diff options
author | Daniel Henrique Barboza <dbarboza@ventanamicro.com> | 2023-04-06 15:03:32 -0300 |
---|---|---|
committer | Alistair Francis <alistair.francis@wdc.com> | 2023-05-05 10:49:50 +1000 |
commit | 1ffa805c9d07bf1059d11a183540a3f07d02ea11 (patch) | |
tree | 9ebf8076aa0363211bf8147d684f80ccfc27b74c /hw/misc | |
parent | 66247edc8b6fb36d6b905babcd795068ea989ad5 (diff) | |
download | qemu-1ffa805c9d07bf1059d11a183540a3f07d02ea11.zip qemu-1ffa805c9d07bf1059d11a183540a3f07d02ea11.tar.gz qemu-1ffa805c9d07bf1059d11a183540a3f07d02ea11.tar.bz2 |
target/riscv: sync env->misa_ext* with cpu->cfg in realize()
When riscv_cpu_realize() starts we're guaranteed to have cpu->cfg.ext_N
properties updated. The same can't be said about env->misa_ext*, since
the user might enable/disable MISA extensions in the command line, and
env->misa_ext* won't caught these changes. The current solution is to
sync everything at the end of validate_set_extensions(), checking every
cpu->cfg.ext_N value to do a set_misa() in the end.
The last change we're making in the MISA cfg flags are in the G
extension logic, enabling IMAFG if cpu->cfg_ext.g is enabled. Otherwise
we're not making any changes in MISA bits ever since realize() starts.
There's no reason to postpone misa_ext updates until the end of the
validation. Let's do it earlier, during realize(), in a new helper
called riscv_cpu_sync_misa_cfg(). If cpu->cfg.ext_g is enabled, do it
again by updating env->misa_ext* directly.
This is a pre-requisite to allow riscv_cpu_validate_set_extensions() to
use riscv_has_ext() instead of cpu->cfg.ext_N to validate the MISA
extensions, which is our end goal here.
Signed-off-by: Daniel Henrique Barboza <dbarboza@ventanamicro.com>
Reviewed-by: Weiwei Li <liweiwei@iscas.ac.cn>
Reviewed-by: Alistair Francis <alistair.francis@wdc.com>
Message-Id: <20230406180351.570807-2-dbarboza@ventanamicro.com>
Signed-off-by: Alistair Francis <alistair.francis@wdc.com>
Diffstat (limited to 'hw/misc')
0 files changed, 0 insertions, 0 deletions