory (os error 2)

error: aborting due to previous error

The toolchains path is not add to the PATH. Fix it.

Signed-off-by: Rong Fan <fanrong03@baidu.com>
Acked-by: Joakim Bech <joakim.bech@linaro.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>

GICv3 is newer and more featured version of the interrupt controller.

Signed-off-by: Aleksey Kazantsev <Alexey.Kazantsev@kaspersky.com>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>

Invoke optee_rust/ci/qemu-check.exp in target "check-rust" to
trigger CI test for Rust applications.
Run build and check progress:
$ make CFG_TEE_CORE_LOG_LEVEL=0 OPTEE_RUST_ENABLE=y check-rust

Signed-off-by: Yuan Zhuang <zhuangyuan04@baidu.com>
Reviewed-by: Jerome Forissier <jerome@forissier.org>
Tested-by: Jerome Forissier <jerome@forissier.org>

When upgrading from Buildroot 2021.2 to 2021.08, the following error
occurs:

 $ make buildroot
 ...
   GEN     /home/jerome/work/optee_repo_qemu_v8/out-br/Makefile

 Your PATH contains spaces, TABs, and/or newline (\n) characters.
 This doesn't work. Fix you PATH.
 make[1]: *** [Makefile:23: _all] Error 2
 make[1]: Leaving directory '/home/jerome/work/optee_repo_qemu_v8/out-br'
 make: *** [common.mk:323: buildroot] Error 2

The space comes from the line that appends the toolchain to the PATH in
optee_rust_examples_ext.mk. Fix it.

Signed-off-by: Jerome Forissier <jerome@forissier.org>

Drop ION driver configs from poplar.conf, as since
linaro-swg/linux.git branch optee [1] was rebased onto kernel
v5.12, where ION allocator was removed.

Link: [1] https://github.com/linaro-swg/linux/tree/optee-v5.12-20210628


Acked-by: Joakim Bech <joakim.bech@linaro.org>
Acked-by: Jerome Forissier <jerome@forissier.org>
Signed-off-by: Igor Opaniuk <igor.opaniuk@foundries.io>

In the current version of linux kernel eMMC is enumerated as mmcblk0
instead of mmcblk1 as before, which causes hang while trying to mount
root partition:

[    1.631806] mmc0: new HS200 MMC card at address 0001
[    1.641686] mmcblk0: mmc0:0001 8WPD3R 7.28 GiB
...
[    2.791837] Waiting for root device /dev/mmcblk1p3...

Acked-by: Joakim Bech <joakim.bech@linaro.org>
Acked-by: Jerome Forissier <jerome@forissier.org>
Signed-off-by: Igor Opaniuk <igor.opaniuk@foundries.io>

Current ZynqMP build script is based on Xilinx Petalinux tool, which
requires the installation of a proprietary tool. The following patch
removes this dependency and gets closed to the common build Makefile
used by other platforms.

The build is done using Xilinx out-of-tree repositories for Linux,
U-Boot and ATF, while using upstream OP-TEE and Buildroot repositories.

Finally the only non standard piece is the usage of pre-compiled
firmware images from the Xilinx wiki page.

Signed-off-by: Ibai Erkiaga <ibai.erkiaga-elorza@xilinx.com>
Acked-by: Jerome Forissier <jerome@forissier.org>

Add OPTEE_RUST_ENABLE option to common.mk and define buildroot packages
for OP-TEE Rust examples
Build Rust examples in optee_rust/examples:
$ cd build && make OPTEE_RUST_ENABLE=y CFG_TEE_RAM_VA_SIZE=0x00300000

Signed-off-by: Rong Fan <fanrong03@baidu.com>
Signed-off-by: Yuan Zhuang <zhuangyuan04@baidu.com>
Acked-by: Joakim Bech <joakim.bech@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>
Acked-by: Ruchika Gupta <ruchika.gupta@linaro.org>
Acked-by: Jerome Forissier <jerome@forissier.org>

Xen is built from $(ROOT)/xen and Xen tools are built from
$(ROOT)/buildroot/package/xen. They need to be the same major.minor
version, otherwise the Xen tool commands ('xl ...') are likely to fail
in more or less obscure ways.

