[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-arm] [PATCH v1 3/3] target/arm: Correct exclusive store cmpxchg me
From: |
Alistair Francis |
Subject: |
[Qemu-arm] [PATCH v1 3/3] target/arm: Correct exclusive store cmpxchg memop mask |
Date: |
Fri, 11 Aug 2017 15:17:41 -0700 |
When we perform the atomic_cmpxchg operation we want to perform the
operation on a pair of 32-bit registers. Previously we were just passing
the register size in which was set to MO_32. This would result in the
high register to be ignored. To fix this issue we hardcode the size to
be 64-bits long when operating on 32-bit pairs.
Signed-off-by: Alistair Francis <address@hidden>
---
This was caught with an internal fuzzy tester. These patches fix the
Xilinx 2.10-rc2 tree. I tested with the fuzzy tester (single CPU) and
Linux boot (4 CPUs) on the Xilinx tree. I don't have a good test case to
run on mainline, but am working with some internal teams to get one.
Also linux-user is fully untested.
All tests were with MTTCG enabled.
target/arm/translate-a64.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/target/arm/translate-a64.c b/target/arm/translate-a64.c
index 245175e2f1..49b4d6918d 100644
--- a/target/arm/translate-a64.c
+++ b/target/arm/translate-a64.c
@@ -1913,7 +1913,7 @@ static void gen_store_exclusive(DisasContext *s, int rd,
int rt, int rt2,
tcg_gen_concat32_i64(val, cpu_exclusive_val, cpu_exclusive_high);
tcg_gen_atomic_cmpxchg_i64(tmp, addr, val, tmp,
get_mem_index(s),
- size | MO_ALIGN | s->be_data);
+ MO_64 | MO_ALIGN | s->be_data);
tcg_gen_setcond_i64(TCG_COND_NE, tmp, tmp, val);
tcg_temp_free_i64(val);
} else if (s->be_data == MO_LE) {
--
2.11.0