Djam/arm-trusted-firmware
1
0
Fork 0
mirror of https://github.com/ARM-software/arm-trusted-firmware.git synced 2025-04-16 09:34:18 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

refactor(qemu): change way how we enable cpu features

Browse source 
We have to handle wide selection of cpu cores in one TF-A binary:
- v8.0: a53, a57, a72
- v8.2: a55, a76, n1
- v8.4: v1
- v9.0: a710, n2

And then we have QEMU's hybrid: 'max' which has everything QEMU can
emulate.

TF-A for QEMU platforms was built for v8.5 architecture. But turned out
that 'max' has v8.7 flag now (HCX) which we need to have. And this
enabled set of mandatory features which made TF-A not-bootable on
v8.0/8.2 cpus.

So I decided to follow Arm FVP way and do build for v8.0 with set of
feature flags enabled. This way we have bare minimum to make v8.0 cpus
boot. And then all features from newer cores are enabled with runtime
check which makes them boot.

Tested with BSA/SBSA ACS and Debian Linux 6.5 kernel.

Signed-off-by: Marcin Juszkiewicz <marcin.juszkiewicz@linaro.org>
Change-Id: Ib87bdab992536c65ce0747ce1520682eafc18d39
Signed-off-by: Marcin Juszkiewicz <marcin.juszkiewicz@linaro.org>
This commit is contained in:
Marcin Juszkiewicz 2023-11-13 16:18:49 +01:00
parent e7781c84e9
commit a97f4665d1
1 changed files with 38 additions and 19 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

57
plat/qemu/common/common.mk
View file

@ -29,18 +29,6 @@ QEMU_CPU_LIBS := lib/cpus/aarch64/aem_generic.S \
lib/cpus/aarch64/qemu_max.S
PLAT_INCLUDES += -Iinclude/plat/arm/common/${ARCH}
# Cpu core architecture level:
# v8.0: a53, a57, a72
# v8.2: a55, a76, n1
# v8.4: v1
# v9.0: a710, n2
#
# let treat v9.0 as v8.5 as they share cpu features
# https://developer.arm.com/documentation/102378/0201/Armv8-x-and-Armv9-x-extensions-and-features
ARM_ARCH_MAJOR := 8
ARM_ARCH_MINOR := 5
endif
PLAT_BL_COMMON_SOURCES := ${PLAT_QEMU_COMMON_PATH}/qemu_common.c \
@ -91,7 +79,44 @@ BL31_SOURCES += ${QEMU_CPU_LIBS} \
# CPU flag enablement
ifeq (${ARCH},aarch64)
# Later QEMU versions support SME and SVE.
# Cpu core architecture level:
# v8.0: a53, a57, a72
# v8.2: a55, a76, n1
# v8.4: v1
# v9.0: a710, n2
#
#
# We go v8.0 by default and will enable all features we want
ARM_ARCH_MAJOR := 8
ARM_ARCH_MINOR := 0
# 8.0
ENABLE_FEAT_CSV2_2 := 2
# 8.1
ENABLE_FEAT_PAN := 2
ENABLE_FEAT_VHE := 2
# 8.2
# TF-A currently does not permit dynamic detection of FEAT_RAS
# so this is the only safe setting
ENABLE_FEAT_RAS := 0
# 8.4
ENABLE_FEAT_SEL2 := 2
ENABLE_FEAT_DIT := 2
# 8.5
ENABLE_FEAT_RNG := 2
ENABLE_FEAT_SB := 2
# 8.6
ENABLE_FEAT_FGT := 2
# 8.7
ENABLE_FEAT_HCX := 2
# SPM_MM is not compatible with ENABLE_SVE_FOR_NS (build breaks)
ifeq (${SPM_MM},1)
ENABLE_SVE_FOR_NS := 0
@ -101,12 +126,6 @@ else
ENABLE_SME_FOR_NS := 2
endif
# QEMU will use the RNDR instruction for the stack protector canary.
ENABLE_FEAT_RNG := 2
# QEMU 7.2+ has support for FGT and Linux needs it enabled to boot on max
ENABLE_FEAT_FGT := 2
# Treating this as a memory-constrained port for now
USE_COHERENT_MEM := 0