Add a build time check and fail early if versions do not match.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Ruchika Gupta <ruchika.gupta@linaro.org>

XEN_GZ is not used so remove it.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Ruchika Gupta <ruchika.gupta@linaro.org>

Since Linux commit dcb3b06d9c34 ("tee: optee: replace might_sleep with
cond_resched") in v5.11, setting CONFIG_PREEMPT=y in kconfigs/qemu.conf
is not needed anymore. It was a workaround for the missing call to
cond_resched() in the TEE driver. At that time, might_sleep() together
with preemption enabled allowed to get rid of some RCU warning messages
when running long operations in secure world such as large key generation
(test case: "xtest -l 1 4007_rsa"). With cond_resched() the preemption
settings do not matter anymore.

As a result of this change, QEMUv8 is unmodified (CONFIG_PREEMPT=y is
set by default in the kernel's arch/arm64/configs/defconfig), but 32-bit
QEMU now has CONFIG_PREEMPT disabled. Both platforms are tested and
4007_rsa runs as expected with no warning.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>

Current linux config file generation does not add the cross compilation
flag, leading potentially to compilation errors. Not sure how a proper config
file is generated with current implementation but issues are observerd
specially when buildling the linux image with a single thread.

This commit ensures that both ARCH and CROSS_COMPILE flags are set when
using the merge_config script.

Signed-off-by: Ibai Erkiaga <ibai.erkiaga-elorza@xilinx.com>
Reviewed-by: Jerome Forissier <jerome@forissier.org>

Use soc_term.py instead of the C version and stop building the soc_term
project.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@linaro.org>

Adds soc_term.py, a direct replacement for the soc_term tool originally
written in C [1]. This Python version is slightly simpler and does not
need to be compiled (obviously). It is therefore reasonable to have it
here in build.git for use on QEMU and QEMUv8. We can later get rid of
the soc_term project in the manifest file of those platforms.

This script is also useful with FVP, although fvp.xml does not use it;
in a custom work flow the 'xterm' command is replaced by a script which
connects the FVP telnet ports to soc_term.py via the socat command.

Link: [1] https://github.com/linaro-swg/soc_term


Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>

Fix issue in ftpm_optee_ext/Config.in where BR2_PACKAGE_FTPM_OPTEE_EXT
both selects and depends on BR2_PACKAGE_OPTEE_OS_EXT. Keep only the
select rule.

The issue was found from a OP-TEE build using GCC-11. The build failed
with the following message:

build/br-ext/package/ftpm_optee_ext/Config.in:1:error: recursive dependency detected!
build/br-ext/package/ftpm_optee_ext/Config.in:1:	symbol BR2_PACKAGE_FTPM_OPTEE_EXT depends on BR2_PACKAGE_OPTEE_OS_EXT
build/br-ext/package/optee_os_ext/Config.in:1:	symbol BR2_PACKAGE_OPTEE_OS_EXT is selected by BR2_PACKAGE_FTPM_OPTEE_EXT

Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

When VIRTFS_AUTOMOUNT=y and/or PSS_AUTOMOUNT=y, QEMU prints the
following warning:

 qemu-system-aarch64: warning: 9p: degraded performance: a reasonable
 high msize should be chosen on client/guest side (chosen msize is <=
 8192). See https://wiki.qemu.org/Documentation/9psetup#msize

 for
 details.

