On 7/15/2026 4:07 PM, [email protected] wrote:
From: Frédéric Pétrot <[email protected]>

Lq shares the cmo opcode space, but the cbos must have rd = 0.
Now ensure that they are matched only in that case.
This implies that lq can be recognized as such only when rd != 0.
Update the decoder file accordingly.


Fixes: 9273cda722 ("disas/riscv.c: add 'cbo' insns to disassembler")

Reported-by: Julien Thillard <[email protected]>
Signed-off-by: Frédéric Pétrot <[email protected]>
---

This patch adds style issues that you're fixing with patch 3:

$ ./scripts/checkpatch.pl 0002-disas-riscv.c-Correct-disasm-of-lq.patch
ERROR: suspect code indent for conditional statements (15, 18)
#46: FILE: disas/riscv.c:2899:
+               if ((inst >> 7) & 0b11111) {
+                  op = rv_op_lq;

ERROR: trailing statements should be on next line
#50: FILE: disas/riscv.c:2903:
+                  case 0: op = rv_op_cbo_inval; break;

ERROR: trailing statements should be on next line
#51: FILE: disas/riscv.c:2904:
+                  case 1: op = rv_op_cbo_clean; break;

ERROR: trailing statements should be on next line
#52: FILE: disas/riscv.c:2905:
+                  case 2: op = rv_op_cbo_flush; break;

ERROR: trailing statements should be on next line
#53: FILE: disas/riscv.c:2906:
+                  case 4: op = rv_op_cbo_zero; break;

total: 5 errors, 0 warnings, 39 lines checked


What we want to do is to squash patch 3 into this patch to not have the style
issues being added in the first place.


Assuming patch 3 is squashed into this one:


Reviewed-by: Daniel Henrique Barboza <[email protected]>



  disas/riscv.c              | 19 ++++++++++---------
  target/riscv/insn32.decode |  2 ++
  2 files changed, 12 insertions(+), 9 deletions(-)

diff --git a/disas/riscv.c b/disas/riscv.c
index 490bda8e56..831ff526ad 100644
--- a/disas/riscv.c
+++ b/disas/riscv.c
@@ -2889,21 +2889,22 @@ static void decode_inst_opcode(rv_decode *dec, rv_isa 
isa)
              case 2:
                 /*
                  * 'lq' shares the "(...) 010 ..... 0001111" opcode space
-                * with 'cbo' insns.
+                * with 'cbo' insns, but assumes rd != 0.
                  *
                  * cbo_inval  0000000 00000 ..... 010 00000 0001111
                  * cbo_clean  0000000 00001 ..... 010 00000 0001111
                  * cbo_flush  0000000 00010 ..... 010 00000 0001111
                  * cbo_zero   0000000 00100 ..... 010 00000 0001111
-                *
-                * Anything that doesn't match these will default to 'lq'.
                  */
-               switch (inst >> 20) {
-               case 0: op = rv_op_cbo_inval; break;
-               case 1: op = rv_op_cbo_clean; break;
-               case 2: op = rv_op_cbo_flush; break;
-               case 4: op = rv_op_cbo_zero; break;
-               default: op = rv_op_lq; break;
+               if ((inst >> 7) & 0b11111) {
+                  op = rv_op_lq;
+               } else {
+                  switch (inst >> 20) {
+                  case 0: op = rv_op_cbo_inval; break;
+                  case 1: op = rv_op_cbo_clean; break;
+                  case 2: op = rv_op_cbo_flush; break;
+                  case 4: op = rv_op_cbo_zero; break;
+                  }
                 }
              }
              break;
diff --git a/target/riscv/insn32.decode b/target/riscv/insn32.decode
index 21272fdb50..aa02dae3c9 100644
--- a/target/riscv/insn32.decode
+++ b/target/riscv/insn32.decode
@@ -216,6 +216,8 @@ ldu      ............   ..... 111 ..... 0000011 @i
    ]
# *** RVI128 lq ***
+  # *** Catches an lq with rd = 0, which we disallow
+  illegal  ------------   ----- 010 00000 0001111
    lq       ............   ..... 010 ..... 0001111 @i
  }
  sq       ............   ..... 100 ..... 0100011 @s


Reply via email to