From: Pu Lehui <pulehui@huawei.com>

Optimize cmpxchg instruction with amocas.w and amocas.d
Zacas instructions.

Signed-off-by: Pu Lehui <pulehui@huawei.com>
---
 arch/riscv/net/bpf_jit.h        | 27 +++++++++++++++++++++++++++
 arch/riscv/net/bpf_jit_comp64.c | 18 ++----------------
 2 files changed, 29 insertions(+), 16 deletions(-)

diff --git a/arch/riscv/net/bpf_jit.h b/arch/riscv/net/bpf_jit.h
index 2351fba5d3e7..0790f40b7e9d 100644
--- a/arch/riscv/net/bpf_jit.h
+++ b/arch/riscv/net/bpf_jit.h
@@ -1294,6 +1294,33 @@ static inline void emit_bswap(u8 rd, s32 imm, struct rv_jit_context *ctx)
 	emit_mv(rd, RV_REG_T2, ctx);
 }
 
+static inline void emit_cmpxchg(u8 rd, u8 rs, u8 r0, bool is64, struct rv_jit_context *ctx)
+{
+	int jmp_offset;
+
+	if (rv_ext_enabled(ZACAS)) {
+		emit(is64 ? rvzacas_amocas_d(r0, rs, rd, 1, 1) :
+		     rvzacas_amocas_w(r0, rs, rd, 1, 1), ctx);
+		if (!is64)
+			emit_zextw(r0, r0, ctx);
+		return;
+	}
+
+	if (is64)
+		emit_mv(RV_REG_T2, r0, ctx);
+	else
+		emit_addiw(RV_REG_T2, r0, 0, ctx);
+	emit(is64 ? rv_lr_d(r0, 0, rd, 0, 0) :
+	     rv_lr_w(r0, 0, rd, 0, 0), ctx);
+	jmp_offset = ninsns_rvoff(8);
+	emit(rv_bne(RV_REG_T2, r0, jmp_offset >> 1), ctx);
+	emit(is64 ? rv_sc_d(RV_REG_T3, rs, rd, 0, 1) :
+	     rv_sc_w(RV_REG_T3, rs, rd, 0, 1), ctx);
+	jmp_offset = ninsns_rvoff(-6);
+	emit(rv_bne(RV_REG_T3, 0, jmp_offset >> 1), ctx);
+	emit_fence_rw_rw(ctx);
+}
+
 #endif /* __riscv_xlen == 64 */
 
 void bpf_jit_build_prologue(struct rv_jit_context *ctx, bool is_subprog);
diff --git a/arch/riscv/net/bpf_jit_comp64.c b/arch/riscv/net/bpf_jit_comp64.c
index a6a9fd9193e5..8e813809d305 100644
--- a/arch/riscv/net/bpf_jit_comp64.c
+++ b/arch/riscv/net/bpf_jit_comp64.c
@@ -599,10 +599,9 @@ static int emit_atomic_ld_st(u8 rd, u8 rs, const struct bpf_insn *insn,
 static int emit_atomic_rmw(u8 rd, u8 rs, const struct bpf_insn *insn,
 			   struct rv_jit_context *ctx)
 {
-	u8 r0, code = insn->code;
+	u8 code = insn->code;
 	s16 off = insn->off;
 	s32 imm = insn->imm;
-	int jmp_offset;
 	bool is64;
 
 	if (BPF_SIZE(code) != BPF_W && BPF_SIZE(code) != BPF_DW) {
@@ -673,20 +672,7 @@ static int emit_atomic_rmw(u8 rd, u8 rs, const struct bpf_insn *insn,
 		break;
 	/* r0 = atomic_cmpxchg(dst_reg + off16, r0, src_reg); */
 	case BPF_CMPXCHG:
-		r0 = bpf_to_rv_reg(BPF_REG_0, ctx);
-		if (is64)
-			emit_mv(RV_REG_T2, r0, ctx);
-		else
-			emit_addiw(RV_REG_T2, r0, 0, ctx);
-		emit(is64 ? rv_lr_d(r0, 0, rd, 0, 0) :
-		     rv_lr_w(r0, 0, rd, 0, 0), ctx);
-		jmp_offset = ninsns_rvoff(8);
-		emit(rv_bne(RV_REG_T2, r0, jmp_offset >> 1), ctx);
-		emit(is64 ? rv_sc_d(RV_REG_T3, rs, rd, 0, 1) :
-		     rv_sc_w(RV_REG_T3, rs, rd, 0, 1), ctx);
-		jmp_offset = ninsns_rvoff(-6);
-		emit(rv_bne(RV_REG_T3, 0, jmp_offset >> 1), ctx);
-		emit_fence_rw_rw(ctx);
+		emit_cmpxchg(rd, rs, regmap[BPF_REG_0], is64, ctx);
 		break;
 	default:
 		pr_err_once("bpf-jit: invalid atomic RMW opcode %02x\n", imm);
-- 
2.34.1