Let's set the buffer size to 64K mainly to silence this warning. The
performance is noticeably increased too although it doesn't really
matters for typical OP-TEE use cases.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>

Signed-off-by: Aleksey Kazantsev <Alexey.Kazantsev@kaspersky.com>
Reviewed-by: Jerome Forissier <jerome@forissier.org>

After trying to use upstream EDK2 on my HiKey board without success,
I reached a point where even 'make recovery' would not make the board
bootable again. The recovery process failed like so:

 $ make recovery
 [...]
 Waiting for device... [35][34][33][32][31][30][29][28][27][26][25][24][23][22][21][20][19][18]
 Sending /home/jerome/work/optee_repo_hikey/build/../l-loader/recovery.bin ...
 Done

 fastboot flash loader /home/jerome/work/optee_repo_hikey/build/../l-loader/l-loader.bin
 < waiting for any device >
 Sending 'loader' (39 KB)                           FAILED (remote: 'invalid partition')
 fastboot: error: Command failed
 make: *** [Makefile:351: recovery] Error 1

The solution to this problem is mentioned in [1] and consists in
running 'fastboot getvar partition-size:ptable' before the fastboot
flash command.

Link: [1] https://github.com/96boards/documentation/issues/751

.
Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Victor Chong <victor.chong@linaro.org>

qemu-check.exp needs the environment variable $XEN_BOOT to be set,
or the following error is displayed:
 $ make check
 ...
 # no such variable
     (read trace on "::env(XEN_BOOT)")
     invoked from within
 "if {$::env(XEN_BOOT) == "y"} {
         info " (Xen Dom0)"
 }"
     (file "/<...>/qemu-check.exp" line 119)

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Ruchika Gupta <ruchika.gupta@linaro.org>

Replaces virt-make-fs with mke2fs since it's more robust on Ubuntu and
also a bit faster.

Fixes build problems like:
virt-make-fs -t ext4
/home/jens/work/repos/qemu_v8_xen/build/../out/bin/xen_files
/home/jens/work/repos/qemu_v8_xen/build/../out/bin/xen.ext4
Image Name:   Root file system
Created:      Wed Jun 30 19:34:06 2021
Image Type:   AArch64 Linux RAMDisk Image (gzip compressed)
Data Size:    31978230 Bytes = 31228.74 KiB = 30.50 MiB
Load Address: 44000000
Entry Point:  44000000
libguestfs: error: tar_in: tar subcommand failed on directory: /: tar:
./rootfs.cpio.gz: Wrote only 6144 of 10240 bytes
tar: Exiting with failure status due to previous errors
make: *** [Makefile:362: xen-create-image] Error 1

Reviewed-by: Jerome Forissier <jerome@forissier.org>
Tested-by: Jerome Forissier <jerome@forissier.org>
Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>

This patch enables Measured Boot on TF-A and builds the TSS tools
and the TPM Kernel Module for the FVP toolkit.

The functionality is disabled by default. To enable it, build with
MEASURED_BOOT=y.

Signed-off-by: Javier Almansa Sobrino <javier.almansasobrino@arm.com>
Reviewed-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

TF-A v2.4 and earlier is used with the SCP binaries provided in the
vexpress-firmware git. With TF-A v2.5 is supposed to use the SCP
binaries released at [1] instead. So update to download the SCP binaries
from [1] instead.

Link [1]: https://downloads.trustedfirmware.org/tf-a/css_scp_2.8.0/juno


Acked-by: Jerome Forissier <jerome@forissier.org>
Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>

make-virt-fs is used to generate xen.ext4 but is passed '-t vfat'
instead of '-t ext4'. Fix that.

In fact '-t vfat' happens to be working in general, presumably because
the actual format of the image is detected at runtime. However, I could
not manage to make a VFAT image work in a Docker container based on
Ubuntu 21.04 and with the following kernel image package installed:

  linux-image-kvm/hirsute-updates,now 5.11.0.1009.9 amd64 [installed]

