On 2023/6/12 17:47, Christoph Müllner wrote:
On Mon, Jun 12, 2023 at 8:25 AM LIU Zhiwei <baxiantai@gmail.com> wrote:
On 2023/5/30 21:18, Christoph Muellner wrote:
From: Christoph Müllner <christoph.muellner@vrull.eu>
The disassembler needs the available extensions in order
to properly decode instructions in case of overlapping
encodings (e.g. for vendor extensions).
Let's use the field 'disassemble_info::private_data' to store
our RISCVCPUConfig pointer.
Signed-off-by: Christoph Müllner <christoph.muellner@vrull.eu>
---
target/riscv/cpu.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/target/riscv/cpu.c b/target/riscv/cpu.c
index 5b7818dbd1..6f0cd9a0bb 100644
--- a/target/riscv/cpu.c
+++ b/target/riscv/cpu.c
@@ -819,6 +819,9 @@ static void riscv_cpu_disas_set_info(CPUState *s,
disassemble_info *info)
{
RISCVCPU *cpu = RISCV_CPU(s);
CPURISCVState *env = &cpu->env;
+ RISCVCPUConfig *cfg = &cpu->cfg;
+
+ info->private_data = cfg;
I don't know if this field will be overridden by the binutils. Can we
extend the struct disassemble_info, and add some fields like supporting
for Capstone?
Initially I wanted to add a new field, but then I noticed that the field
'disassemble_info::private_data' is used for a similar purpose by
disas/cris.c, disas/m68k.c, and dias/xtensa.c.
So I decided to not add yet another field, which only serves one architecture.
I think you can CC these arch maintainers to see if it need some
specially process before using the private_data.
But if that's the preferred way, then I can change.
I prefer this way, but not insist on if it really works using the
private_data.