[PATCH v2] target/hppa: Generate illegal instruction exception for 64-bit instructions

Helge Deller posted 1 patch 1 year, 7 months ago
Failed in applying to current master (apply log)
[PATCH v2] target/hppa: Generate illegal instruction exception for 64-bit instructions
Posted by Helge Deller 1 year, 7 months ago
Qemu currently emulates a 32-bit CPU only, and crashes with this error
when it faces a 64-bit load (e.g.  "ldd 0(r26),r0") or a 64-bit store
(e.g. "std r26,0(r26)") instruction in the guest:

ERROR:../qemu/tcg/tcg-op.c:2822:tcg_canonicalize_memop: code should not be reached

Add checks for 64-bit sizes and generate an illegal instruction
exception if necessary.

Signed-off-by: Helge Deller <deller@gmx.de>
---
v2: Move checks to trans_ld() and trans_st() as suggested by Richard

diff --git a/target/hppa/translate.c b/target/hppa/translate.c
index 8b861957e0..a32036c4ce 100644
--- a/target/hppa/translate.c
+++ b/target/hppa/translate.c
@@ -2899,14 +2899,22 @@ static bool trans_cmpiclr(DisasContext *ctx, arg_rri_cf *a)

 static bool trans_ld(DisasContext *ctx, arg_ldst *a)
 {
-    return do_load(ctx, a->t, a->b, a->x, a->scale ? a->size : 0,
+    if (unlikely(TARGET_REGISTER_BITS == 32 && a->size > MO_32)) {
+        return gen_illegal(ctx);
+    } else {
+        return do_load(ctx, a->t, a->b, a->x, a->scale ? a->size : 0,
                    a->disp, a->sp, a->m, a->size | MO_TE);
+    }
 }

 static bool trans_st(DisasContext *ctx, arg_ldst *a)
 {
     assert(a->x == 0 && a->scale == 0);
-    return do_store(ctx, a->t, a->b, a->disp, a->sp, a->m, a->size | MO_TE);
+    if (unlikely(TARGET_REGISTER_BITS == 32 && a->size > MO_32)) {
+        return gen_illegal(ctx);
+    } else {
+        return do_store(ctx, a->t, a->b, a->disp, a->sp, a->m, a->size | MO_TE);
+    }
 }

 static bool trans_ldc(DisasContext *ctx, arg_ldst *a)
Re: [PATCH v2] target/hppa: Generate illegal instruction exception for 64-bit instructions
Posted by Richard Henderson 1 year, 7 months ago
On 9/28/22 11:49, Helge Deller wrote:
> Qemu currently emulates a 32-bit CPU only, and crashes with this error
> when it faces a 64-bit load (e.g.  "ldd 0(r26),r0") or a 64-bit store
> (e.g. "std r26,0(r26)") instruction in the guest:
> 
> ERROR:../qemu/tcg/tcg-op.c:2822:tcg_canonicalize_memop: code should not be reached
> 
> Add checks for 64-bit sizes and generate an illegal instruction
> exception if necessary.
> 
> Signed-off-by: Helge Deller <deller@gmx.de>
> ---
> v2: Move checks to trans_ld() and trans_st() as suggested by Richard


Reviewed-by: Richard Henderson <richard.henderson@linaro.org>


r~