The error messages are as follows:

root@32b61ad4d7f3:~/optee_repo_qemu_v8/build# LIBGUESTFS_DEBUG=1 LIBGUESTFS_TRACE=1 virt-make-fs -t vfat /root/optee_repo_qemu_v8/build/../out/bin/xen_files /root/optee_repo_qemu_v8/build/../out/bin/xen.ext4
[...]
libguestfs: trace: mount_options "utf8" "/dev/sda" "/"
guestfsd: => mkfs (0x116) took 0.47 secs
guestfsd: <= mount_options (0x4a) request length 68 bytes
commandrvf: stdout=n stderr=y flags=0x0
commandrvf: udevadm --debug settle -E /dev/sda
SELinux enabled state cached to: disabled
No filesystem is currently mounted on /sys/fs/cgroup.
Failed to determine unit we run in, ignoring: No data available
command: mount '-o' 'utf8' '/dev/sda' '/sysroot//'
[   11.673076] squashfs: Unknown parameter 'utf8'
[   11.694908] fuseblk: Unknown parameter 'utf8'
command: mount returned 32
command: mount: stderr:
mount: /sysroot: wrong fs type, bad option, bad superblock on /dev/sda, missing codepage or helper program, or other error.
ocaml_exn: 'mount_options' raised 'Failure' exception
guestfsd: error: mount exited with status 32: mount: /sysroot: wrong fs type, bad option, bad superblock on /dev/sda, missing codepage or helper program, or other error.
guestfsd: => mount_options (0x4a) took 0.40 secs
libguestfs: trace: mount_options = -1 (error)
libguestfs: error: mount_options: mount exited with status 32: mount: /sysroot: wrong fs type, bad option, bad superblock on /dev/sda, missing codepage or helper program, or other error.
libguestfs: trace: close
libguestfs: closing guestfs handle 0x55f423503430 (state 2)
[...]

With '-t ext4', the issue is gone.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>
Acked-by: Ruchika Gupta <ruchika.gupta@linaro.org>

Update broken URLs in the issue and PR templates.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Reviewed-by: Joakim Bech <joakim.bech@linaro.org>
Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>

Adds a helper script to create a Xen DomU from a Dom0 root shell.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Reviewed-by: Ruchika Gupta <ruchika.gupta@linaro.org>

Use security_model mapped-xattr for QEMU_PSS_HOST_DIR.
This allows folders/files below /data/tee to be owned by
any uid/gid of the QEMU context.

Fixes https://github.com/OP-TEE/build/issues/478



Signed-off-by: Christoph Gellner <cgellner@de.adit-jv.com>
Suggested-by: Jerome Forissier <jerome@forissier.org>
Reviewed-by: Jerome Forissier <jerome@forissier.org>

Build script updated to use Petalinux 2020.2 release. Custom recipes
removed in order to use upstream recipes insted for optee and python
devtool dependencies.

Signed-off-by: Ibai Erkiaga <ibai.erkiaga-elorza@xilinx.com>
Acked-by: Michael Grand <michael.grand.mg@gmail.com>
Tested-by: Michael Grand <michael.grand.mg@gmail.com>

The readability check of the host kernel makes sense only when
XEN_BOOT=y, so avoid annoying users that are not concerned by this.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Ruchika Gupta <ruchika.gupta@linaro.org>

Xen tools are required to launch the other DomU's from
Dom0. Enable them in Buildroot. Since Xen is built outside
Buildroot, it is essential to maintain the same version of
Xen built in build system and Xen tools built in Buildroot.

DomU needs a rootfs without the Xen tools. So another target
is introduced to build rootfs for DOMU without Xen tools.

To launch & run DomU:

make XEN_BOOT=y run

From Dom0 prompt:

> cd /mnt/host/qemu_v8/xen
> xl create guest.cfg	# This will create a DomU
To goto DomU's console:
> xl console domu

To run and check xtests from Dom0 and DomU:
make XEN_BOOT=y check

Signed-off-by: Ruchika Gupta <ruchika.gupta@linaro.org>
Reviewed-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

Introduce option XEN_BOOT to add xen in the bootflow.

When built with XEN_BOOT=y, boot flow currently supported is :

BL1 -> BL2 -> BL31 -> OP-TEE -> uboot -> Xen -> Dom0

xtests can be run over Dom0

Signed-off-by: Ruchika Gupta <ruchika.gupta@linaro.org>
Reviewed-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

QEMU supports the virtualization extensions only in emulated
GICv3. So, add an option to build TF-A, OP-TEE with GICv3.
To enable this, use GICV3=y with make.

Signed-off-by: Ruchika Gupta <ruchika.gupta@linaro.org>
Reviewed-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Jens Wiklander <jens.wiklander@linaro.org>

Commit c993ec99 ("qemu_v8.mk: add scripts_gdb kernel build target")
has added "$(MAKE) -C $(LINUX_PATH) $(LINUX_COMMON_FLAGS) scripts_gdb" to the
linux: recipe. Since linux: depends on linux-common which already runs
"$(MAKE) -C $(LINUX_PATH) $(LINUX_COMMON_FLAGS)", and since
$(LINUX_COMMON_FLAGS) contains "Image", the Image target is made twice.
It is more efficient to remove the added line and add scripts_gdb to
LINUX_COMMON_FLAGS instead.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Peter Griffin <peter.griffin@linaro.org>

Add an option to build and use U-Boot instead of edk2 as BL33
in the bootflow. To use this option: make UBOOT=y

Signed-off-by: Ruchika Gupta <ruchika.gupta@linaro.org>
Reviewed-by: Jerome Forissier <jerome@forissier.org>
Tested-by: Jerome Forissier <jerome@forissier.org>
Acked-by: Etienne Carriere <etienne.carriere@linaro.org>
Reviewed-by: Jens Wiklander <jens.wiklander@linaro.org>

Update the toolchain built with the buildroot sdk target used on
non-x86_64 hosts to GCC version 10.2.

Tested-by: Jerome Forissier <jerome@forissier.org> (QEMUv8 64/32, host uname -m: aarch64)
Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>

The needed precompiled toolchains provided by Arm are only available for
x86_64 hosts. For all other host architectures we build it with
Buildroot itself.

Reviewed-by: Jerome Forissier <jerome@forissier.org>
Tested-by: Jerome Forissier <jerome@forissier.org> (QEMUv8 64/64, host uname -m: aarch64)
Signed-off-by: Jens Wiklander <jens.wiklander@linaro.org>

Update the default GCC version to v10.2.

Fixes: https://github.com/OP-TEE/optee_os/issues/4233



Signed-off-by: Joakim Bech <joakim.bech@linaro.org>
Tested-by: Joakim Bech <joakim.bech@linaro.org> (QEMU v7, v8)
Acked-by: Jerome Forissier <jerome@forissier.org>

ION has been deprecated in the upstream Linux kernel since v5.11 [1].
Consequently, it was removed from the linaro-swg/linux branch optee
shortly after the release 3.13.0 of OP-TEE. Remove the corresponding
kernel configuration flags which are now unused.

Link: [1] https://github.com/torvalds/linux/commit/e722a295cf493388dae474745d30e91e1a2ec549


Signed-off-by: Jerome Forissier <jerome@forissier.org>
Reviewed-by: Joakim Bech <joakim.bech@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@linaro.org>

Commit f1ca9916 ("qemu: disable kernel CONFIG_JFFS2_FS") was a
workaround to an issue that does not seem to be reproducible anymore.
That was more than five years ago, and the root cause was never clearly
identified. Many things have changed since.
I did not meet any problem running "make check" with CONFIG_JFFS2_FS=y
on my local development machine (it is enabled by default in the Linux
kernel's arch/arm/configs/qemu_defconfig). Therefore it seems safe to
remove this CONFIG_JFFS2_FS=n.

Signed-off-by: Jerome Forissier <jerome@forissier.org>
Reviewed-by: Joakim Bech <joakim.bech@linaro.org>
Reviewed-by: Etienne Carriere <etienne.carriere@linaro.org>

Explicitly set CFG_TEE_BENCHMARK to either ON or OFF upon br-ext
config switch BR2_PACKAGE_OPTEE_BENCHMARK_EXT and not rely on
the default configuration in component CMake files.

Signed-off-by: Etienne Carriere <etienne.carriere@linaro.org>
Reviewed-by: Jerome Forissier <jerome@forissier.org>